Introduction to Programming using Python

Programming Course for Biologists at the Pasteur Institute

by Katja Schuerer, Corinne Maufrais, Catherine Letondal, Eric Deveaud, and Marie-Agnes Petit

Introduction to Programming using Python [http://www.python.org/]: Programming Course for Biologists at the Pasteur Institute
by Katja Schuerer, Corinne Maufrais, Catherine Letondal, Eric Deveaud, and Marie-Agnes Petit Published February, 21 2007 Copyright 2007 Pasteur Institute [http://www.pasteur.fr/] The objective of this course is to teach programming concepts to biologists. It is thus aimed at people who are not professional computer scientists, but who need a better control of computers for their own research. This programming course is part of a course in informatics for biology [http://www.pasteur.fr/formation/infobio/infobioen.html]. If you are already a programmer, and if you are just looking for an introduction to Python, you can go to this Python course [http://www.pasteur.fr/recherche/unites/sis/formation/python/] (in Bioinformatics). PDF version of this course [support.pdf] This course is still under construction. Comments are welcome. Handouts for practical sessions (still under construction) will be available on request. Contact: [email protected]

Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1. First session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2. Documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3. Why Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.4. Programming Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2. Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.1. Data, values and types of values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.2. Variables or naming values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.3. Variable and keywords, variable syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4. Namespaces or representing variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.5. Reassignment of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3. Statements, expressions and functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.1. Statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2. Sequences or chaining statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.3. Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.4. Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.5. Composition and Evaluation of Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4. Communication with outside . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.1. Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.2. Formatting strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.3. Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 5. Program execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 5.1. Executing code from a le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 5.2. Interpreter and Compiler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 6. Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 6.1. Values as objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 6.2. Working with strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7. Branching and Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 7.1. Conditional execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 7.2. Conditions and Boolean expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 7.3. Logical operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 7.4. Alternative execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 7.5. Chained conditional execution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 7.6. Nested conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 7.7. Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 8. Dening Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 8.1. Dening Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 8.2. Parameters and Arguments or the difference between a function denition and a function call 47 8.3. Functions and namespaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 8.4. Boolean functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 9. Collections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.1. Datatypes for collections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 9.2. Methods, Operators and Functions on Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

10.

11.

12.

13.

14.

15.

16.

9.3. Methods, Operators and Functions on Dictionaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 9.4. What data type for which collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 10.1. Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 10.2. The for loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 10.3. The while loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 10.4. Comparison of for and while loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 10.5. Range and Xrange objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.6. The map function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 10.7. Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 Nested data structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 11.1. Nested data structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 11.2. Identity of objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 11.3. Copying complex data structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 11.4. Modifying nested structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 12.1. Handle les in programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 12.2. Reading data from les . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 12.3. Writing in les . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 12.4. Design problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 12.5. Documentation strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Recursive functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 13.1. Recursive functions denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 13.2. Flow of execution of recursive functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 13.3. Recursive data structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 14.1. General Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 14.2. Python built-in exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 14.3. Raising exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 14.4. Dening exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Modules and packages in Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 15.1. Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 15.1.1. Using modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 15.1.2. Building modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 15.1.3. Where are the modules? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 15.1.4. How does it work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 15.1.5. Running a module from the command line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 15.2. Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 15.2.1. Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 15.3. Getting information on available modules and packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 Scripting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 16.1. Using the system environment: os and sys modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 16.2. Running Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 16.3. Parsing command line options with getopt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 16.4. Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 16.5. Searching for patterns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

16.5.1. Introduction to regular expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16.5.2. Regular expressions in Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16.5.3. Prosite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16.5.4. Searching for patterns and parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17. Object-oriented programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.2. What are objects and classes? An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.2.1. Objects description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.2.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.2.3. Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.2.4. Creating objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.3. Dening classes in Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.4. Combining objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5. Classes and objects in Python: technical aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5.1. Namespaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5.2. Objects lifespan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5.3. Objects equality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5.4. Classes and types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17.5.5. Getting information on classes and instances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18. Object-oriented design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2. Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2.1. Software quality factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2.2. Large scale programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2.3. Modularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2.4. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.2.5. Reusability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.3. Abstract Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.3.1. Denition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.3.2. Information hiding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.3.3. Using special methods within classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.4. Inheritance: sharing code among classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.4.2. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.5. Flexibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.5.1. Summary of mechanisms for exibility in Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.5.2. Manual overloading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18.6. Object-oriented design patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

128 129 133 134 135 135 135 135 135 136 137 138 141 144 144 148 149 150 150 153 153 153 153 153 154 156 156 157 157 159 161 162 162 169 172 172 172 174 185

List of Figures
1.1. History of programming languages(Source) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1. Namespace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2. Reassigning values to variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 4.1. Interpretation of formatting templates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 5.1. Comparison of compiled and interpreted code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 5.2. Execution of byte compiled code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 6.1. String indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 7.1. Flow of execution of a simple condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 7.2. If statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 7.3. Block structure of the if statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 7.4. Flow of execution of an alternative condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 7.5. Multiple alternatives or Chained conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 7.6. Nested conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 7.7. Multiple alternatives without elif . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 8.1. Function denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 8.2. Blocks and indentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 8.3. Stack diagram of function calls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 9.1. Comparison some collection datatypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 10.1. The for loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 10.2. Flow of execution of a while statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 10.3. Structure of the while statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 10.4. Passing functions as arguments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 11.1. Representation of nested lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 11.2. Accessing elements in nested lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 11.3. Representation of a nested dictionary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 11.4. List comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 11.5. Copying nested structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 11.6. Modifying compound objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 12.1. ReBase le format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 12.2. Flowchart of the processing of the sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 13.1. Stack diagram of recursive function calls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 13.2. A phylogenetic tree topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 13.3. Tree representation using a recursive list structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 14.1. Exceptions class hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 15.1. Module namespace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 15.2. Loading specic components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 16.1. Manual parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 16.2. Event-based parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 16.3. Parsing: decorated grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 16.4. Parsing result as a hierarchical document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 16.5. Pattern searching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 16.6. Python regular expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 16.7. Python regular expressions: classes and methods summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

17.1. 17.2. 17.3. 17.4. 17.5. 17.6. 17.7. 17.8. 18.1. 18.2. 18.3. 18.4. 18.5. 18.6. 18.7. 18.8.

A DNA object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Representation showing objects methods as counters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A Protein object. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Protein and DNA objects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Classes and instances namespaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Class attributes in class dictionary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Classes methods and bound methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Types of classes and objects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Components as a language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic binding (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic binding (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . UML diagram for inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Multiple Inheritance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Delegation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A composite tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

135 136 142 143 145 148 148 150 154 158 166 167 168 169 176 180

List of Tables
3.1. Order of operator evaluation (highest to lowest) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.1. String formatting: Conversion characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.2. String formatting: Modiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.3. Type conversion functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 6.1. String methods, operators and builtin functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 6.2. Boolean methods and operators on strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 7.1. Boolean operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 9.1. Sequence types: Operators and Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 9.2. List methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 9.3. Dictionary methods and operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 12.1. File methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 12.2. File modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 18.1. Stack class interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 18.2. Some of the special methods to redene Python operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

List of Examples
5.1. Executing code from a le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 8.1. More complex function denition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 8.2. Function to check whether a character is a valid amino acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 10.1. Translate a cds sequence into its corresponding protein sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 10.2. First example of a while loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 10.3. Translation of a cds sequence using the while statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 11.1. A mixed nested datastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 12.1. Reading from les . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 12.2. Restriction of a DNA sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 14.1. Filename error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 14.2. Raising an exception in case of a wrong DNA character . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 14.3. Raising your own exception in case of a wrong DNA character . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 14.4. Exceptions dened in Biopython . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 15.1. A module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 15.2. Using the Bio.Fasta package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 16.1. Walking subdirectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 16.2. Running a program (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 16.3. Running a program (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 16.4. Running a program (3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 16.5. Getopt example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 16.6. Searching for the occurrence of PS00079 and PS00080 Prosite patterns in the Human Ferroxidase protein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 17.1. DNA, a class for DNA sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 18.1. A Stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 18.2. Stack class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 18.3. Dening operators for the DNA class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 18.4. Inheritance example (1): sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 18.5. Curve class: manual overloading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 18.6. An uppercase sequence class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 18.7. A composite tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

List of Exercises
3.1. Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 5.1. Execute code from a le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 7.1. Chained conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 7.2. Nested condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 10.1. Repetitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 10.2. Write the complete codon usage function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 10.3. Rewrite for as while . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 11.1. Representing complex structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 12.1. Multiple sequences for all enzymes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 15.1. Locating modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 15.2. Bio.SwissProt package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 15.3. Using a class from a module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 15.4. Import from Bio.Clustalw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 16.1. Basename of the current working directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 16.2. Finding les in directories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 18.1. Operators for the DNA class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 18.2. Example of an abstract framework: Enzyme parser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 18.3. An analyzed sequence class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 18.4. A partially editable sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

Chapter 1. Introduction

Chapter 1. Introduction
1.1. First session
Python 2.4.2 (#1, Dec 20 2005, 16:25:40) [GCC 4.0.0 (Apple Computer, Inc. build 5026)] on darwin Type "help", "copyright", "credits" or "license" for more information. >>> 1 + 5 6 >>> 2 * 5 10 >>> 1 / 2 0 Is it the right answer? >>> float(1 / 2) 0.0 >>> 1 / 2.0 0.5 >>> float(1)/2 0.5

>>> aaa aaa >>> len(aaa) 3 What happened? >>> len(aaa) + len(ttt) 6 >>> len(aaa) + len(ttt) + 1 7 >>> aaa + ttt aaattt >>> aaa + 5 Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: cannot concatenate str and int objects Read carefully the error message, and explain it. How to protect you from this kind of problem? >>> type(1) <type int> >>> type(1)

Chapter 1. Introduction

<type str> >>> type(1.0) <type float> You can associate a name to a value: >>> a = 3 >>> a 3 The interpreter displays the value (3) of the variable (a). >>> myVar = one sentence >>> myVar one sentence >>> 1string = one string File "<stdin>", line 1 1string = one string ^ SyntaxError: invalid syntax

Read carefully the error message, and explain it. >>> myvar Traceback (most recent call last): File "<stdin>", line 1, in ? NameError: name myvar is not defined What appended? >>> >>> 2 >>> 10 >>> >>> 10 >>> >>> 10 a = 2 a a * 5 b = a * 5 b a = 1 b

Why hasnt b changed? What is the difference between: >>> b = a * 5 and: >>> b = 5 ?

Chapter 1. Introduction

>>> a = 1 in this case a is a number >>> a + 2 3 >>> a = 1 in this case a is a string >>> a + 1 Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: cannot concatenate str and int objects What do you conclude about the type of a variable?

Some magical stuff, that will be explained later: >>> from string import * We can also perform calculus on strings: >>> codon=atg >>> codon * 3 atgatgatg >>> seq1 = agcgccttgaattcggcaccaggcaaatctcaaggagaagttccggggagaaggtgaaga >>> seq2 = cggggagtggggagttgagtcgcaagatgagcgagcggatgtccactatgagcgataata

How do you concatenate seq1 and seq2 in a single string?

>>> seq = seq1 + seq2 What is the length of the string seq? >>> len(seq) 120 Does the string seq contain the ambiguous n base? >>> n in seq False Does it contain an adenine base? >>> a in seq True >>> seq[1] g Why? Because in computer science, strings are numbered from 0 to string length - 1 so the first character is: >>> seq[0] a

Chapter 1. Introduction

Display the 12th base. >>> seq[11] t Find the index of the last character. >>> len(seq) 120 So, because we know the sequence length, by: >>> seq[119] a we can display the last character

But this is not true for all the sequences we will work on. Find a more generic way to do it. >>> seq[len(seq) - 1] a Python provides a special form to get the characters from the end of a string: >>> seq[-1] a >>> seq[-2] t Find a way to get the first codon from the sequence >>> seq[0] + seq[1] + seq[2] agc Python provides a form to get slices from strings: >>> seq[0:3] agc >>> seq[3:6] gcc How many of each base does this sequence contains? >>> 35 >>> 21 >>> 44 >>> 12 count(seq, a) count(seq, c) count(seq, g) count(seq, t)

Count the percentage of each base on the sequence.

Chapter 1. Introduction

Example for the adenine representation >>> long = len(seq) >>> nb_a = count(seq, a) >>> (nb_a / long) * 100 0 What happened? How 35 bases from 120 could be 0 percent? This is due to the way the numbers are represented inside the computer. >>> float(nb_a) / long * 100 29.166666666666668 Now, let us say that you want to find specific pattern on a DNA sequence: >>> dna = """tgaattctatgaatggactgtccccaaagaagtaggacccactaatgcagatcctgga tccctagctaagatgtattattctgctgtgaattcgatcccactaaagat""" >>> EcoRI = GAATTC >>> BamHI = GGATCC Looking at the sequence you will see that EcoRI is present twice and BamHI just once: tgaattctatgaatggactgtccccaaagaagtaggacccactaatgcagatcctgga ~~~~~~ ~~~ tccctagctaagatgtattattctgctgtgaattcgatcccactaaaga ~~~ ~~~~~~ >>> count(dna, EcoRI) 0 Why ?? >>> atgc == atgc True >>> atgc == gcta False >>> atgc == ATGC False why are atgc and ATGC different? We can change the case of a string: >>> EcoRI = lower(EcoRI) >>> EcoRI gaattc >>> count(dna, EcoRI) 2 >>> find(dna, EcoRI) 1

Chapter 1. Introduction

>>> find(dna, EcoRI, 2) 88 >>> BamHI = lower(BamHI) >>> count(dna, BamHI) 0 Why ? Tip: display the sequence:

>>> dna tgaattctatgaatggactgtccccaaagaagtaggacccactaatgcagatcctgga\ntccctagctaagatgtattattctgctgtgaattcgatcc

What is this \n character? How to remove it?

>>> dna = replace(dna, \n, ) >>> dna tgaattctatgaatggactgtccccaaagaagtaggacccactaatgcagatcctggatccctagctaagatgtattattctgctgtgaattcgatccca >>>find(dna, BamHI) 54 Using the mechanisms we have learnt so far, produce the complement of the dna sequence.

1.2. Documentation 1.3. Why Python

The reasons to use Python as a rst language to learn programming are manyfold. First, there are studies that show that Python is well designed for beginners [Wang2002] and the language has been explicitely designed by its author to be easier to learn [Rossum99]. Next, it is more and more often used in bioinformatics as a generalpurpose programming language, to both build components and applications [Mangalam2002]. Another very important reason is the object-orientation, that is necessary not just for aesthetics but to scale to modern large-scale programming [Booch94][Meyer97]. Finally, a rich library of modules for scripting and network programming are essential for bioinformatics which very often relies on the integration of existing tools.

1.4. Programming Languages

What can computers do for you? Computers can execute tasks very rapidly, but in order to achieve this they need an accurate description of the task. They can handle a greater amount of input data than you can. But they

Chapter 1. Introduction

can not design a strategy to solve problems for you. So if you can not gure out the procedure that solve your problem computers cannot help you. The Computers own language. Computers do not understand any of the natural languages such as English, French or German. Their proper language, also called machine language, is only composed of two symbols 0 and 1, or power on - off. They have a sort of a dictionary containing all valid words of this language. These words are the basic instructions, such as add 1 to some number, are two values the same or copy a byte of memory to another place. The execution of these basic instructions are encoded by hardware components of the processor. Programming languages. Programming languages belongs to the group of formal languages. Some other examples of formal languages are the system of mathematical expressions or the languages chemists use to describe molecules. They have been invented as intermediate abstraction level between humans and computers. Why do not use natural languages as programming languages? Programming languages are design to prevent problems occurring with natural language. Ambiguity Natural languages are full of ambiguities and we need the context of a word in order to choose the appropriate meaning. minute for example is used as a unit of time as a noun, but means tiny as adjective: only the context would distinguish the meaning. Natural languages are full of redundancy helping to solve ambiguity problems and to minimize misunderstandings. When you say We are playing tennis at the moment., at the moment is not really necessary but underlines that it is happening now. Natural languages are full of idioms and metaphors. The most popular in English is probably It rains cats and dogs.. Besides, this can be very complicated even if you speak a foreign language very well.

Redundancy

Literacy

Programming languages are foreign languages for computers. Therefore you need a program that translates your source code into the machine language. Programming languages are voluntarily unambiguous, nearly context free and non-redundant, in order to prevent errors in the translation process. History of programming languages. It is instructive to try to communicate with a computer in its own language. This let you learn a lot about how processors work. However, in order to do this, you will have to manipulate only 0s and 1s. You will need a good memory, but probably you would never try to write a program solving real world problems at this basic level of machine code. Because humans have difculties to understand, analyze and extract informations of sequences of zeros and ones, they have written a language called Assembler that maps the instruction words to synonyms that give an idea of what the instruction does, so for instance 0001 became add. Assembler increased the legibility of the code, but the instruction set remained basic and depended on the hardware of the computer. In order to write algorithms for solving more complex problems, there was a need for machine independent higher level programming languages with a more elaborated instruction set than the low level Assembler. The rst ones were Fortran and C and a lot more have been invented right now. A short history of a subset of programming languages is shown in Figure 1.1.

Chapter 1. Introduction

Figure 1.1. History of programming languages(Source)

1956

Fortran I Lisp Cobol Algol 60

1958

1960

1962

1964

PL/I

1966

1968

Smalltalk Pascal Prolog

1970

1972

1974

Scheme Fortran 77 C (K&R) Smalltalk 80 Ada 83 C++ Common Lisp Eiffel Perl Tcl Python Java Scheme R4RS Fortran 90 Haskell SML Caml ML

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

Ruby OCaml Java 2 (v1.2) C# C++ (ISO)

1996

1998

2000

2002

Chapter 2. Variables

Chapter 2. Variables
2.1. Data, values and types of values
In the rst session we have explored some basic issues about a DNA sequence. The specic DNA sequence atcgat was one of our data. For computer scientists this is also a value. During the program execution values are represented in the memory of the computer. In order to interpret these representations correctly values have a type.

Type
Types are sets of data or values sharing some specic properties and their associated operations. We have modeled the DNA sequence, out of habit, as a string. 1 Strings are one of the basic types that Python can handle. In the gc calculation we have seen two other ones: integers and oats. If you are not sure what sort of data you are working with, you can ask Python about it.
>>> type(atcgat) <type str> >>> type(1) <type int> >>> type(1) <type str>

2.2. Variables or naming values

If you need a value more than once or you need the result of a calculation later, you have to give it a name to remember it. Computer scientists also say binding a value to a name or assign a value to a variable.

Binding
Binding is the process of naming a value.

Variable
Variables are names bound to values. You can also say that a variable is a name that refers to a value.
>>> EcoRI = GAATTC

For Python the model is important because it knows nothing about DNA but it knows a lot about strings.

Chapter 2. Variables

So the variable EcoRI is a name that refers to the string value GAATTC. The construction used to give names to values is called an assignment. Python, as a lot of other programming languages, use the sign = to assign value to variables. The two sides of the sign = can not be interchanged. The left side has always to be a variable and the right side a value or a result of a calculation.

Caution
Do not confuse the usage of = in computer science and mathematics. In mathematics, it represents the equality, whereas in Python it is used to give names. So all the following statements are not valid in Python:
>>> GAATTC = EcoRI SyntaxError: cant assign to literal >>> 1 = 2 SyntaxError: cant assign to literal

We will see later how to compare things in Python (Section 11.2).

2.3. Variable and keywords, variable syntax

Python has some conventions for variable names. You can use any letter, the special characters _ and every number provided you do not start with it. White spaces and signs with special meanings in Python, as + and - are not allowed.

Important
Python variable names are case-sensitive, so EcoRI and ecoRI are not the same variable.
>>> EcoRI = GAATTC >>> ecoRI Traceback (most recent call last): File "<stdin>", line 1, in ? NameError: name ecoRI is not defined >>> ecori Traceback (most recent call last): File "<stdin>", line 1, in ? NameError: name ecori is not defined >>> EcoRI GAATTC

Among the words you can construct with these letters, there are some reserved words for Python and can not be used as variable names. These keywords dene the language rules and have special meanings in Python. Here is the list of all of them:

10

Chapter 2. Variables

and else import raise

assert except in return

break exec is try

class finally lambda while

continue def for from not or yield

del global pass

elif if print

2.4. Namespaces or representing variables

How does Python nd the value referenced by a variable? Python stores bindings in a Namespace.

Namespace
A namespace is a mapping of variable names to their values. You can also think about a namespace as a sort of dictionary containing all dened variable names and the corresponding reference to their values.

Reference
A reference is a sort of pointer to a location in memory. Therefore you do not have to know where exactly your value can be found in memory, Python handles this for you via variables.

Figure 2.1. Namespace

Namespace

EcoRI gc

GAATTC

0.546 105 Memory space

Figure 2.1 shows a representation of some namespace. Values which have not been referenced by a variable, are not accessible to you, because you can not access the memory space directly. So if a result of a calculation is returned, you can use it directly and forget about it after that. Or you can create a variable holding this value and then access this value via the variable as often as you want.
>>> from string import *

11

Chapter 2. Variables

>>> cds = """atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaa tttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaacagctttaccatgcggtg ctgcgccatcaggtggtgtttctacgcgatcaagctattacgccgcagcagcaacgcgcgctggc ccagcgttttggcgaattgcatattcaccctgtttacccgcatgccgaaggggttgacgagatca tcgtgctggatacccataacgataatccgccagataacgacaactggcataccgatgtgacattt attgaaacgccacccgcaggggcgattctggcagctaaagagttaccttcgaccggcggtgatac gctctggaccagcggtattgcggcctatgaggcgctctctgttcccttccgccagctgctgagtg ggctgcgtgcggagcatgatttccgtaaatcgttcccggaatacaaataccgcaaaaccgaggag gaacatcaacgctggcgcgaggcggtcgcgaaaaacccgccgttgctacatccggtggtgcgaac gcatccggtgagcggtaaacaggcgctgtttgtgaatgaaggctttactacgcgaattgttgatg tgagcgagaaagagagcgaagccttgttaagttttttgtttgcccatatcaccaaaccggagttt caggtgcgctggcgctggcaaccaaatgatattgcgatttgggataaccgcgtgacccagcacta tgccaatgccgattacctgccacagcgacggataatgcatcgggcgacgatccttggggataaac cgttttatcgggcggggtaa""".replace("\n","") >>> float(count(cds, G) + count(cds, C))/ len(cds) 0.54460093896713613

Here the result of the gc-calculation is lost.

>>> gc = float(count(cds, G) + count(cds, C))/ len(cds) >>> gc 0.54460093896713613

In this example you can remember the result of the gc calculation, because it is stored in the variable gc.

2.5. Reassignment of variables

It is possible to reassign a new value to an already dened variable. This will destroy the reference to its former value and create a new binding to the new value. This is shown in Figure 2.2.

Figure 2.2. Reassigning values to variables

Namespace

EcoRI gc

GAATTC

0.546 0.45 105 Memory space

12

Chapter 2. Variables

Note
In Python, it is possible to reassign a new value with a different type to a variable. This is called dynamic typing, because the type of the variable is assigned dynamically. Note that this is not the case in all programming languages. Sometimes, as in C, the type of variables is assigned statically and has to be declared before use. This is some way more secure because types of variables can be checked only by examining the source code, whereas that is not possible if variables are dynamically typed.

13

Chapter 2. Variables

14

Chapter 3. Statements, expressions and functions

Chapter 3. Statements, expressions and functions

3.1. Statements
In our rst practical lesson, the rst thing we did, was the invocation of the Python interpreter. During the rst session we entered statements that were read, analyzed and executed by the interpreter.

Statement
Statements are instructions or commands that the Python interpreter can execute. Each statement is read by the interpreter, analyzed and then executed.

3.2. Sequences or chaining statements

Program
A program is a sequence of statements that can by executed by the Python interpreter.

Sequence
Sequencing is a simple programming feature that allows to chain instructions that will be executed one by one from top to bottom. Later we are going to learn more complicated ways to control the ow of a program, such as branching and repetition.

3.3. Functions
Function
Functions are named sequences of statements that execute some task. We have already used functions, such as:
>>> type(GAATTC) <type str> >>> len(cds) 852

For example len is a function that calculates the length of things and we asked here for the length of our DNA sequence cds.

15

Chapter 3. Statements, expressions and functions

Function call
Function calls are statements that execute or call a function. The Python syntax of function calls is the function name followed by a comma separated list of arguments inclosed into parentheses. Even if a function does not take any argument the parentheses are necessary. Differences between function calls and variables. As variable names, function names are stored in a namespace with a reference to their corresponding sequence of statements. When they are called, their name is searched in the namespace and the reference to their sequence of statements is returned. The procedure is the same as for variable names. But unlike them, the following parentheses indicate that the returned value is a sequence of statements that has to be executed. Thats why they are even necessary for functions which are called without arguments.

Arguments of functions
Arguments are values provided to a function when the function is called. We will see more about them soon.

3.4. Operations
Operations and Operators
Operations are basic functions with their own syntax. They have a special Operator (a sign or a word) that is the same as a function name. Unary Operators, operations which take one argument, are followed by their argument, and secondary operators are surrounded by their two arguments. Here are some examples:
>>> GTnnAC + GAATTC GTnnACGAATTC >>> GAATTC * 3 GAATTCGAATTCGAATTC >>> n in GTnnAC 1

This is only a simpler way of writing these functions provided by Python, because humans are in general more familiar with this syntax closely related to the mathematical formal language.

3.5. Composition and Evaluation of Expressions

Composition and Expression
Composition is a way to combine functions. The combination is also called an Expression.

16

Chapter 3. Statements, expressions and functions

We have already used it. Here is the most complex example we have seen so far:
>>> float(count(cds, G) + count(cds, C)) / len(cds) 0.54460093896713613

What is happening when this expression is executed? The rst thing to say is that it is a mixed expression of operations and function calls. Lets start with the function calls. If a function is called with an argument representing a composed expression, this one is executed rst and the result value is passed to the calling function. So the cds variable is evaluated, which returns the value that it refers to. This value is passed to the len function which returns the length of this value. The same happens for the float function. The operation count(cds, G) + count(cds, C) is evaluated rst, and the result is passed as argument to float. Lets continue with the operations. There is a precedence list, shown in Table 3.1, for all operators, which determines what to execute rst if there are no parentheses, otherwise it is the same as for function calls. So, for the operation count(cds, G) + count(cds, C) the two count functions are executed rst on the value of the cds variable and G and C respectively. And the two counts are added. The result value of the addition is then passed as argument to the float function followed by the division of the results of the two functions float and len.

Table 3.1. Order of operator evaluation (highest to lowest)

Operator +x, -x, ~x x ** y x * y, x / y,x % y x + y, x - y x << y, x >> y x & y x | y x < y, x <= y, x > y, x >= y, x == y, x != y, x <> y, x is y, x is not y, x in s, x not in s< not x x and y lambda args: expr Name Unary operators Power (right associative) Multiplication, division, modulo Addition, subtraction Bit shifting Bitwise and Bitwise or Comparison, identity, sequence membership tests

Logical negation Logical and Anonymous function

So, as in mathematics, the innermost function or operation is evaluated rst and the result is passed as argument to the enclosing function or operation. It is important to notice that variables are evaluated and only their values are passed as argument to the function. We will have a closer look at this when we talk about function denitions in Section 8.1.

Exercise 3.1. Composition

Have a look at this example. Can you explain what happens? If you cant please read this section once again.

17

Chapter 3. Statements, expressions and functions

>>> from string import * >>> replace(replace(replace(cds, A, a), T, A), a, T)

18

Chapter 4. Communication with outside

Chapter 4. Communication with outside

4.1. Output
We saw in the previous chapter how to export information outside of the program using the print statement. Lets give a little bit more details of its use here. The print statements can be followed by a variable number of values separated by commas. Without a value print puts only a newline character on the standard output, generally the screen. If values are provided, they are transformed into strings, and then are written in the given order, separated by a space character. The line is terminated by a newline character. You can suppress the nal newline character by adding a comma at the end of the list. The following example illustrates all these possibilities:
#! /usr/local/bin/python from string import * dna = "ATGCAGTGCATAAGTTGAGATTAGAGACCCGACAGTA" gc = float(count(dna, G) + count(dna, C))/ len(dna) print gc print print print print print print print "the gc percentage of dna:", dna, "is:", gc "the gc percentage of dna:", dna " is:", gc "the gc percentage of dna:", dna, "is:", gc

producing the following output:

caroline:~> python print_gc.2.py 0.432432432432 the gc percentage of dna: ATGCAGTGCATAAGTTGAGATTAGAGACCCGACAGTA is: 0.432432432432 the gc percentage of dna: ATGCAGTGCATAAGTTGAGATTAGAGACCCGACAGTA is: 0.432432432432 the gc percentage of dna: ATGCAGTGCATAAGTTGAGATTAGAGACCCGACAGTA is: 0.432432432432 caroline:~>

4.2. Formatting strings

19

Chapter 4. Communication with outside

Important
All data printed on the screen have to be character data. But values can have different types. Therefore they have to be transformed into strings beforehand. This transformation is handled by the print statement. It is possible to control this transformation when a specic format is needed. In the examples above, the oat value of the gc calculation is written with lots of digits following the dot which are not very signicant. The next example shows a more reasonable output:
>>> print "%.3f" % gc 0.432 >>> print "%3.1f %%" % (gc*100) 43.2 % >>> print "the gc percentage of dna: %10s... is: %4.1f %%." % (dna, gc*100) the gc percentage of dna: ATGCAGTGCA... is: 43.2 %

Figure 4.1 shows how to interpret the example above. The % (modulo) operator can be used to format strings. It is preceded by the formatting template and followed by a comma separated list of values enclosed in parentheses. These values replace the formatting place holders in the template string. A place holder starts with a % followed by some modiers and a character indicating the type of the value. There has to be the same number of values and place holders.

20

Chapter 4. Communication with outside

Figure 4.1. Interpretation of formatting templates

percent operator preceeded by the formating string and followed by a tuple of values that will replace the placholders values replacing the format placeholder if there are more than one they have to be separated by commas and enclosed by parentheses

formatting string

>>> print "%3.1f %%" % (gc*100) 43.2 %

%
indicates that a format follows
.

f
letter indicating the type of the value to format

total number of digits

number of digits following the dot

Table 4.1 provides the characters that you can use in the formatting template and Table 4.2 gives the modiers of the formatting character.

Important
Remember that the type of a formatting result is a string and no more the type of the input value.
>>> "%.1f" % (gc*100) 43.2 >>> res = "%.1f" % (gc*100) >>> at = 100 - res Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: unsupported operand type(s) for -: int and str >>> res 43.2 >>>

Table 4.1. String formatting: Conversion characters

21

Chapter 4. Communication with outside

Formatting character d,i o,x f,e,E s r %

Output decimal or long integer octal/hexadecimal integer normal, E notation of oating point numbers strings or any object that has a str() method string, use the repr() function of the object literal %

Example "%d" % 10 "%o" % 10 "%e" % 10.0 "%s" % [1, 2, 3] "%r" % [1, 2, 3]

Result 10 12 1.000000e+01 [1, 2, 3] [1, 2, 3]

Table 4.2. String formatting: Modiers

Modier name in parentheses Action selects the key name in a mapping object left, right alignment zero lled string minimum eld width precision Example "%(num)d %(str)s" % { num:1, str:dna} "%-10s" % "dna" "%04i" % 10 "%10s" % "dna" "%4.2f" % 10.1 Result 1 dna

-,+ 0 number . number

dna_______ 0010 _______dna 10.10

4.3. Input
As you can print results on the screen, you can read data from the keyboard which is the standard input device. Python provides the raw_input function for that, which is used as follows:
>>> nb = raw_input("Enter a number, please:") Enter a number, please:12

The prompt argument is optional and the input has to be terminated by a return.

Important
raw_input always returns a string, even if you entered a number. Therefore you have to convert the string input by yourself into whatever you need. Table 4.3 gives an overview of all possible type conversion function.
>>> nb 12 >>> type(nb) <type str>

22

Chapter 4. Communication with outside

>>> nb = int(nb) >>> nb 12 >>> type(nb) <type int>

Notice that a user can enter whatever he wants. So, the input is probably not what you want, and the type conversion can therefore fail. It is careful to test before converting input strings.
>>> nb = raw_input("Please enter a number:") Please enter a number:toto >>> nb toto >>> int(nb) Traceback (most recent call last): File "<stdin>", line 1, in ? ValueError: invalid literal for int(): toto

The following function controls the input:

def read_number(): while 1: nb = raw_input("Please enter a number:") try: nbconv = int(nb) except: print nb, "is not a number." continue else: break return nb

and produces the following output:

>>> read_number() Please enter a number:toto toto is not a number. Please enter a number:12 12

Table 4.3. Type conversion functions

Function int(x [,base]) long(x [,base]) Description converts x to an integer converts x to a long integer

23

Chapter 4. Communication with outside

float(x) complex(real [,imag]) str(x) repr(x) eval(str) tuple(s) list(s) chr(x) unichr(x) ord(c) hex(x) oct(x)

converts x to a oating-point number creates a complex number converts x to a string representation converts x to an expression string evaluates str and returns an object converts a sequence object to a tuple converts a sequence object to a list converts an integer to a character converts an integer to a Unicode character converts a character to its integer value converts an integer to a hexadecimal string converts an integer to an octal string

24

Chapter 5. Program execution

Chapter 5. Program execution

5.1. Executing code from a le
Until now we have only worked interactively during an interpreter session. But each time we leave our session all denitions made are lost, and we have to re-enter them again in the next session of the interpreter whenever we need them. This is not very convenient. To avoid that, you can put your code in a le and then pass the le to the Python interpreter. Here is an example:

Example 5.1. Executing code from a le

Take the code for the cds translation as example and put it in a le named gc.py:

from string import * cds = """atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaa tttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaacagctttaccatgcggtg ctgcgccatcaggtggtgtttctacgcgatcaagctattacgccgcagcagcaacgcgcgctggc ccagcgttttggcgaattgcatattcaccctgtttacccgcatgccgaaggggttgacgagatca tcgtgctggatacccataacgataatccgccagataacgacaactggcataccgatgtgacattt attgaaacgccacccgcaggggcgattctggcagctaaagagttaccttcgaccggcggtgatac gctctggaccagcggtattgcggcctatgaggcgctctctgttcccttccgccagctgctgagtg ggctgcgtgcggagcatgatttccgtaaatcgttcccggaatacaaataccgcaaaaccgaggag gaacatcaacgctggcgcgaggcggtcgcgaaaaacccgccgttgctacatccggtggtgcgaac gcatccggtgagcggtaaacaggcgctgtttgtgaatgaaggctttactacgcgaattgttgatg tgagcgagaaagagagcgaagccttgttaagttttttgtttgcccatatcaccaaaccggagttt caggtgcgctggcgctggcaaccaaatgatattgcgatttgggataaccgcgtgacccagcacta tgccaatgccgattacctgccacagcgacggataatgcatcgggcgacgatccttggggataaac cgttttatcgggcggggtaa""".replace("\n","") gc = float(count(cds, g) + count(cds, c))/ len(cds) print gc

and now pass this le to the interpreter:

caroline:~/python_cours> python gc.py 0.54460093896713613

Tip
You can name your le as you like. However, there is a convention for les containing python code to have a py extension.

25

Chapter 5. Program execution

You can also make your le executable if you put the following line at the beginning of your le, indicating that this le has to be executed with the Python interpreter:
#! /usr/local/bin/python

(Dont forget to set the x execution bit under UNIX system.) Now you can execute your le:
caroline:~/python_cours> ./gc.py 0.54460093896713613

This will automatically call the Python interpreter and execute all the code in your le. You can also load the code of a le in a interactive interpreter session with the -i option:
caroline:~/python_cours> python -i gc.py 0.54460093896713613 >>>

This will start the interpreter, execute all the code in your le and than give you a Python prompt to continue:

>>> cds atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaatttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaac >>>cds="atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaatttcgggtgccgacctgacgcgcccgtt" >>>cds atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaatttcgggtgccgacctgacgcgcccgtt >>>gc 0.54460093896713613

Important
It is important to remember that the Python interpreter executes code from top to bottom, this is also true for code in a le. So, pay attention to dene things before you use them.

Exercise 5.1. Execute code from a le

Take all expressions that we have written so far and put them in a le.

26

Chapter 5. Program execution

Important
Notice that you have to ask explicitly for printing a result when you execute some code from a le, while an interactive interpreter session the result of the execution of a statement is printed automatically. So to view the result of the translate function in the code above, the print statement is necessary in the le version, whereas during an interactive interpreter session we have never written it.

5.2. Interpreter and Compiler

Lets introduce at this point some concepts of execution of programs written in high level programming languages. As we have already seen, the only language that a computer can understand is the so called machine language. These languages are composed of a set of basic operations whose execution is implemented in the hardware of the processor. We have also seen that high level programming languages provide a machine-independent level of abstraction that is higher than the machine language. Therefore, they are more adapted to a human-machine interaction. But this also implies that there is a sort of translator between the high level programming language and the machine languages. There exists two sorts of translators:

Interpreter
An Interpreter is a program that implements or simulates a virtual machine using the base set of instructions of a programming language as its machine language. You can also think of an Interpreter as a program that implements a library containing the implementation of the basic instruction set of a programming language in machine language. An Interpreter reads the statements of a program, analyzes them and then executes them on the virtual machine or calls the corresponding instructions of the library.

Interactive interpreter session

During an interactive interpreter session the statements are not only read, analyzed and executed but the result of the evaluation of an expression is also printed. This is also called a READ - EVAL - PRINT loop.

Important
Pay attention, the READ - EVAL - PRINT loop is only entered in an interactive session. If you ask the interpreter to execute code in a le, results of expression evaluations are not printed. You have to do this by yourself.

Compiler
A Compiler is a program that translates code of a programming language in machine code, also called object code. The object code can be executed directly on the machine where it was compiled.

27

Chapter 5. Program execution

Figure 5.1 compares the usage of interpreters and compilers.

Figure 5.1. Comparison of compiled and interpreted code

Interpreter
virtual machine

source code

processor

Compiler

So using a compiler separates translation and execution of a program. In contrast of an interpreted program the source code is translated only once. The object code is machine-dependent meaning that the compiled program can only be executed on a machine for which it has been compiled, whereas an interpreted program is not machine-dependent because the machinedependent part is in the interpreter itself. Figure 5.2 illustrates another concept of program execution that tries to combine the advantage of more effective execution of compiled code and the advantage of machine-independence of interpreted code. This concept is used by the JAVA programming language for example and in a more subtle way by Python.

28

Chapter 5. Program execution

Figure 5.2. Execution of byte compiled code

source code
virtual machine

Interpreter

Compiler processor

bytecode

In this case the source code is translated by a compiler in a sort of object code, also called byte code that is then executed by an interpreter implementing a virtual machine using this byte code. The execution of the byte code is faster than the interpretation of the source code, because the major part of the analysis and verication of the source code is done during the compilation step. But the byte code is still machine-independent because the machine-dependent part is implemented in the virtual machine. We will see later how this concept is used in Python (Section 15.1).

29

Chapter 5. Program execution

30

Chapter 6. Strings

Chapter 6. Strings
So far we have seen a lot about strings. Before giving a summary about this data type, let us introduce a new syntax feature.

6.1. Values as objects

We have seen that strings have a value. But Python values are more than that. They are objects.

Object
Objects are things that know more than their values. In particular, you can ask them to perform specialized tasks that only they can do. Up to now we have used some special functions handling string data available to us by the up to now magic statement from string import *. But strings themselves know how to execute all of them and even more. Look at this:
>>> motif = "gaattc" >>> motif.upper() GAATTC >>> motif gaattc >>> motif.isalpha() 1 >>> motif.count(n) 0 >>> motif = GAATTC_ >>> motif + motif GAATTC_GAATTC_ >>> motif * 3 GAATTC_GAATTC_GAATTC_

At the rst glance this looks a little bit strange, but you can read the . (dot) operator as: ask object motif to do something as: transform motif in an uppercase string (upper), ask whether it contains only letters (isalpha) or count the number of n characters. Objects as namespaces. How does it work? All objects have their own namespace containing all variable and function names that are dened for that object. As already describeb in Section 2.4 you can see all names dened for an object by using the dir function:
>>> dir(motif) [__add__, __class__, __contains__, __delattr__, __eq__, __ge__, __getattribute__, __getitem__, __getslice__, __gt__, __hash__, __init__, __le__, __len__, __lt__, __mul__, __ne__, __new__,

31

Chapter 6. Strings

__reduce__, __repr__, __rmul__, __setattr__, __str__, capitalize, center, count, decode, encode, endswith, expandtabs, find, index, isalnum, isalpha, isdigit, islower, isspace, istitle, isupper, join, ljust, lower, lstrip, replace, rfind, rindex, rjust, rstrip, split, splitlines, startswith, strip, swapcase, title, translate, upper]

The dot operator is used to access this namespace. It will look up in the namespace of the object for the name following the dot operator.
>>> motif.__class__ <type str> >>> motif.replace(a, A) gAAttc

Variables and functions dened in object namespaces are called attributes and methods of the object.

Attribute
An attribute is a variable dened in a namespace of an object, which can only be accessed via the object himself.

Method
Methods are functions dened in a namespace of an object. This is just a little introduction to objects making it possible to use the object syntax for the basic types in Python. We will give further explanation into object-oriented programming in Chapter 17.

6.2. Working with strings

Strings
Strings are sequences or ordered collections of characters. You can write them in Python using quotes, double quotes, triple quotes or triple double quotes. The triple quote notation permits to write strings spanning multiple lines with keeping any line feed.
>>> ATGCA ATGCA >>> "ATGCA" ATGCA >>> """ATGATA ... AGAGA"""

32

Chapter 6. Strings

ATGATA\nAGAGA

The rst thing that we would sum up is how to extract characters or substrings . Characters are accessible using their position in the string which has to be enclosed into brackets following the string. The position number can be positive or negative, which means starting at the beginning or the end of the string. Substrings are extracted by providing an interval of the start and end position separated by a colon. Both positions are optional which means either to start at the beginning of the string or to extract the substring until the end. When accessing characters, it is forbidden to access position that does not exist, whereas during substring extraction, the longest possible string is extracted.
>>> motif = GAATTC >>> motif[0] G >>> motif[-1] C >>> motif[0:3] GAA >>> motif[1:3] AA >>> motif[:3] GAA >>> motif[3:] TTC >>> motif[3:6] TTC >>> motif[3:-1] TT >>> motif[-3:-1] TT >>> motif[:] GAATTC >>> motif[100] Traceback (most recent call last): File "<stdin>", line 1, in ? IndexError: string index out of range >>> motif[3:100] TTC >>> motif[3:2]

Caution
Figure 6.1 compares positive and negative indices. Be careful, forward string indices starts always with 0, whereas backward string indices starts with -1.

33

Chapter 6. Strings

Figure 6.1. String indices

0 motif =

G A A T T C
6 5 4 3 2 1 EcoRI[1:4] == EcoRI[5:2]

Caution
It is also important to notice that the character at the end position during a substring extraction is never included in the extracted substring.

Warning
Strings are immutable in Python. This means you can neither change characters or substrings. You have always to create a new copy of the string.
>>> motif[1] = n Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: object doesnt support item assignment >>> motif[:1] + n + motif[2:] GnATTC

A list of all other methods, function and operators and their action on string objects are summarized in Table 6.1 and Table 6.2).

Table 6.1. String methods, operators and builtin functions

Method, Operator, Function s + t s * 3 len(s) min(s), max(s) Description Concatenation Repetition Returns the length of s Returns the smallest, largest character of s, depending on their position in the ASCII code

34

Chapter 6. Strings

s.capitalize() s.center([width]) s.count(sub [,start [, end]]) s.encode([encoding [, errors]]) s.expandtabs([tabsize]) s.find(sub [, start [, end]]) s.index(sub [,start [, end]]) s.join(words) s.ljust(width) s.lower() s.lstrip() s.replace(old, new [, maxrep]) s.rfind(sub [, start [, end]]) s.rindex(sub [,start [, end]])

Capitalize the rst character of s Centers s in a eld of length width Counts occurrences of sub between start and end Encode s using encoding as code and error Expands tabs Finds the rst occurrence of sub between start and end same as find but raise an exception if no occurrence is found Joins the list of words with s as delimiter Left align s in a string of length width Returns a lowercase version of s Removes all leading whitespace characters of s Replace maximal maxrep versions of substring old with substring new Finds the last occurrence of substring sub between start and end Same as rfind but raise an exception if sub does not exists Right-align s in a string of length width Removes trailing whitespace characters Split s into maximal maxsplit words using sep as separator (default whitespace) Split s into lines, if keepends is 1 keep the trailing newline Removes trailing and leading whitespace characters Returns a copy of s with lowercase letters turn into uppercase and vice versa Returns a title-case version of s (all words capitalized) Translate s using translation table table and removing characters in string delchars Returns an uppercase version of s

s.rjust(width) s.rstrip() s.split([sep [, maxsplit]])) s.splitlines([keepends]) s.strip() s.swapcase() s.title() s.translate(table [, delchars]) s.upper()

Table 6.2. Boolean methods and operators on strings

Method or operator s < <=, >=, > t Description Checks if s appears before, before or at the same point, after or at the same point, after than t in an alphabetically sorted dictionary Checks if r appears between s and t in an alphabetically sorted dictionary Checks the identity or difference of s and t Checks if character c appears in s Checks if s ends with sufx Checks whether all characters are alphanumeric

s < <= t >=, > r s ==, !=, is, not is t c in s, c not in s s.endswith(suffix [,start [,end]]) s.isalnum()

35

Chapter 6. Strings

s.isalpha() s.isdigit() s.islower() s.isspace() s.istitle() s.isupper() s.startswith(prefix [, start [, end]]))

Checks whether all characters are alphabetic Checks whether all characters are digits Checks whether all characters are lowercase Checks whether all characters are whitespace Checks whether s is title-case meaning all words are capitalized Checks whether s starts with prex between start and end

36

Chapter 7. Branching and Decisions

Chapter 7. Branching and Decisions

7.1. Conditional execution
Sometimes the continuation of a program depends on a condition. We would either execute a part of the program if this condition is fullled or adapt the behavior of the program depending on the truth value of the condition.

Branching or conditional statements

Branching is a feature provided by programming languages that makes it possible to execute a sequence of statements among several possibilities. The simplest case of a conditional statement can be expressed by the if statement.
>>> from string import * >>> seq = ATGAnnATG >>> if n in seq: ... print "sequence contains undefined bases" ... nb = count(seq, n) sequence contains undefined bases

Figure 7.1 shows a general schema of a simple conditional statement.

Figure 7.1. Flow of execution of a simple condition

if condition : block of statements executed only if condition is true
if block false

true

condition

The if statement has to be followed by a condition. Then a block is opened by a colon. This block contains the sequence of statements that has to be executed if the condition is fullled. Figure 7.3 and Figure 7.2 highlight the structural parts of the if statement.

37

Chapter 7. Branching and Decisions

Figure 7.2. If statement

condition sequence of instructions ONLY executed if the condition is true

>>> if n in seq : ... print "sequence contains undefined bases" nb = count(seq, n) ... sequence contains undefined bases
body of the if statement

Figure 7.3. Block structure of the if statement

header line block inititiation

>>> if n in seq : ... print "sequence contains undefined bases" nb = count(seq, n) ... sequence contains undefined bases
same indentation to indicate a block block of code body of the if statement

7.2. Conditions and Boolean expressions

The condition in the if statement has to be a boolean expression.

Conditions or Boolean expressions

Conditional or boolean expressions are expressions that are either true or false. Here are some examples of simple boolean expressions or conditions:
>>> 1 < 0 False >>> 1 > 0 True >>> 1 == 0 False >>> n in ATGCGTAnAGTA True >>> A > C False >>> AG < AA

38

Chapter 7. Branching and Decisions

False >>> AG != AC True >>> len(ATGACGA) >= 10 False

In Python the value true is represented by 1 and the value false by 0. Table 7.1 lists all boolean operators and their action on strings and numbers.

Table 7.1. Boolean operators

Operator <, <=, >=, > ==, !=, is, is nota in, not in
a

Action on strings Action on numbers alphabetically sorted, lower/lower or lower/lower or equal/greater equal/greater or equal/greater than equal/greater than identity identity membership -

or

We will explain the difference between these two identity operators later (Section 11.2).

Important
Do not confuse the assignment sign = with the logical operator ==. The second one is used to compare two things and check their equality whereas the rst one is used to bound values to variables. Python does not accept an assignment as a condition but there are other programming languages that use the same syntax for these two statements, but they do not warn when you use = instead of ==.

7.3. Logical operators

The three logical operators not, and and or enable you to compose boolean expressions and by this way to construct more complex conditions. Here are some examples:
>>> seq = ATGCnATG >>> n in seq or N in seq True >>> A in seq and C in seq True >>> n not in seq False >>> len(seq) > 100 and n not in seq False >>> not len(seq) > 100 True

39

Chapter 7. Branching and Decisions

Caution
The or operation is not an exclusive or as it is sometimes used in the current languages. An or expression is also true if both subexpressions are true.
>>> seq = ATGCnATG >>> A in seq or C in seq True

7.4. Alternative execution

If the condition of a if statement is not fullled no statement is executed.
>>> seq = ATGACGATAG >>> if n in seq: print "sequence contains undefined characters" >>>

An alternative sequence of statements, that will be executed if the condition is not fullled, can be specied with the else statement.
>>> seq = ATGACGATAG >>> if n in seq: ... print "sequence contains undefined bases" ... else: print "sequence contains only defined bases" sequence contains only defined bases

Here the if and else are followed by a block containing the statements to execute depending on the truth value of the condition. In this case exactly one of them is executed, which is illustrated in Figure 7.4

40

Chapter 7. Branching and Decisions

Figure 7.4. Flow of execution of an alternative condition

if condition: block of statements executed only if condition is true else : block of statements executed only if the condition is false

true

condition

false

if block

else block

7.5. Chained conditional execution

In Python, you can specify more than one alternative:
>>> seq = vATGCAnATG >>> base = seq[0] >>> base v >>> if base in ATGC: ... print "exact nucleotid" ... elif base in bdhkmnrsuvwxy: ... print "ambiguous nucleotid" ... else: ... print "not a nucleotid" ... ambiguous nucleotid

The elif statement is used to give an alternative condition. What happens when this is executed? The conditions are evaluated from top to bottom. In our example with base = v as rst condition, the if condition base in ATGC is false, so the next condition, that of the elif statement is evaluated. base in bdhkmnrsuvwxy is true, so the block of statements of this clause is executed and ambiguous nucleotid is printed. Then the evaluation of the condition is stopped and the ow of execution continues with the statement following the if-elif-else construction.

Multiple alternative conditions

Multiple alternative conditions are conditions that are tested from top to bottom. The clause of statements for the rst alternative that is evaluated as true is executed. So there is exactly one alternative that is executed, even

41

Chapter 7. Branching and Decisions

if there are more than one that are true. In this case the clause of the rst true condition encountered is chosen. Figure 7.5 illustrates this.

Figure 7.5. Multiple alternatives or Chained conditions

true

if condition : block of statements executed only if condition is true elifsecond_condition

condition false

true

:
if block elif block

second condition

false

block of statements executed only if second_condition is true and condition is false else : block of statements executed only if all conditions are false

else block

The else statement is optional. But it is more safe to catch the case where all of the given conditions are false.

Exercise 7.1. Chained conditions

The elif statement only facilitates the writing and legibility of multiple alternative conditions. How would you write a multiple condition without this statement (Solution 7.1)? Hint: See the scheme of Figure 7.5.

7.6. Nested conditions

However, construction with multiple alternatives are sometimes not sufcient and you need to nest condition like this:
>>> primerLen = len(primer) >>> primerGC = float(count(primer, g) + count(primer, c))/ primerLen >>> if primerGC > 50: ... if primerLen > 20: ... PCRprogram = 1 ... else: ... PCRprogram = 2

42

Chapter 7. Branching and Decisions

... else: ... PCRprogram = 3

Exercise 7.2. Nested condition

Why is it impossible to write the above example as chained condition? Figure 7.6 shows the scheme of nested conditions.

Figure 7.6. Nested conditions

if condition: if second_condition : block of statements executed only if condition and second_condition are true else : block of statements executed only if condition is true else : if third_condition : block of statements executed only if condition is false and third_condition is true else : block of statements executed only if condition is false and third_condition is false
true true condition false

second condition

false

true

third condition

false

if_if block

if_else block

else_if block

else_else block

Why are they joint here first?

What is known about the second condition here?

Sometimes you can simplify nested conditions by constructing more complex conditions with boolean operators.
>>> primerLen = len(primer) >>> primerGC = float(count(primer, g) + count(primer, c))/ primerLen >>> if primerGC > 50: ... if primerLen > 20: ... PCRprogram = 1 ... else: ... PCRprogram = 2

43

Chapter 7. Branching and Decisions

can be expressed as:

>>> if primerGC > 50 and primerLen > 20: ... PCRprogram = 1 ... else: ... PCRprogram = 2

Caution
Even if the second version is easier to read, be careful and always check whether the complex condition you have written, is what you really want. Such errors are called semantic error. They can not be detected by the interpreter because the syntax of the program is correct, even if it is not necessarily what you want to compute.

7.7. Solutions
Solution 7.1. Chained conditions
Exercise 7.1

Figure 7.7. Multiple alternatives without elif

if condition: block of statements executed only if condition is true elifsecond_condition : block of statements executed only if the second condition is true else : block of statements executed only if all conditions are false if condition: block of statements executed only if condition is true else : if second_condition : block of statements executed only if the second condition is true else : block of statements executed only if all conditions are false

44

Chapter 8. Dening Functions

Chapter 8. Dening Functions

8.1. Dening Functions
In Section 3.3 we have learnt how to apply or call functions. So lets remember the example calculating the GC-percentage of a DNA sequence.
>>> float(count(cds, G) + count(cds, C)) / len(cds)

This calculates the gc percentage of the specic DNA sequence cds, but we can use the same formula to calculate the gc percentage of other DNA sequences. The only thing to do is to replace cds by the new DNA sequence in the formula. But it is not very convenient to remember the formula and retype it all the time. It would be much easier to type the following instead.
>>> gc(ATGCAT) 0.33333333333333331 >>> gc(cds) 0.54460093896713613

The only thing we have to remember is the name of the new function and its use.

Abstraction
The possibility to dene such new function executing tasks specied by yourself, is an abstraction feature, provided by all high level programming languages.

Important
It is also important to notice that functions have to be dened before they are called. You can not use something that is not dened. Here is the syntax of such a new denition in Python:
>>> from string import * >>> def gc(seq): ... return float(count(seq, G) + count(seq, C)) / len(seq)

Lets have a closer look at this denition. Figure 8.1 illustrates the structure of a function denition.

45

Chapter 8. Dening Functions

Figure 8.1. Function denitions

statement indicating a new function definition in Python that has to be followed by the function name and a comma separated list of parameters and a colon indication the start of a new block

second prompt indicating a new block

>>> def gc(seq) : ... return float(count(seq, G) + count(seq, C)) / len(seq)

indentation defining a block statement used to return back a result of a calculation during a function call the parameter (a sort of place holder) is replaced by its argument value

def and return are basic instructions.

Basic instruction
Basic instructions are statements that dene the language rules and the semantic of Python. They constitute the basic set of instructions of Python. Each basic instruction has its own syntax that you have to learn in order to master the programming language. The return basic instruction is used to return the result of a function back, in our example the value of the GC percentage of the specied DNA sequence. The def basic instruction indicates to Python that a function denition follows. It has to be followed by the new function name and a comma separated list of parameter names enclosed into parentheses.

Parameter
Parameters are variable names. When the function is called they are bound in the same order to the arguments given. The body of a function contains the piece of code needed to execute the subtask of the function. In the example above, the body contains only the return statement. Here is a more complex example that excludes ambiguous bases from the GC percentage calculation.
>>> from string import * >>> def gc(seq): ... nbases = count(seq, N) ... gcpercent = float(count(seq, G) + count(seq, C)) / (len(seq) - nbases) ... return gcpercent

In this example the body of the function contains three instructions (2 assignments and the return statement). The body of a function follows the denition line and is written as an indented block initiated by a colon.

46

Chapter 8. Dening Functions

Block
Blocks are structure elements of a program, that are used to group instructions. They are initiated by a colon and are separated from the rest of the code by the same indentation step in Python. In our example, the function body is written as a block to separate it from the rest of the program. The Python interpreter used it to detect where the function body starts and ends.

Important
In other programming languages special words or characters are used to indicate the begin and the end of blocks and the indentation is only used to make the code readable for humans. In Python the indentation is used explicitly to do this. So in the above example of a function denition the end of the body is detected when the next line starts without any indentation step. This is also illustrated in Figure 8.2.

Example 8.1. More complex function denition

>>> from string import * >>> def gc(seq): ... nbases = count(seq, N) ... gcpercent = float(count(seq, G) + count(seq, C)) / (len(seq) - nbases) ... return gcpercent ... >>> gc(ATGCN) 0.5

Figure 8.2. Blocks and indentation

definition line block inititiation sequence of instructions

>>> def gc(seq) : ... nbases = count(seq, N) ... gcpercent = float(count(seq, G) + count(seq, C)) / (len(seq)nbases) ... ... return gcpercent >>> gc(ATGCN) function body block of code 0.5 same indentation to
indicate a block

47

Chapter 8. Dening Functions

8.2. Parameters and Arguments or the difference between a function denition and a function call
As we have already said, functions are named sequences of statements that execute a specic task. A function denition is the construction that gives a name to this special sequence of statements. Whereas a function call is the execution of this sequence of statements. We have also seen that the execution can be parameterized. This means, that there are special variables, the parameters, in the function which are bound to values, the arguments, during the function call. Flow of execution. Some precision about the ow of execution. If you call a function the statements of the function body are executed. The execution of the current sequence of statements areis interrupted and the ow of execution jumps to the sequence of statements named by the function. The function statements are executed and then the ow of execution continues with the sequence from where the function was called.

Caution
The statements in the function body are only executed at a function call, so all errors in this piece of code are only reported at the execution. Now if we remember that within a function you can call other functions, the above scheme becomes rapidly disturbing, but Python keeps a track from where a function is called and there is only one statement at a time that is executed. You can visualize this using a stack diagram of the function calls.

Stack
A stack is an ordered set of things which can be accessed in two ways. You can put things on the top of the stack, and only take back the thing on the top of the stack. A common example is a stack of plates (it is hard to get a plate from within the stack). A stack is sometimes called a LIFO an abbreviation of last-in-rst-out. So what Python does is to handle a stack of function calls. If a function is called, Python puts the position from where it is called on the stack, executes the function and when the function execution is nished Python gets the position to continue back from the stack. Figure 8.3 shows the stack diagram of the execution of the gc function (Example 8.1).

48

Chapter 8. Dening Functions

Figure 8.3. Stack diagram of function calls

gc(ATGCNNATGC)
count s sub x y x x y seq ATGCNNATGC nbases 2 gcpercent
ATGCNNATGC G 2

count

s sub x y x x y

ATGCNNATGC C 2 2

x y x x y

2 2

float count s sub

ATGCNNATGC N

float /

float /

float /

x x y

4 4.0

gc

seq ATGCNNATGC nbases gcpercent

gc

seq ATGCNNATGC nbases gcpercent

gc

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent

len

s x y x y

ATGCNNATGC

2 4.0

x y x y

9 2 4.0

x y

4.0

x y

4.0 7

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent

gc

seq ATGCNNATGC nbases 2 gcpercent 0.5714285

8.3. Functions and namespaces

Even if it is confusing at the beginning, the dna variable in the function call is not the same as the dna parameter of the gc function. During the function call the global variable is evaluated and its value ATGACT is assigned to the parameter dna of the gc function. How does Python distinguish the two variables? It can do so, because they are not in the same namespace. All functions have two associated namespaces, a local and a global namespace. The global namespace is the namespace where the function is dened and the local namespace is a namespace created at the function denition where parameters are dened. The local namespace contains also all local variables. Local variables are variables that are bound to value in the function body. When a function is dened, its source code is analyzed and all used names are sorted in the two namespaces. If a variable name is searched during a function execution, Python looks in the local namespace for local variables and in the global namespace for global variables. If a global variable is not found there, the builtin namespace is searched. If it is not found there, an error is produced.

builtin namespace
The builtin namespace is a namespace that contains all predened function of Python. You can ask Python which names are dened in the global and the local namespace using the functions globals() and locals().
>>> from string import * >>> def gc(seq): ... nbases = count(seq, N) ... gcpercent = float(count(seq, G) + count(seq, C)) / (len(seq) - nbases)

49

Chapter 8. Dening Functions

... print "local namespace:", locals() ... print "global namespace:", globals() ... print "names in the current namespace:", dir() ... return gcpercent ... >>> seq = ATGACGATAGGAGANNTATAGAN >>> gc(seq) local namespace: {gcpercent: 0.34999999999999998, nbases: 3, seq: ATGACGATAGGAGANNTATAGAN} global namespace: {ascii_lowercase: abcdefghijklmnopqrstuvwxyz, upper: <function upper at 0x1623f0>, punctuation: "!#$%&\()*+,-./:;<=>?@[\\]^_{|}~, letters: abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ, seq: ATGACGATAGGAGANNTATAGAN, lstrip: <function lstrip at 0x160cf0>, uppercase: ABCDEFGHIJKLMNOPQRSTUVWXYZ, ascii_letters: abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ, replace: <function replace at 0x160b90>, capwords: <function capwords at 0x160b10>, index_error: <class exceptions.ValueError at 0x119700>, expandtabs: <function expandtabs at 0x160a50>, strip: <function strip at 0x160cb0>, ascii_uppercase: ABCDEFGHIJKLMNOPQRSTUVWXYZ, find: <function find at 0x160810>, gc: <function gc at 0x160b50>, rjust: <function rjust at 0x160990>, ljust: <function ljust at 0x160950>, whitespace: \t\n\x0b\x0c\r , rindex: <function rindex at 0x1625a0>, capitalize: <function capitalize at 0x160ad0>,

atol_error: <class exceptions.ValueError at 0x119700>, octdigits: 01234567, lower: <function lower at 0x1623b0>, splitfields: <function split at 0x160d70>, split: <function split at 0x160d70>, rstrip: <function rstrip at 0x160d30>, translate: <function translate at 0x160a90>, __doc__: None, printable: 0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!"#$%&\()*+,-./:;<=>?@[\ digits: 0123456789, joinfields: <function join at 0x162520>, index: <function index at 0x162560>, __builtins__: <module __builtin__ (built-in)>, swapcase: <function swapcase at 0x162430>, atof_error: <class exceptions.ValueError at 0x119700>, atoi: <function atoi at 0x1608d0>, hexdigits: 0123456789abcdefABCDEF, atol: <function atol at 0x160910>, __name__: __main__, atof: <function atof at 0x160890>, count: <function count at 0x1625e0>, lowercase: abcdefghijklmnopqrstuvwxyz, join: <function join at 0x162520>, center: <function center at 0x1609d0>, rfind: <function rfind at 0x160850>, atoi_error: <class exceptions.ValueError at 0x119700>, maketrans: <built-in function maketrans>, zfill: <function zfill at 0x160a10>} names in the current namespace: [gcpercent, nbases, seq] 0.34999999999999998

50

Chapter 8. Dening Functions

The dir function shows all the names dened in the current namespace.
>>> from string import * >>> dir() [__builtins__, __doc__, __name__, ascii_letters, ascii_lowercase, ascii_uppercase, atof, atof_error, atoi, atoi_error, atol, atol_error, capitalize, capwords, center, count, digits, expandtabs, find, gc, hexdigits, index, index_error, join, joinfields, letters, ljust, lower, lowercase, lstrip, maketrans, octdigits, printable, punctuation, replace, rfind, rindex, rjust, rstrip, seq, split, splitfields, strip, swapcase, translate, upper, uppercase, whitespace, zfill]

8.4. Boolean functions

Whenever a condition is too complex, you can write a boolean function.

Boolean function
Boolean functions are function that return truth values (in Python either 0 or 1). It is good style to give them a name that indicates its nature, often the function name starts with is. Here is an example that tests if a character is a valid amino acid:
>>> from string import * >>> def isAminoAcid(aa): ... AA = upper(aa) ... if AA < A or AA > Z: ... return 0 ... if AA in JOU: ... return 0 ... return 1

you can also write it as follow:

>>> from string import * >>> def isAminoAcid(aa): ... AA = upper(aa) ... if AA < A: ... ok = 0 ... elif AA > Z: ... ok = 0

51

Chapter 8. Dening Functions

... ... ... ... ...

elif AA in JOU: ok = 0 else: ok = 1 return ok

or using boolean operators:

Example 8.2. Function to check whether a character is a valid amino acid

>>> def isAminoAcid(aa): ... return (A <= aa <= Z or a <= aa <= z) and aa not in jouJOU

Using this function makes your code easier to understand because the function name expresses what you want to test:
>>> prot = ATGAFDWDWDAWDAQW >>> oneaa = prot[0] >>> if isAminoAcid(oneaa): ... print ok ... else: print not a valid amino acid ok

52

Chapter 9. Collections

Chapter 9. Collections
9.1. Datatypes for collections
In the proceeding chapter we have seen that strings can be interpreted as collections of characters. But they are a very restricted sort of collection, you can only put characters in a string and a string is always ordered. But we need often to handle collections of all sorts of objects and sometimes these collections are not even homogeneous, meaning that they may contain objects of different types. Python provides several predened data types that can manage such collections. The two most used structures are called Lists and Dictionaries. Both can handle collections of different sorts of objects, but what are their differences?

List
Lists are mutable ordered collections of objects of different sorts. The objects are accessible using their position in the ordered collection.

Dictionary
Dictionaries are mutable unordered collections which may contain objects of different sorts. The objects can be accessed using a key. Here are some examples comparing a list version of a collection of enzymes pattern and a dictionary version of the same collection. Lists are created using a comma separated list of all elements enclosed into brackets, whereas dictionaries are enclosed into braces and contain a comma separated list of key-value pairs, each separated by a colon.
>>> patternList = [ gaattc, ggatcc, aagctt ] >>> patternList [gaattc, ggatcc, aagctt] >>> patternDict = { EcoRI : gaattc, BamHI : ggatcc, HindIII : aagctt } >>> patternDict { EcoRI : gaattc, BamHI : ggatcc, HindIII : aagctt }

To access the elements we use the position in the list and the key for the dictionary.

Important
List indices have the same properties as string indices, in particular they start with 0 (remember Figure 6.1).
>>> patternList[0] gaattc >>> patternDict[EcoRI]

53

Chapter 9. Collections

gaattc

As for strings you can get the number of elements in the collection, as well as the smallest and the greatest element.
>>> len(patternList) 3 >>> len(patternDict) 3

Lists can be sliced but dictionaries cannot. Remember that dictionaries are unordered collections, so getting a slice does not make any sense.
>>> digest = patternList[:1] >>> digest [gaattc]

You can ask whether an element is in the collection. Dictionaries have two possibilities to perform this.
>>> gaattc in patternList 1 >>> patternDict.has_key(HindIII) 1 >>> HindIII in patternDict 1

Unlike strings both collections are mutable. This means that you can remove, add or even change their elements.
>>> del patternList[0] >>> patternList [ggatcc, aagctt] >>> patternList[0] = gaattc >>> patternList [gaattc, aagctt] >>> patternList.append(ggattc) >>> patternList [gaattc, aagctt, ggattc] >>> del patternList[:2] >>> patternList [ggattc] >>> del patternDict[EcoRI] >>> patternDict {BamHI: ggatcc, HindIII: aagctt} >>> patternDict[EcoRI] = gaattc >>> patternDict {BamHI: ggatcc, EcoRI: gaattc, HindIII: aagctt} >>> patternDict[BamHI] =

54

Chapter 9. Collections

>>> patternDict {BamHI: , EcoRI: gaattc, HindIII: aagctt}

As for strings you cannot access elements that do not exist.

>>> patternList[10] Traceback (most recent call last): File "<stdin>", line 1, in ? IndexError: list index out of range >>> patternDict[ScaI] Traceback (most recent call last): File "<stdin>", line 1, in ? KeyError: ScaI

Figure 9.1 compares different actions on collections for strings, lists and dictionaries.

Figure 9.1. Comparison some collection datatypes

Action Strings List [a, b, ..., n] L[i] Dictionnary {keya: a, keyb: b, ..., keyn:n } D[key]

Creation Access to an element

"..." ..., "..." s[i] c in s Not Possible s = s[:i1]+s[i+1:] Not Possible s=s[:i1]+new+s[i+1:] Not Possible s=s + new Not Possible s=s[:i]+s[k:] Not Possible s=s[:i]+news+s[k:] Not Possible s=s+news

Membership

e in L

key in D D.has_key(D) del D[key]

Remove en element

del L[i]

Change an element

L[i]=new

D[key]=new

Add an element Remove consecutive element Change consecutive element Add more than one element

L.append(e)

D[newkey]=val

del L[i:k]

Not Possible, not ordered but, remove all D.clear() Not Possible

L[i:k]=Lnew L.extend(newL) or L + L + Lnew

D.update(newD)

9.2. Methods, Operators and Functions on Lists

55

Chapter 9. Collections

Table 9.1 remembers the action of builtin functions and operators on list objects and Table 9.2 summarizes all methods of list objects.

Table 9.1. Sequence types: Operators and Functions

Operator/Function [ ... ], ( ... ... " s + t s * n s[i] s[i:k] x in s x not in s for a in s len(s) min(s) max(s) s[i] = x s[i:k] = t del s[i]
a

), "

Action creation concatenation repetition a indication slice membership iteration length return smallest element return greatest element index assignment slice assignment deletion

Action on Numbers

addition multiplication

Important shallow copy (see Section 11.2)

Table 9.2. List methods

Method list(s) s.append(x) s.extend(t) s.count(x) s.index(x) s.insert(i,x) s.pop([i]) s.remove(x) s.reverse()b s.sort([cmp])b
a b

Operation converts any sequence object to a list append a new element concatenationa count occurrences of x nd smallest position where x occurs in s insert x at position i removes i-th element and return it remove element reverse sort according to the cmp function

equal to the + operator but inplace in place operation

56

Chapter 9. Collections

Important
It is important to know whether a function or method, that is applied to a mutable objects, modies this object internally or whether it returns a new object containing these modications. Look at the following example that shows two possibilities to concatenate lists. The + operator creates a new list whereas the method extend adds one list to the other:
>>> l1 = [ EcoRI, BamHI ] >>> l2 = [ HindIII ] >>> l1 [EcoRI, BamHI] >>> l2 [HindIII] >>> l1 + l2 [EcoRI, BamHI, HindIII] >>> l1 [EcoRI, BamHI] >>> l2 [HindIII] >>> l1.extend(l2) >>> l1 [EcoRI, BamHI, HindIII] >>> l2 [HindIII]

9.3. Methods, Operators and Functions on Dictionaries

Contrary to strings and lists the ordering of the elements in dictionaries does not matter. Elements are accessed by a unique key rather than by an index number.

Important
It is important to notice that dictionary keys have to be unique and immutable. Table 9.3 gives an overview of the methods and operations on dictionaries.

Table 9.3. Dictionary methods and operations

Method or Operation d[key] d[key] = val del d[key] d.clear() len(d) Action get the value of the entry with key key in d set the value of entry with key key to val delete entry with key key removes all entries number of items

57

Chapter 9. Collections

d.copy() d.has_key(key) d.keys() d.values() d.items() d.update(new) d.get(key [, otherwise]) d.setdefaults(key [, val]) d.popitem()
a

makes a shallow copya returns 1 if key exists, 0 otherwise gives a list of all keys gives a list of all values returns a list of all items as tuples (key,value) adds all entries of dictionary new to d returns value of the entry with key key if it exists otherwise returns otherwise same as d.get(key), but if key does not exists sets d[key] to val removes a random item and returns it as tuple

Important shallow copy (see Section 11.2)

9.4. What data type for which collection

We have seen so far three collection types in Python. Which one should you chose in your applications? Because they all have their advantages and disadvantages making some actions easy or difcult to handle, the choice depends on your data and on what you would do with these data. To recognize strings is the easiest one, but be aware if you need to change characters, since you will have to construct a new string. So a mutable ordered collection datatype, such as a list, could be more adapted. For list and dictionaries there are two important issues to consider: 1. Does the ordering of your collection matter? 2. Is it sufcient to access the elements by their position or do you prefer to access the collection with a more complex, sometimes more descriptive, key rather than a number? If you need an ordered collection you do not have any choice left, because only lists take care of the ordering. However the way of accessing a collection is sometimes not so easy and often worth further consideration.

58

Chapter 10. Repetitions

Chapter 10. Repetitions

10.1. Repetitions
Sometimes you need to do the same thing several times with different data. Whereas humans make often errors during a repetitive procedure, computers execute repetitive tasks very well. One possibility to write repetitions is to repeat the code as often as needed, but this is neither elegant nor readable, and above all you cannot do this if the number of repetition times is not constant or is not known beforehand. Just as now, we have seen all that we need to do repetitive tasks.

Exercise 10.1. Repetitions

Try to nd out how to perform a repetitive task, only with the statements seen so far. It is not very easy if you have not seen this and we will discuss this later (Chapter 13). Because repetitions are common programming tasks, high level programming languages provide special structures and statements for this purpose. Python has two of them: for and while. Let us see the for statement rst.

10.2. The for loop

Verify protein sequences. Let us start with an example that checks whether all characters of a protein sequence are valid amino acid characters. How could we do this by hand? One possibility is to start at the beginning of the protein sequence and check each character one by one. Here is a more formal description of the algorithm: 1. for each character of the protein sequence a. if the current character is not an amino acid print the invalid character b. end if 2. end for In Python this can be performed with the for statement.

59

Chapter 10. Repetitions

The for loop

The for loop enables to iterate on an ordered collection of objects and to execute the same sequence of statements for each element. This sort of iteration is also known as a traversal, because the collection is traversed from left to right and a particular task is applied to each element. So in our example, the collection of elements is a string which we traverse with the for loop. We also have to specify what a valid amino acid is, because computers do not know about it. Nevertheless we have already written a function doing this (Example 8.2).
>>> protein = "SERLSITPLGPYIGAQIJSGADLTRPLSDNQFEQLYHAVLRHQVVFLRDQAITPQQQRALA" >>> for aa in protein: ... if not isAminoAcid(aa): ... print "protein contains unknown amino acid: ", aa ... protein contains unknown amino acid: J

As def and if, the for statement is also a statement containing a block. The header line gives the two essential information for the traversal. It starts with for followed by a variable name to which each element of the collection is assigned during the execution of the block of statements. The second half of the header indicates the collection in which the elements can be found. As in all functions, operations and expressions, the collection can be specied by a value, a variable or even a composed expression.

Figure 10.1. The for loop

iteration variable collection to traverse use of the iteration variable in the body

for aa in protein : if isAminoAcid( aa): print "protein contains unknown amino acid: ",aa
sequence of statements to repeat for each element

Getting the position of invalid amino acids. As written above we cannot show the position of the invalid amino acid in the protein sequence, because only the amino acid character is known in the body of the for loop. Therefore, we have to iterate over the indices of the sequence. Here is the algorithm: 1. for each position in the protein sequence a. if the character at this position is not an amino acid print the position and the invalid character

60

Chapter 10. Repetitions

b. end if 2. end for What are the possible values of the positions? Remember the string indices, they start at 0 and end at one before the length of the string. Python has a function that gives the possibility to create a list of numbers that is called range. Using this function, we can translate the algorithms above into Python:
>>> for i in range(len(protein)): ... if not isAminoAcid(protein[i]): ... print "protein contains unknown amino acid " , protein[i] , \ ... " at position ", i

Note
There is a convention in most programming languages to use i, k, l as variable names for the iteration variable. Translate cds sequences. Lets take another example. Say that we have a cds sequence and that we would like to know such things as the corresponding amino acid sequence or the codon usage. How could we achieve this? For sure, it looks like a traversal of the cds sequence, but the things we are looking for do not correspond anymore to one nucleotide of the cds sequence but to a codon which is a three letter substring. If we could access the codons one by one, the translation into the amino acid sequence would look like this: for each codon in the cds sequence: add the amino acid corresponding to the current codon to the protein sequence In the body of the loop we need to establish the correspondence between a codon sequence and the amino acid. We know how to do this by hand, by looking up a codon in the codon translation table. A dictionary is the data type that gives us such sort of access to the collection of codons.
>>> code = { ... ... ... ... ... ... ... ... ... ... ... ... ttt: ttc: tta: ttg: ctt: ctc: cta: ctg: att: atc: ata: atg: gtt: F, F, L, L, L, L, L, L, I, I, I, M, V, tct: tcc: tca: tcg: cct: ccc: cca: ccg: act: acc: aca: acg: gct: S, S, S, S, P, P, P, P, T, T, T, T, A, tat: tac: taa: tag: cat: cac: caa: cag: aat: aac: aaa: aag: gat: Y, Y, *, *, H, H, Q, Q, N, N, K, K, D, tgt: tgc: tga: tgg: cgt: cgc: cga: cgg: agt: agc: aga: agg: ggt: C, C, *, W, R, R, R, R, S, S, R, R, G,

61

Chapter 10. Repetitions

... ... ... ... } >>> code[atg] M

gtc: V, gcc: A, gac: D, ggc: G, gta: V, gca: A, gaa: E, gga: G, gtg: V, gcg: A, gag: E, ggg: G

Now let us go back to the rst part of the problem: getting the codons. If we know where a codon starts, we only have to extract the substring of length 3 starting from this position. Here is the algorithm: for each third position in the cds sequence: get the substring of length three starting from this position Fortunately, the range function can take an optional step argument.
>>> [0, >>> [4, >>> [0, range(10) 1, 2, 3, 4, 5, 6, 7, 8, 9] range(4,10) 5, 6, 7, 8, 9] range(0,10,3) 3, 6, 9]

We can now print the codon list:

>>> cds = "atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaataa" >>> for i in range(0,len(cds),3): ... print cds[i:i+3], ... atg agt gaa cgt ctg agc att acc ccg ctg ggg ccg tat atc ggc gca

62

Chapter 10. Repetitions

caa taa taa

If we replace the print statement with the access to the dictionary of translation, we can translate the codon into the amino acid.

Example 10.1. Translate a cds sequence into its corresponding protein sequence
>>> def translate(cds, code): ... prot = "" ... for i in range(0,len(cds),3): ... codon = cds[i:i+3] ... prot = prot + code[codon] ... return prot >>> translate(cds, code) MSERLSITPLGPYIGAQ*

What about the computing of the codon usage? We do not need the translation table anymore. But we have to count each codon now. We also need a data structure accessible by codons, although it is not to get information, but to store the result. Here is the algorithm: do not forget that accessing a key which does not exist in a dictionary, is not allowed. for each codon in the coding sequence: a. if the codon is already in the dictionary of the usage: then add 1 to the count b. otherwise: put the codon in the dictionary with count = 1

63

Chapter 10. Repetitions

Here is the corresponding Python code:

>>> def count_codons(cds): ... usage = {} ... for i in range(0,len(cds),3): ... codon = cds[i:i+3] ... if usage.has_key(codon): ... usage[codon] += 1 ... else: ... usage[codon] = 1 ... return usage ... >>> count_codons(cds) {acc: 1, atg: 1, atc: 1, gca: 1, agc: 1, ggg: 1, att: 1, ctg: 2, taa: 1, ggc: 1, tat: 1, ccg: 2, agt: 1, caa: 1, cgt: 1, gaa: 1}

Exercise 10.2. Write the complete codon usage function

Transform the count_codons function into a function getting the real codon usage. Hint: You need to divide each count by the total number of codons.

10.3. The while loop

The while loop
A while loop is composed of two parts, a condition and a sequence of statements to repeat. During the execution, the sequence of statements are repeated as long as the condition is true. Figure 10.2 illustrates the ow of execution.

Figure 10.2. Flow of execution of a while statement

while condition : block of statements executed as long as condition is true

while block

true

condition false

64

Chapter 10. Repetitions

Before saying more, let us illustrate that with an example. We have already shown how to use the find method of strings to nd pattern (Section 1.1). Lets try now to nd all occurrences of a specic pattern in a DNA sequence. Here is one possible way to proceed:

Procedure 10.6. Find all occurrences of a pattern in a sequence

INPUT: a DNA sequence seq and a pattern pat OUTPUT: a list of positions matches containing all start positions of pat in seq 1. matches <- empty list 2. current_match <- position of the rst occurrence of pattern pat in sequence seq or -1 if there is not anyone 3. as long as current_match is not -1: a. matches <- matches + current_matches b. current_match <- position of the next occurrence of pattern pat in sequence seq after current_match or -1 if there is not anyone 4. return matches

Example 10.2. First example of a while loop

And here is the implementation in Python:
def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches

Lets follow with a second example that rewrites the translate function (Example 10.1) using a while loop instead of the for traversal.

Example 10.3. Translation of a cds sequence using the while statement

This version follows the same procedure as the implementation with the for statement.
code = {ttt: F, tct: S, tat: Y, tgt: C, ttc: F, tcc: S, tac: Y, tgc: C, tta: L, tca: S, taa: *, tga: *, ttg: L, tcg: S, tag: *, tgg: W, ctt: L, cct: P, cat: H, cgt: R, ctc: L, ccc: P, cac: H, cgc: R,

65

Chapter 10. Repetitions

cta: ctg: att: atc: ata: atg: gtt: gtc: gta: gtg: }

L, L, I, I, I, M, V, V, V, V,

cca: ccg: act: acc: aca: acg: gct: gcc: gca: gcg:

P, P, T, T, T, T, A, A, A, A,

caa: cag: aat: aac: aaa: aag: gat: gac: gaa: gag:

Q, Q, N, N, K, K, D, D, E, E,

cga: cgg: agt: agc: aga: agg: ggt: ggc: gga: ggg:

R, R, S, S, R, R, G, G, G, G

def translate_while(cds,code): prot = "" last_i = len(cds)-3 i = 0 while i < last_i: codon = cds[i:i+3] prot = prot + code[codon] i += 3 return prot

In Python a while loop is written using the while statement which is as the for and the if statements a block statement. The header of the block statement contains the conditional expression and the body contains the sequence of statements to repeat (Figure 10.3).

Figure 10.3. Structure of the while statement

initalization of the changing variable variable which change during the repetition condition of the repetion in the header of the while statement

matches = [] current_match = seq.find(pat) while current_match != 1:

matches.append(current_match) current_match = seq.find(pat, current_match+1)

new value of the changing variable sequence of statements to repeat in the body of the while statement

Look at the following example. Do you know what happened during the execution?

66

Chapter 10. Repetitions

while 1: print "hello"

It simply never stops, because 1 is always true and therefore it is neither an algorithm nor a program. But it illustrates an important feature of a while loop. The conditional expression has to contain at least one variable whose value changes during the execution of the sequence of statements. This variable is often a simple iterator, which is increased or decreased until it reaches a limit. One example of such an iterator is the variable i in Example 10.3. The behavior of the current_match variable in Example 10.2 is more complicated, because it represents two situations. Its value is the start position of a pattern occurrence or it is -1 if there are no more occurrences. In this example the second argument of the find function ensures that the next occurrence is found but no occurrence missed. Because the iteration variable is used in the condition and the condition is evaluated rst when a while loop is entered, it has to be dened before entering the loop. This means there are two different affectations of this variable. A rst affectation or initialization before the loop is entered and a second affectation statement in the repeated sequence of statements that gives it the next value. In general, this is the last statement of the sequence. Even if it is not always the case, and sometimes not necessary, it is a convention to put it at the end if possible because it is easy to nd if it has been forgotten.

Warning
Conditions of while loops are frequent error sources. They are sometimes difcult to detect because they are semantic errors that produce wrong results on all or sometimes only special input data sets. You have to ensure that the iteration variable changes and that the end condition is always reached.

10.4. Comparison of for and while loops

What are the differences between while and for? And in which cases it is more appropriated to use one or the other? There are many cases where it is only a question of feeling. But the while loop is the more general one because its loop condition is more exible and can be more complicated than that of a for loop. The condition of a for loop is always the same and implicit in the construction. A for loop stops if there are no more elements in the collection to treat. For simple traversals or iterations over index ranges it is a good advice to use the for statement because it handles the iteration variable for you, so it is more secure than while where you have to handle the end of the iteration and the change of the iteration variable by yourself. The while loop can take every boolean expression as condition and permits therefore more complex end conditions. It is also the better choice if the iteration variable does not change evently by the same step or if there are more than one iteration variable. Even if you can handle more than one iteration variable in a for statement, the collections from which to choose the values must have the same number of elements.

67

Chapter 10. Repetitions

Exercise 10.3. Rewrite for as while

Imagine that there is no for statement in the Python language. How can you rewrite it as while statement? Take the following pseudocode as example:
for var in collection: do something with var

The for loop is a special case of the

10.5. Range and Xrange objects

We have already introduced the range function which generates a list of integers from a start to an end value (the last is not included) with a regular interval. This function is very useful while writing for loops because it gives the possibility to iterate rather over the indices of a collection than over their elements. So you have not only access to the value of the element but also its position in the collection. Sometimes you need to iterate over a great interval of integers. In this case the list generated by the range function would be long and therefore take a lot of memory space and time to be constructed. In such cases it is more appropriated to use the xrange function which is used in the same way as the range function. But rather than to create the list of integers physically in memory, it creates an object that calculates the new value if it is asked for the next element in the list.

10.6. The map function

We will nish this chapter with the introduction of the function map which allows to apply a function to all arguments of an ordered collection. The result is always a list even if the ordered collection is a string. If the function takes more than one argument, as many collections as arguments have to be specied as arguments. map takes the rst argument from the rst collection, the second one from the second collection and so on.
>>> map(isAminoAcid, "atgahryuox") [1, 1, 1, 1, 1, 1, 1, 0, 0, 1] >>> def add10(n): ... return n+10 ... >>> map(add10, [0,1,2,3,4,5]) [10, 11, 12, 13, 14, 15] >>> def add(a,b): return a+b >>> map(add, [1,2,3], [1,2,3]) [2,4,6]

68

Chapter 10. Repetitions

Passing functions as arguments. This is the rst example where we pass a function as an argument. Even if this looks strange at the beginning, function names are references as variable names which are bound to the sequences of statements dened in the body of the function denition. The sequence of statements can be viewed as an object which has the particularity that it can be called with arguments. Figure 10.4 shows what happens when passing a function as argument. This illustrates that there is no difference between variables and functions as arguments.

Figure 10.4. Passing functions as arguments

global namespace
isAminoAcid prot
sequence of statements that checks if a character is an amino acid

local namespace: map(isAminAcid, prot)

function sequence athncfxuoa

Another example that handles function names as references is renaming a function as follows:
>>> isAA = isAminoAcid >>> isAA(a) 1

The only difference between function and variable names is that function names are bound to objects that can be called using the function call syntax.
>>> isAminoAcid <function isAminoAcid at 0x111ad0> >>> isAminoAcid(a) 1

Without the parenthesis the object bound to the name isAminoAcid is shown by the interpreter, whereas adding the parentheses will call the function object bound to isAminoAcid.

69

Chapter 10. Repetitions

Rewriting map with for. map with a function that takes only one argument, can be rewritten as follows.. Rewrite map(func, collection) 1. res < empty list 2. for each element in collection: res < res appended with func(element) 3. return res

10.7. Solutions
Solution 10.1. Rewrite for as while
Exercise 10.3
while collection: var = collection[0] do something with var collection = collection[1:]

or using an index
i = 0 while i < len(collection): var = collection[i] do something with var i += 1

70

Chapter 11. Nested data structures

Chapter 11. Nested data structures

11.1. Nested data structures
Composed data structures, such as list, dictionaries and tuples, can be nested in Python. This means that you can use these structures as elements of themselves.
>>> l = [1, [2, 3, [4]], [5, 6]]

The list above contains 3 elements.

>>> len(l) 3

The rst element is the integer 1, whereas the second and third element are lists themselves.
>>> 1 >>> [2, >>> [5, l[0] l[1] 3, [4]] l[2] 6]

Figure 11.1 shows a representation of this list. Each element of a list is represented by a box containing a pointer to the value of the element. Like pointers in a namespace representation, these pointers are references. Remember that a reference is an address to a memory location (Section 2.4).

Figure 11.1. Representation of nested lists

71

Chapter 11. Nested data structures

As far we have said that a list is an ordered collection of some objects, but rather to contain the objects themselves, a list contains references to them. How can we access to the elements of the internal list? Figure 11.1 makes clear there are different levels of the nested structure. Each time we follow a reference we go to the next level of the structure. To access the value 2 of our example list, we could store the value of the second element in a variable which creates the new reference l1 to this sublist. And then ask for the rst element of l1.
>>> >>> [2, >>> 2 l1 = l[1] l1 3, [4]] l1[0]

But we can also follow the references directly without creating a new variable.
>>> l[1][0] 2

Figure 11.2 colors the path of the references we have followed to access the value 2.

Figure 11.2. Accessing elements in nested lists

1 l1 2 3 5 6

4
72

Chapter 11. Nested data structures

Nested dictionaries. Dictionaries can be nested in the same way as lists. Here is an example of a dictionary that stores restriction enzyme data.
>>> enzdict = { EcoRI: {pattern: GAATTC, cut_position: 1}, BamHI: {pattern: GGATCC, cut_position: 1}} >>> enzdict { EcoRI: {pattern: GAATTC, cut_position: 1}, BamHI: {pattern: GGATCC, cut_position: 1}} >>> enzdict[EcoRI][pattern] GAATTC

Figure 11.3 shows a representation of enzdict.

Figure 11.3. Representation of a nested dictionary

pattern EcoRI BamHI pattern cut_position cut_position

GAATTC 1

GGATCC 1

Mixed structures. In the same way as above you can construct mixed structures. Here is an example of a restriction enzyme dictionary that stores, in addition to the pattern, all occurrences of the pattern in a sequence.

Example 11.1. A mixed nested datastructure

>>> enzdict = { EcoRI: {pattern: GAATTC, occ: [0, 109, 601]}, BamHI: {pattern: GGATCC, occ: [31, 59]}

Exercise 11.1. Representing complex structures

Try to draw a representation of the the mixed structure in Example 11.1.

11.2. Identity of objects

73

Chapter 11. Nested data structures

In ??? we have said that there exists the two operators == and is for comparing objects in Python without explaining their differences. Both operators compare the identity of objects or values, but the is operator looks if their memory locations are the same, whereas the == operator rather does a content based comparison. Therefore for simple objects, such as integers, and immutable objects, such as strings, the operators return the same result, but their behavior is different for composed objects as lists and dictionaries. The following listing shows some examples:
>>> 1 >>> 0 >>> 1 >>> 1 >>> 1 >>> 1 >>> >>> >>> 1 >>> >>> 1 >>> 0 >>> >>> 1 >>> >>> 0 10.0 == 10 10.0 is 10 10 is 10 10.0 is 10.0

"atg" == "atg" "atg" is "atg" start = "atg" end = start start == end

L1 = [1, 2, 3] L1 == [1, 2, 3] L1 is [1, 2, 3] L2 = L1 L1 is L2 L3 = L1[:] L1 is L3

>>> [1, [2, 3]] == [1, 2, 3] 0 >>> [1, [2, 3]] == [1, [2, 3]] 1

Figure 11.4 shows a representation to illustrate the examples of list comparisons done above.

74

Chapter 11. Nested data structures

Figure 11.4. List comparison

anonymous list compared to L1

L1 1 L2 L3
Identity of objects. In Python all objects have an identier that is unique for each object. Therefore two objects are the same in the sense of is if they have the same identier. You can ask for the identier of an object in Python using the id function:
>>> id(10) 1104200 >>> id(10.0) 1424356 >>> id("atg") 1183232 >>> id(start) 1183232 >>> id(L1) 1177744 >>> id(L2) 1177744 >>> id(L3) 1121328 >>> id([1, 2, 3]) 1174048

(the integers are the same as those of L1 for lisibility reasons they are doubled here)

Warning
In Python strings are immutable and handled as the same in the sense of the is function if they contain the same sequence of characters, but this is not the case in all programming languages. In C for example, even strings with the same sequence of characters can have different memory locations.

75

Chapter 11. Nested data structures

11.3. Copying complex data structures

In the list examples of the above Section 11.2 we have given the example:
>>> L3 = L1[:]

and shown that L3 and L1 are not the same in memory

>>> L3 is L1 0

This is due to the fact that slicing a list creates a copy of a list and therefore L3 is only a copy of L1 but not the same object in memory. Lets look at a second example with a nested list structure:
>>> >>> [1, >>> >>> 0 >>> 1 L1 = [1, [2, 3], 4] L1 [2, 3], 4] L2 = L1[:] L1 is L2 L1[1] is L2[1]

Figure 11.5 illustrates what happens. In fact slicing creates only a shallow copy of compound objects by creating a new compound object and populating the new object with references to the members of the old list.

Figure 11.5. Copying nested structures

L1

L2

76

Chapter 11. Nested data structures

11.4. Modifying nested structures

Why is it important to know about the identity of compound object elements? Look at the following:
>>> L1 = [1, [2, 3], 4] >>> L2 = L1[:] >>> >>> [1, >>> [1, L2[1][0] = 5 L2 [5, 3], 4] L1 [5, 3], 4]

This example modies the rst element of the list that is the second element of L2. But the second element of L1 is the same in memory as the second element of L2. That is the reason why L1 is also modied, whereas if we modie the rst level of L1, L2 remains unmodied because that only changes independent references in the two list.
>>> [1, >>> [1, >>> >>> [1, >>> [1, L1 [5, 3], L2 [5, 3], L1[1] = L1 6, 4] L2 [5, 3],

4] 4] 6

4]

Figure 11.6 illustrates in red the rst modication and in green the second one.

77

Chapter 11. Nested data structures

Figure 11.6. Modifying compound objects

6 L1

5 2 3

L2
Copying references occurs also when you use variables.
>>> a = [1, 2] >>> b = a >>> a is b 1 >>> c = [a, 3] >>> c [[1, 2], 3] >>> c[0] is a 1 >>> a[1] = 0 >>> c [[1, 0], 3] >>> a [1, 0] >>> b [1, 0] >>> >>> [0, >>> [1, c[0] = 0 c 3] a 0]

Independent copies. It is possible to get an independent copy via the deepcopy function dened in the copy module.

78

Chapter 11. Nested data structures

>>> >>> [1, >>> [1, >>> >>> [1, >>> 0

import copy L1 2, 4] L2 [1, 3], 4] L3 = copy.deepcopy(L2) L3 [1, 3], 4] L3[1] is L2[1]

79

Chapter 11. Nested data structures

80

Chapter 12. Files

Chapter 12. Files

During a program execution data are stored in memory. But if the program ends the data are cleared from memory. In this chapter we will explain how programs can communicate data permanently. In the above sections we have seen how we can communicate interactively with programs during their execution. However, if we need to store data permanently, we have to store them in les.

Files
Files are entities containing character data which are stored on external devices, such as disks or CD-ROMs.

12.1. Handle les in programs

Files are like books. You open them to start working, then read or write in them and you close them when you have nished your work. However, you have always to know where you are in the book. As children use their ngers when they start to learn reading, you manipulate a le pointer which indicates your current position in the le. File data are ordered collections, such as lists. But you have to traverse all elements to get to a position, whereas lists can by accessed directly using an index. Opening. When you open a le in Python using the built-in function open, you have to indicate, in addition to its name, whether you want to read from it or write in it. The default is reading. Working. The le object created by the open function has methods to read from the le, to write in it and to move the le pointer to another position. Although Python can handle les containing binary data as well as text data, le objects have special functions to work with text les, such as reading line per line. At this point we like to show two possibilities to handle text le data line per line. The rst one uses the loop while:
infile="<some-file-name"> infh=open(infile) line=infh.readline() while line: #do something with the line line=infh.readline() infh.close()

81

Chapter 12. Files

And the second one the loop for.

Example 12.1. Reading from les

infile="<some-file-name>" infh=open(infile) for line in infh.xreadlines() #do something with the line infh.close()

There exists two le methods, readlines and xreadlines, with the same properties as range and xrange. The rst one reads all the lines of a le and puts them in a list, whereas the second one allows to iterate over all lines without creating the list. The rst one is appropriated for small les, but if you are not sure about the number of lines in your le, the second one prevents to overload the memory. Table 12.1 gives a list of all common used le methods and Table 12.2 explains all possible modes to work with les.

Table 12.1. File methods

Method read([n]) readline([n]) readlines() xreadlines() write(s) writelines(l) close() seek(offset [, mode]) Action reads at most n bytes; if no n is specied, reads the entire le reads a line of input, if n is specied reads at most n bytes reads all lines and returns them in a list reads all lines but handles them as a XRangeTypea writes strings s writes all strings in list l as lines closes the le changes to a new le position=start + offset. start is specied by the mode argument: mode=0 (default), start = start of the le, mode=1, start = current le position and mode=2, start = end of the le

See Section 10.5 for more informations

Table 12.2. File modes

Mode r w a Description read write append

82

Chapter 12. Files

[rwa]b r+ w+

[reading,writing,append] as binary data (required on Windows) update+reading (output operations must ush their data before subsequent input operations) truncate to size zero followed by writing

Closing. Although Python closes all opened les when the interpreter exits, it is a good idea to close them explicitly. Why ???

Important
Pay attention, while open is a built in function creating a le object, close is a method of the created le object.

12.2. Reading data from les

Let us work on a concrete example. In Section 10.3, we have written the findpos function that nds all occurrences of a pattern in a sequence. In the following example, we will show how we can read restriction site patterns of enzymes described in the ReBase [http://www.rebase.org] database. ReBase database format. In our example we will use an excerpt of the gcgenz.dat le of the restriction enzyme ReBase database. The information for each restriction enzyme is stored in one line of this le. The information about isoschizomers and the header of the le containing explanations have been omitted in the excerpt that we will use. Figure 12.1 shows a part of this le. Each line starts with the enzyme name, followed by the number of bases after which it cuts the restriction site pattern, the sequence of the restriction site, the length of the overhang and some comments about the isoschizomeres and the availability of the enzyme.

Figure 12.1. ReBase le format

AarI AatII AccI Acc65I AciI AclI AfeI AflII AflIII AgeI 11 CACCTGCnnnnnnnn_ 5 G_ACGTC -4 ! 2 GTmk_AC 2 ! 1 1 2 3 1 1 1 GGTAC_C CCG_C AACG_TT AGCGCT CTTAA_G ACryG_T ACCGG_T 4 2 2 0 4 4 4 ! ! ! ! ! ! ! 4 ! ZraI FblI,XmiI KpnI,Asp718I Psp1406I Eco47III,Aor51HI,FunI BfrI,BspTI,Bst98I,MspCI,Vha464I AsiAI,BshTI,CspAI,PinAI >F 201,386 >AEFGIKMNOR 624 >ABEGJKMNORSU 288,374,751 >FGINR 505 >N 497 >IN 140 >IN 15 >ABJKNO 722 >BGMNS 722 >GJNR 739

Using the general scheme of reading lines from les shown above (Example 12.1), the following shows one possibility for extracting the restriction site pattern from the line.

83

Chapter 12. Files

Procedure 12.1. read_rebase

INPUT: a le in rebase format OUTPUT: a dictionary enz_dict containing all restriction patterns indexed by their name 1. enz_dict <- empty dictionary 2. infh <- open le for reading 3. for each line read from infh: a. b. c. d. e. split line in its elds name <- rst eld pat <- third eld clean pat to get a string containing only the sequence recognized by the restriction enzyme add the cleaned pat to enz_dict with name as key

4. close infh 5. return enz_dict here is one possibility how to translate this procedure into Python:
def get_site_only(pat): newpat = "" for c in pat: if c.isalpha(): newpat +=c return newpat def read_rebase(filename): enz_dict={} infh= open(filename) for line in infh.xreadlines(): fields = line.split() name = fields[0] pat = fields[2] enz_dict[name] = get_site_only(pat) infh.close() return enz_dict print read_rebase("../data/rebase.dat")

12.3. Writing in les

Let us continue our restriction site example. Because we have got the enzyme pattern from the ReBase database, we can now process our sequence with all these patterns using the findpos function (Example 10.2). There is only one restriction: at the moment the findpos function can only nd exact restriction patterns, so we have to exclude all patterns containing ambiguous bases.

84

Chapter 12. Files

INPUT: a dictionary enz_dict containing all restriction site patterns accessible by enzyme name, a sequence seq to search for OUTPUT: list of start position of every occurrence for each pattern in the dictionary. for each key enzname in enz_dict: a. pat <- pattern of enzname in enz_dict b. if pat is an exact pattern i. occs <- ndpos(seq,pat) ii. print all occs The print_matches in the following listing of functions prints the results of the analysis on the screen.
seq = "" def isexact(pat): for c in pat.upper(): if c not in ATGC: return 0 return 1 def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches

def print_matches(enz, matches): if matches: print "Enzyme %s matches at:" % enz, for m in matches: print m, print else: print "No match found for enzyme %s." % enz for enzname in enz_dict.keys(): pat = enz_dict[enzname] if isexact(pat): print_matches(enzname, findpos(seq, pat))

85

Chapter 12. Files

In order to store the results permanently, we will see now how we can write the information in a le. As for reading, we have to open a le although now in a writing mode, write our results in the le and then close it.
def print_matches(enz, matches): ofh = open("rebase.res", "w") if matches: print >>ofh, "Enzyme %s matches at:" % enz, for m in matches: print >>ofh, m, print >>ofh else: print >>ofh, "No match found for enzyme %s." % enz ofh.close()

The problem with this print_matches function is that it prints only the result of last enzyme. Because if we close the le after writing the information, the next time we will open the le for writing the next result, we will overwrite the old result. We have two possibilities to solve this. First, we can open the le to append at the end of the le. Or second, we can open the le for writing in the main stream of the program and then pass the le object as argument to print_matches, and close the le only when all results have been written. We prefer the second solution.
seq = "" def isexact(pat): for c in pat.upper(): if c not in ATGC: return 0 return 1 def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches def print_matches(ofh, enz, matches): if matches: print >>ofh, "Enzyme %s matches at:" % enz, for m in matches: print >>ofh, m, print >>ofh else: print >>ofh, "No match found for enzyme %s." % enz ofh = open("rebase.res", "w") for enzname in enz_dict.keys(): pat = enz_dict[enzname]

86

Chapter 12. Files

if isexact(pat): print_matches(ofh, enzname, findpos(seq, pat)) ofh.close()

Although it is possible to use the write or writelines methods of the le object, we have shown in the above example how to pass a le object to the print statement. Which syntax you will use in your own code, is a question of taste. But the code could be difcult to read if you mix them.

12.4. Design problems

Here is the complete program to nd all occurrences of restriction sites from a set of enzymes in a sequence. The enzyme data are read from a le and the results are also stored in a le.
def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches def isexact(pat): for c in pat.upper(): if c not in ATGC: return 0 return 1 def print_matches(ofh, enz, matches): if matches: print >>ofh, "Enzyme %s matches at:" % enz, for m in matches: print >>ofh, m, print >>ofh else: print >>ofh, "No match found for enzyme %s." % enz def get_site_only(pat): newpat = "" for c in pat: if c.isalpha(): newpat += c return newpat def read_rebase(filename): enz_dict = {} infh = open(filename) for line in infh.xreadlines(): fields = line.split()

87

Chapter 12. Files

name = fields[0] pat = fields[2] enz_dict[name] = get_site_only(pat) infh.close() return enz_dict def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches

seq = """atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaa tttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaacagctttaccatgcggtg ctgcgccatcaggtggtgtttctacgcgatcaagctattacgccgcagcagcaacgcgcgctggc ccagcgttttggcgaattgcatattcaccctgtttacccgcatgccgaaggggttgacgagatca tcgtgctggatacccataacgataatccgccagataacgacaactggcataccgatgtgacattt attgaaacgccacccgcaggggcgattctggcagctaaagagttaccttcgaccggcggtgatac gctctggaccagcggtattgcggcctatgaggcgctctctgttcccttccgccagctgctgagtg ggctgcgtgcggagcatgatttccgtaaatcgttcccggaatacaaataccgcaaaaccgaggag gaacatcaacgctggcgcgaggcggtcgcgaaaaacccgccgttgctacatccggtggtgcgaac gcatccggtgagcggtaaacaggcgctgtttgtgaatgaaggctttactacgcgaattgttgatg tgagcgagaaagagagcgaagccttgttaagttttttgtttgcccatatcaccaaaccggagttt caggtgcgctggcgctggcaaccaaatgatattgcgatttgggataaccgcgtgacccagcacta tgccaatgccgattacctgccacagcgacggataatgcatcgggcgacgatccttggggataaac cgttttatcgggcggggtaa""".replace("\n","").upper()

enzdict = read_rebase("../data/rebase.dat") ofh = open("rebase.res", "w") for enzname in enzdict.keys(): pat = enzdict[enzname] if isexact(pat): print_matches(ofh, enzname, findpos(seq, pat)) ofh.close()

We have written this program step by step by broadening the problem at each step, but we have never looked at the global design of the program. Therefore we have used two strategies while reading and writing. In the reading step all data are read at once, whereas the results are written each time we got them. In biological problems it is often necessary to handle lots of data and this can lead to problems if you try to handle all of them in memory, as we have done with the enzyme data. Sometimes we need all the data to solve a problem,

88

Chapter 12. Files

but often we can treat them each at a time. If the dataset is small, the choice of the strategy does not change a lot, but if you handle a large amount of data, it is worth asking whether you really need it all in memory? In our restriction problem it is not really necessary to store all the enzyme data in the memory. It is possible to treat enzyme by enzyme with one loop that reads the enzyme data, processes the sequence and writes the result.

Example 12.2. Restriction of a DNA sequence

def isexact(pat): for c in pat.upper(): if c not in ATGC: return 0 return 1 def print_matches(ofh, enz, matches): if matches: print >>ofh, "Enzyme %s matches at:" % enz, for m in matches: print >>ofh, m, print >>ofh else: print >>ofh, "No match found for enzyme %s." % enz def get_site_only(pat): newpat = "" for c in pat: if c.isalpha(): newpat += c return newpat def findpos(seq, pat): matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches

seq = """atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaa tttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaacagctttaccatgcggtg ctgcgccatcaggtggtgtttctacgcgatcaagctattacgccgcagcagcaacgcgcgctggc ccagcgttttggcgaattgcatattcaccctgtttacccgcatgccgaaggggttgacgagatca tcgtgctggatacccataacgataatccgccagataacgacaactggcataccgatgtgacattt attgaaacgccacccgcaggggcgattctggcagctaaagagttaccttcgaccggcggtgatac gctctggaccagcggtattgcggcctatgaggcgctctctgttcccttccgccagctgctgagtg ggctgcgtgcggagcatgatttccgtaaatcgttcccggaatacaaataccgcaaaaccgaggag gaacatcaacgctggcgcgaggcggtcgcgaaaaacccgccgttgctacatccggtggtgcgaac gcatccggtgagcggtaaacaggcgctgtttgtgaatgaaggctttactacgcgaattgttgatg tgagcgagaaagagagcgaagccttgttaagttttttgtttgcccatatcaccaaaccggagttt caggtgcgctggcgctggcaaccaaatgatattgcgatttgggataaccgcgtgacccagcacta

89

Chapter 12. Files

tgccaatgccgattacctgccacagcgacggataatgcatcgggcgacgatccttggggataaac cgttttatcgggcggggtaa""".replace("\n","").upper()

ifh = open("../data/rebase.dat") ofh = open("rebase.res", "w") line = ifh.readline() while line: fields = line.split() name = fields[0] pat = get_site_only(fields[2]) if isexact(pat): print_matches(ofh, name, findpos(seq, pat)) line = ifh.readline() else: line = ifh.readline() ofh.close() ifh.close()

Important
Notice that there are two les opened at the same time during the loop. Figure 12.2 shows three owchart comparing our rst solution, with the new design and a version that rst reads all data in, handles them and writes all results at the end.

90

Chapter 12. Files

Figure 12.2. Flowchart of the processing of the sequence

open the input file

open the input file

open the input file

open output file

close the input file

close the input file

open output file close the output file

close the input file open output file

close the output file

close the output file

Exercise 12.1. Multiple sequences for all enzymes

Think about alternative solutions, their advantages and problems if you need to process more than one sequence for all restriction enzymes.

12.5. Documentation strings

The code of the program written in this section (Example 12.2) lls about one page. If you put it away for a while and look on it a week later, it might be difcult to remember all the choices made during the development and it might take a lot of time to understand it. This fact underlines the importance to document programs.

91

Chapter 12. Files

For this purpose you can use comments that are ignored by the interpreter. In Python, all lines starting with a # character are comments. But Python has a more powerful mechanism to document programs. If the rst statement of a function or le is a string, this string is used as documentation string which is available by the pydoc command in a shell or the help function of the interpreter. Here is a documented version of our restriction program.
#!/usr/bin/python """ program that finds all occurences of restriction enzymes in a sequence at the moment the sequence is contained in the seq variable todo: reads the sequence from a field enzyme data are read from the file: rebase.dat and results are written to file: restrict.res restriction: the program finds only occurences of EXACT restriction patterns """ def isexact(pat): """ checks if a pattern is an exact DNA sequence all strings containing only the following characters are considered to be an exact DNA sequence: atgcATGC """ for c in pat.upper(): if c not in ATGC: return 0 return 1 def print_matches(ofh, enz, matches): """ write starting positions of occurences of a restriction site to a file positions in the sequence starts by 1 """ if matches: print >>ofh, "Enzyme %s matches at:" % enz, for m in matches: print >>ofh, m+1, print >>ofh else: print >>ofh, "No match found for enzyme %s." % enz def get_site_only(pat): """ clean a pattern string read from Rebase to obtain only the recognition site sequence """ newpat = "" for c in pat:

92

Chapter 12. Files

if c.isalpha(): newpat += c return newpat def findpos(seq, pat): """ find all occurences of restriction site pattern in a sequence RETURN VALUE: a list containing the start positions of all occurences RESTRICTION: can only process exact patterns, searching with ambiguous pattern strings would not produce an error """ matches = [] current_match = seq.find(pat) while current_match != -1: matches.append(current_match) current_match =seq.find(pat, current_match+1) return matches # starting main stream seq = """atgagtgaacgtctgagcattaccccgctggggccgtatatcggcgcacaaa tttcgggtgccgacctgacgcgcccgttaagcgataatcagtttgaacagctttaccatgcggtg ctgcgccatcaggtggtgtttctacgcgatcaagctattacgccgcagcagcaacgcgcgctggc ccagcgttttggcgaattgcatattcaccctgtttacccgcatgccgaaggggttgacgagatca tcgtgctggatacccataacgataatccgccagataacgacaactggcataccgatgtgacattt attgaaacgccacccgcaggggcgattctggcagctaaagagttaccttcgaccggcggtgatac gctctggaccagcggtattgcggcctatgaggcgctctctgttcccttccgccagctgctgagtg ggctgcgtgcggagcatgatttccgtaaatcgttcccggaatacaaataccgcaaaaccgaggag gaacatcaacgctggcgcgaggcggtcgcgaaaaacccgccgttgctacatccggtggtgcgaac gcatccggtgagcggtaaacaggcgctgtttgtgaatgaaggctttactacgcgaattgttgatg tgagcgagaaagagagcgaagccttgttaagttttttgtttgcccatatcaccaaaccggagttt caggtgcgctggcgctggcaaccaaatgatattgcgatttgggataaccgcgtgacccagcacta tgccaatgccgattacctgccacagcgacggataatgcatcgggcgacgatccttggggataaac cgttttatcgggcggggtaa""".replace("\n","").upper() # open the input and output file ifh = open("../data/rebase.dat") ofh = open("rebase.res", "w") # process enzyme by enzyme line = ifh.readline() while line: # extract enzyme name and pattern string fields = line.split() name = fields[0] pat = get_site_only(fields[2]) # find pattern and write the result if isexact(pat): print_matches(ofh, name, findpos(seq, pat)) # get the next enzyme

93

Chapter 12. Files

line = ifh.readline() # close opened files ofh.close() ifh.close()

and the result of the documentation formatted with pydoc

Python Library Documentation: module restrict_documented NAME restrict_documented - Program that find all occurrences of restriction enzymes in a sequence

FILE FILE /home/maufrais/cours_python/code/restrict_documented.py DESCRIPTION At the moment the sequence is contained in the seq variable Enzyme data are red from the file: rebase.dat and results are written to file: restrict.res restriction: the program finds only occurrences of EXACT restriction patterns FUNCTIONS get_site_only(pat) clean a pattern string red from Rebase to obtain only the recognition site sequence isexact(pat) checks if a pattern is an exact DNA sequence all strings containing only the following characters are considered to be an exact DNA sequence: atgcATGC print_matches(ofh, enz, matches) write starting positions of occurrences of a restriction site to a file positions in the sequence starts by 1 findpos(seq, pat) find all occurrences of restriction site pattern in a sequence RETURN VALUE: a list containing the start positions of all occurrences RESTRICTION: can only process exact patterns, searching with ambiguous pattern strings would not produce an error DATA __file__ = ./restrict_documented.pyc

94

Chapter 12. Files

__name__ = restrict_documented fields = [ZraI, 3, "GACGTC", 0, !, AatII, >I, 136] ifh = <closed file ../data/rebase.dat, mode r> line = name = ZraI ofh = <closed file rebase.res, mode w> pat = GACGTC seq = ATGAGTGAACGTCTGAGCATTACCCCGCTGGGGCCGTATATCGGCGC...CATCGGGCGACGA...

95

Chapter 12. Files

96

Chapter 13. Recursive functions

Chapter 13. Recursive functions

13.1. Recursive functions denitions
In Chapter 10 we have introduced a way to execute repetitive tasks, but we have also mentionned that repetition can be done without the special for and whileloop statements. This chapter will explain this in detail. Translate a cds sequence into its corresponding protein. Lets start with the example of the cds translation. In Chapter 10 we have already written functions that solve this task using either for (Example 10.1) or while (Example 10.3). Lets remind the pseudocode of the examples: INPUT: a cds sequence cds and a genetic code code OUTPUT: the translated cds sequence prot 1. prot <- empty string 2. as long as there are still codons to translate: a. codon <- get next codon from sequence cds b. lookup the corresponding amino acid of codon codon in genetic code code and add it to prot 3. return prot

1. prot <- empty string 2. for each codon in sequence cds: add the corresponding amino acid of the current codon to sequence prot using genetic code code 3. return prot In both examples the translation is dened as: the concatenation of the rst codon of the sequence and the translation of the cds sequence without this codon or in more mathematical terms: translation(cds) = code(cds[:3]) + translation(cds[3:]) This is a recursive denition of the translation function.

97

Chapter 13. Recursive functions

Recursive functions
A recursive function is a function that use itself during the calculation procedure.

Recursive denition
A recursive denition of a function is a denition that uses the function itself in the denition. It is important to notice that we need a terminal condition otherwise the recursion would never stop. In our case the recursion stops if there are no more codons to translate: the protein sequence of the an empty cds sequence is empty or in more mathematical terms: translation("") = "" Therefore the pseudocode can be written as follow without using loop structures: INPUT: a cds sequence cds and a genetic code code OUTPUT: the translated sequence prot 1. if cds is empty: a. *** This is the terminal condition *** b. return empty string 2. otherwise: a. *** This is the recursion *** b. codon <- rst codon of sequence cds c. return the concatenation of the corresponding amino acid of the rst codon of sequence cds in genetic code code and the translation of the rest of the cds sequence

98

Chapter 13. Recursive functions

and implemented in Python like that:

code = {ttt: F, tct: S, tat: Y, tgt: C, ttc: F, tcc: S, tac: Y, tgc: C, tta: L, tca: S, taa: *, tga: *, ttg: L, tcg: S, tag: *, tgg: W, ctt: L, cct: P, cat: H, cgt: R, ctc: L, ccc: P, cac: H, cgc: R, cta: L, cca: P, caa: Q, cga: R, ctg: L, ccg: P, cag: Q, cgg: R, att: I, act: T, aat: N, agt: S, atc: I, acc: T, aac: N, agc: S, ata: I, aca: T, aaa: K, aga: R, atg: M, acg: T, aag: K, agg: R, gtt: V, gct: A, gat: D, ggt: G, gtc: V, gcc: A, gac: D, ggc: G, gta: V, gca: A, gaa: E, gga: G, gtg: V, gcg: A, gag: E, ggg: G } def rectranslate(cds, code): if cds == "": return "" else: codon = cds[:3] return code[codon] + rectranslate(cds[3:], code) print rectranslate("atgattgctggt", code)

13.2. Flow of execution of recursive functions

At rst glance recursive functions look a little bit strange because it seems that we use something that we have not dened. But remember that the statements in the body of a function are only executed when the function is called. At the denition step, the function is just added in the current namespace. Therefore it will be dened when we call it in the body. Writing recursive functions is sometimes difcult, because we doubt that we have dened a function executing the specic task, when we call the it . Therefore it is useful to remember this fact during the development of recursive functions. Stack diagram of recursive functions. Lets have a deeper look at the ow of execution of recursive functions. Figure 13.1 shows the stack diagram of function calls.

99

Chapter 13. Recursive functions

Figure 13.1. Stack diagram of recursive function calls

rectranslate("atgattgctggt", code)

cds

"" return

rectranslate
code ""

cds

"ggt"

rectranslate
code return "G"

cds

"gctggt"

rectranslate
code

return

"AG"

{ ttt: F , ..., ggg: G }

cds

"attgctggt" return

rectranslate
code "IAG"

cds

"atgattgctggt" return "MIAG"

rectranslate
code

Figure 13.1 shows that the translate function is called with a cds sequence of decreasing length and the local function namespaces are piled up until the terminal condition is reached.
=> rectranslate("atgattgctggt", code) ===> rectranslate("attgctggt", code) =====> rectranslate("gctggt", code) =======> rectranslate("ggt", code) =========> rectranslate("", code) =========> return "" =======> return "G" =====> return "AG" ===> return "IAG" => return "MIAG"

100

Chapter 13. Recursive functions

Important
Recursive function calls can take a lot of memory space because for each recursive call the local namespace for the given function call has to be stored. It is important to pay attention on this fact because memory space is a limited resource on a computer.

13.3. Recursive data structures

In the Chapter 11, we have worked with examples using nested list. If we examine them a little bit more, we will see that a nested list can be dened recursively.

Nested list
A nested list is an ordered collection of elements that are: either items or nested lists The items dene the content, whereas the lists dene the structure of the collection. How can we use this? Lets use phylogenetic trees as examples for such a recursive list structure. Figure 13.2 shows an example of a phylogenetic tree topology that can be implemented in Python as follows:
>>> tree = [[[[human, chimpanzee], gorilla], orang-utan], gibbon]

Figure 13.2 shows the representation of tree.

101

Chapter 13. Recursive functions

Figure 13.2. A phylogenetic tree topology

human

chimpanzee

gorilla

orangutan

gibbon

Figure 13.3. Tree representation using a recursive list structure

tree

human

chimpanzee

gorilla

orangutan

gibbon

Working with recursive structures. Assume that we would do something for each branching point in the above tree. A simple traversal using for or while is no more possible, because both loop structures handle only one level of the list. But with recursive functions that we have introduced in Chapter 13, we can do such things with the following strategy:

Procedure 13.4. tree_traversal(tree)

INPUT: a tree tree 1. do what you have to do when you enter the branching point 2. if tree is a list: a. *** recursion step *** b. for each element of tree:

102

Chapter 13. Recursive functions

tree_traversal(element) 3. otherwise: *** end condition of the recursion, here the leaves of the tree *** 4. do what you have to do when you leave the branching point Doing something when you enter the branching point, is also known as preorder traversal of a tree, whereas doing something when leaving a branching point is also called postorder traversal of a tree. We will give two example applications of this strategy for our tree. The rst one prints all species of the trees:
tree = [[[[human, chimpanzee], gorilla], orang-utan], gibbon] import types def print_species(tree): if type(tree) is types.ListType: for child in tree: print_species(child) else: print tree

>>> print_species(tree) human chimpanzee gorilla orang-utan gibbon

and the second one that prints for a tree or a binary tree 2 the two populations that are split at this point.
tree = [[[[human, chimpanzee], gorilla], orang-utan], gibbon] import types def splits(tree): if type(tree) is types.ListType: all_leaves = [] for child in tree: child_leaves = splits(child) print child_leaves, all_leaves += child_leaves
A binary tree is a tree with maximal two subtrees for each branching point.

103

Chapter 13. Recursive functions

print return all_leaves else: return [ tree ]

def binary_splits(tree): if type(tree) is types.ListType: left = binary_splits(tree[0]) right = binary_splits(tree[1]) print left, right return left + right else: return [ tree ]

>>> splits(tree) [human] [chimpanzee] [human, chimpanzee] [gorilla] [human, chimpanzee, gorilla] [orang-utan] [human, chimpanzee, gorilla, orang-utan] [gibbon] [human, chimpanzee, gorilla, orang-utan, gibbon]

Finally, we will show the code that separates the work to do from the tree traversal itself by functions. The traversal function tree_traversal do not change for the task. The user has only to redene the functions do_prework, do_postwork and do_leafwork.
import types def do_prework(node): print "prework:", node def do_postwork(node): print "postwork:", node def is_leaf(node): return type(node) is not types.ListType def do_leafwork(leaf): print "that is a leaf:", leaf def tree_traversal (node): do_prework(node) if not is_leaf(node): for child in node: tree_traversal(child)

104

Chapter 13. Recursive functions

else: do_leafwork(node) do_postwork(node)

>>> tree=[[[[human, chimpanzee], gorilla], orang-utan], gibbon] >>> tree_traversal(tree) prework: [[[[human, chimpanzee], gorilla], orang-utan], gibbon] prework: [[[human, chimpanzee], gorilla], orang-utan] prework: [[human, chimpanzee], gorilla] prework: [human, chimpanzee] prework: human that is a leaf: human postwork: human prework: chimpanzee that is a leaf: chimpanzee postwork: chimpanzee postwork: [human, chimpanzee] prework: gorilla that is a leaf: gorilla postwork: gorilla postwork: [[human, chimpanzee], gorilla] prework: orang-utan that is a leaf: orang-utan postwork: orang-utan postwork: [[[human, chimpanzee], gorilla], orang-utan] prework: gibbon that is a leaf: gibbon postwork: gibbon postwork: [[[[human, chimpanzee], gorilla], orang-utan], gibbon]

105

Chapter 13. Recursive functions

106

Chapter 14. Exceptions

Chapter 14. Exceptions

14.1. General Mechanism
Exceptions are a mechanism to handle errors during the execution of a program. An exception is raised whenever an error occurs:

Example 14.1. Filename error

>>> f = open(my_fil) Traceback (most recent call last): File "<stdin>", line 1, in ? IOError: [Errno 2] No such file or directory: my_fil

An exception can be caught by the code where the error occurred:

try: f = open(my_fil) except IOError, e: print e

Variable e contains the cause of the error:

[Errno 2] No such file or directory: my_fil

14.2. Python built-in exceptions

107

Chapter 14. Exceptions

Python predenes several exceptions (Figure 14.1).

Figure 14.1. Exceptions class hierarchy

Exception SystemError StandardError ArithmeticError FloatingPointError OverflowError ZeroDivisionError AssertionError AttributeError EnvironmentError IOError OSError WindowsError EOFError ImportError KeyboardInterrupt LookupError IndexError KeyError MemoryError NameError RunTimeError SyntaxError UnboundLocalError NotImplementedError TabError IndentationError SystemError TypeError ValueError UnicodeError

AttributeError: when you attempt to access a non-existing attribute (method or variable) of an object. NameError: failure to nd a global name (module, ...) IndexError, KeyError: occurs when attempting to access either an out-of-range index in a list or a non-existing key in a dictionary TypeError: passing an inappropiate value to an operation TabError, IndentationError: two kinds of SyntaxError

108

Chapter 14. Exceptions

14.3. Raising exceptions

You can also raise an exception in your code, if you consider that the program should be interrupted:
if something_wrong: raise Exception

You can associate a message to the raise statement:

if something_wrong: raise Exception, " something went wrong"

Example 14.2. Raising an exception in case of a wrong DNA character

def check_dna(dna, alphabet=atgc): """ using exceptions """ for base in dna: if base not in alphabet: raise ValueError, "%s not in %s" % (base, alphabet) return 1

14.4. Dening exceptions

Python provides a set of pre-dened exception classes that you can specialize by sub-classing to dene specic exceptions for your application (Figure 14.1).

Go to
Since exceptions are dened as classes and by inheritance, you will need some knowledge about classes in order to fully understand this section (see Chapter 17).

Example 14.3. Raising your own exception in case of a wrong DNA character
In the following code, you dene an exception AlphabetError that can be used when the sequence passed to the function does not correspond to the alphabet.
class AlphabetError(ValueError): pass def check_dna(dna, alphabet=atgc):

109

Chapter 14. Exceptions

""" using exceptions """ for base in dna: if base not in alphabet: raise AlphabetError, "%s not in %s" % (base, alphabet) return 1

Denition of a new exception in category ValueError: AlphabetError is a class, that is a subclass of class ValueError. The only statement present in class AlphabetError is pass since AlphabetError does not dene any new behaviour: it is just a new class name.

Example 14.4. Exceptions dened in Biopython

Some Biopython modules dene their own exceptions, such as: ParserFailureError (GenBank package) BadMatrix (SubsMat package)

110

Chapter 15. Modules and packages in Python

Chapter 15. Modules and packages in Python

15.1. Modules
A module is a component providing Python denitions of functions, variables or classes... all corresponding to a given specic thema. All these denitions are contained in a single Python le. Thanks to modules, you can reuse ready-to-use denitions in your own programs. Python also encourages you to build your own modules in a rather simple way.

15.1.1. Using modules

In order to use a module, just use the import statement. Let us take an example. Python comes with numerous modules, and a very useful one is the sys module (sys stands for "system"): it provides information on the context of the run and the environment of the Python interpreter. For instance, consider the following code:
#!/local/bin/python import sys print "arguments: ", sys.argv

Say that you stored in a modexa.py le, and that you run it like this:
./modexa.py 1 a seq.fasta

This will produces the following output:

arguments: [./modexa.py, 1, a, seq.fasta]

Explanation: By using the argv variable dened in the sys module, you can thus access to the values provided on the command line when launching the program. As shown in this example, the access to this information is made possible by: importing the module through the import statement, which provides access to the modules denitions using the argv variable dened in the module by a qualied name: sys.argv.

111

Chapter 15. Modules and packages in Python

You may also select specic components from the module:

from sys import argv print "arguments: ", argv

In this case, you only import one denition (the argv variable) from the sys module. The other denitions are not loaded.

15.1.2. Building modules

You build your own module by creating a Python le. For instance, if the le ValSeq.py contains the following code (adapted from the Biopython ValSeq module):

Example 15.1. A module

# file Valseq.py valid_sequence_dict = { "P1": "complete protein", \ "F1": "protein fragment", "DL": "linear DNA", "DC": "circular DNA", \ "RL": "linear RNA", "RC":"circular RNA", "N3": "transfer RNA", \ "N1": "other" } def find_valid_key(e): for key,value in valid_sequence_dict.items(): if value == e: return key

you can use it by loading it:

import ValSeq

where ValSeq is the module name. You can then access to its denitions, which may be variables, functions, classes, etc...:
>>> print ValSeq.valid_sequence_dict[RL] linear RNA >>> ValSeq.find_valid_key("linear RNA") RL

15.1.3. Where are the modules?

Modules are mainly stored in les that are searched:

112

Chapter 15. Modules and packages in Python

in your current working directory, in PYTHONHOME, where Python has been installed, in a path, i.e a colon (:) separated list of le paths, stored in the environment variable PYTHONPATH. You can check this path through the sys.path variable. Files may be: Python les, sufxed by .py (when loaded for the rst time, compiled version of the le is stored in the corresponding .pyc le), dened as C extensions, built-in modules linked to the Python interpreter.

Exercise 15.1. Locating modules

Sometimes, it is not enough to use pydoc or help. Looking at the source code can bring a better understanding, even if you should of course never use undocumented features. Browse the directory tree PYTHONHOME/site-packages/Bio/.

15.1.4. How does it work?

When importing a module, the interpreter creates a new namespace, in which the Python code of the modules le is run. The interpreter also denes a variable (such as sys, ValSeq, ...) that refers to this new namespace, by which the namespace becomes available to your program (Figure 15.2).

Figure 15.1. Module namespace

import sys

current namespace

...
sys

argv path

...

113

Chapter 15. Modules and packages in Python

open A module is loaded only once, i.e, a second import statement will neither re-execute the code inside the module (see Python reload statement in the reference guides), nor will it re-create the corresponding namespace. When selecting specic denitions from a module:
>>> from ValSeq import find_valid_key >>> find_valid_key("linear RNA") RL

the other components stay hidden. As illustrated in Figure 15.2, no new namespace is created, the imported denition is just added in the current name space.

Figure 15.2. Loading specic components

import ValSeq from ValSeq import find_valid_key

current namespace

current namespace

ValSeq

find_valid_key

valid_sequence_dict find_valid_key

This can causes errors if the denition that is imported needs to access to other denitions of the module, e.g:
>>> print valid_sequence_dict[RL] NameError: name valid_sequence_dict is not defined >>> print ValSeq.valid_sequence_dict[RL] NameError: name ValSeq is not defined

You can also load "all" the components from a module, which makes them available directly into your code:
>>> from ValSeq import * >>> find_valid_key("linear RNA")

114

Chapter 15. Modules and packages in Python

You probably did this many times in order to use the string modules denitions, right? The result of:
>>> from string import *

is that all the denitions of the module are copied in your current namespace.

Caution
Be aware of potential names collision: for instance, if you current namespace contains a denition of a variable called, say: count, it will be destroyed and overloaded by the string modules denition of the count function.

Caution
You can restrict the components being imported by an import * statement. The __all__ variable, also used for packages (Section 15.2), can explicitly list the components that should be directly accessible (see Exercise 15.4).

15.1.5. Running a module from the command line

When the le of a module is run from the command line (instead for being imported):
% python ValSeq.py

the module does not behaves like a module anymore. It is, instead, run within the default __main__ module (i.e not the ValSeq module):
% python -i ValSeq.py >>> ValSeq.find_valid_key("linear RNA") NameError: name ValSeq is not defined >>> find_valid_key("linear RNA") DL

For this reason, the code executed when the module is loaded (e.g: either with import or from the command line) can be made dependent of its current name by testing this name. The current modules name is stored in a special purpose variable __name__:
if __name__ == __main__: # statements that you want to be executed only when the # module is executed from the command line # (not when importing the code by an import statement) print find_valid_key("linear RNA")

115

Chapter 15. Modules and packages in Python

15.2. Packages
A package is a set of modules or sub-packages. A package is actually a directory containing either .py les or sub-directories dening other packages. The dot (.) operator is used to describe a hierarchy of packages and modules. For instance, the module Bio.WWW.ExPASy is located in the le PYTHONHOME/site-packages/Bio/WWW/ExPASy.py. This module belongs to the Bio.WWW package located into the PYTHONHOME/site-packages/Bio/WWW/ directory.

15.2.1. Loading
When loading a package, the __init__.py le is executed. If the __init__.py denes classes, functions, etc... they become available at once, as shown in the following example:

Example 15.2. Using the Bio.Fasta package

>>> >>> >>> >>> >>> >>> import Bio.Fasta handle = open("data/ceru_human.fasta") it = Bio.Fasta.Iterator(pin, Bio.Fasta.SequenceParser()) seq = it.next() print seq.seq it.close()

However, loading a package does not automatically load the inner modules. For instance, even though the Bio.Fasta package directory contains the following les:
% ls Bio/Fasta FastaAlign.py

FastaAlign.pyc

__init__.py

__init__.pyc

this does not imply that importing the Bio.Fasta package loads the Bio.Fasta.FastaAlign module:
>>> import Bio.Fasta >>> Bio.Fasta.FastaAlign.parse_file("data/ceru_human.fasta") AttributeError: module object has no attribute FastaAlign

Issuing:
>>> from Bio.Fasta import *

will however load the Bio.Fasta.FastaAlign, because this module is mentioned in the __all__ attribute in the Bio/Fasta/__init__.py le:

116

Chapter 15. Modules and packages in Python

__all__ = [ FastaAlign, ]

Other attributes of interest for packages and modules: __name__ __path__ __file__

Exercise 15.2. Bio.SwissProt package

Which import statements are necessary to make the following code work?
expasy = ExPASy.get_sprot_raw(CERU_HUMAN) sp = SProt.Iterator(expasy, SProt.RecordParser()) record = sp.next() print record.keywords

???

Exercise 15.3. Using a class from a module

Why does the following code issue an error?
from Bio.SubsMat import FreqTable dict = ... # whatever f = FreqTable(dict, COUNT) TypeError: module object is not callable

???

Exercise 15.4. Import from Bio.Clustalw

Why does the following code not work?
from Bio.Clustalw import *

117

Chapter 15. Modules and packages in Python

a=ClustalAlignment() NameError: name ClustalAlignment is not defined

???

15.3. Getting information on available modules and packages

You can use the help to get the list of available modules:
>>> help("modules") Please wait a moment while I gather a list of all available modules...

To search through the documentation of modules for a specic word, for instance "SProt", also use the help command like this:
>>> help("modules SProt") Here is a list of matching modules. Bio.SwissProt.KeyWList - KeyWList.py Bio.SwissProt.SProt - SProt.py Bio.SwissProt (package) Enter any module name to get more help.

The sys also contains the dictionary of loaded modules.

118

Chapter 16. Scripting

Chapter 16. Scripting

16.1. Using the system environment: os and sys modules
There are modules in the Python library that help you to interact with the system. The sys module. The sys module provides an interface with the Python interpreter: you can retrieve the version, the strings displayed as prompt (by default: >>> and ...), etc... You can nd the arguments that were provided on the command line:
% python -i prog.py myseq.fasta >>> import sys >>> sys.argv [prog.py, myseq.fasta]

The le handle for the standard input, output and error are accessible from the sys module:
>>> sys.stdout.write("a string\n") a string >>> sys.stdin.read() a line another line a line\nanother line\n

You have to enter a Ctl-D here to end the input. The os module. This module is very helpful to handle les and directories, processus, and also to get environment variables (see environ dictionary). One of the most useful component is the os.path module, that you use to get informations on les:
>>> import os.path >>> os.path.exists(myseq.fasta) 1 >>> os.path.isfile(myseq.fasta) 1 >>> os.path.isdir(myseq.fasta) 0 >>> os.path.basename(/local/bin/perl) perl

119

Chapter 16. Scripting

Exercise 16.1. Basename of the current working directory

Write the statements to display the basename of the current working directory. The os.path module provides a method: walk that enables to walk in all the directories from a starting directory and to call a given function on each.

Example 16.1. Walking subdirectories

The following code displays for each directory its name and how many les it contains:
>>> def f(arg, dirname, fnames): ... print dirname, ": ", len(fnames) >>> os.path.walk(., f, None)

The arguments of function f must be: dirname, which is the name of the directory, and fnames which is a list containing the names of the les and subdirectories in dirname. arg is a free parameter, that is passed to walk (here: None).

Exercise 16.2. Finding les in directories

Find a le of a given name and bigger than a given size in a directory and its sub-directories. Only consider les, not directories.

16.2. Running Programs

You can run external programs from a Python program. There are three major tasks to perform in order to run programs from a script: Building the command line. Testing for success and checking for potential errors. Getting results.

120

Chapter 16. Scripting

Example 16.2. Running a program (1)

The simplest way to run a program is by using the system of the os module. The result of the program will be printed on the standard output, which is normally the screen. The return value reports about the success or failure of the execution.
import os cmd="golden swissprot:malk_ecoli" status = os.system(cmd) print "Status: ", status

Another Python module, commands, enables to store the result of the execution in a string:
import commands cmd="golden swissprot:malk_ecoli" output = commands.getoutput(cmd) print "Output: ", output

To get both result and status:

import commands cmd="golden swissprot:malk_ecoli" status, output = commands.getstatusoutput(cmd) print "Output: ", output print "Status: ", status

Example 16.3. Running a program (2)

A more elaborate but lower level interface to run commands is provided by the popen function from the os module. The following script runs a program that fetches a Swissprot entry given its entry name, and prints it on the screen.
import os import string cmd="golden swissprot:malk_ecoli" handle = os.popen(cmd, r) print string.join(handle.readlines()) handle.close()

Builds the command line with a program name and the arguments. Runs the command and stores a handle in the handle variable. A handle for a command is the same kind of objects as a le handle: you open it (with the popen command, read from it, and close it. Reads all the lines from the handle, and prints the joint result.

121

Chapter 16. Scripting

If the program takes time and if you wish to read the result step by step as long as results show up, you can do like this:
import os import sys cmd="blastall -i " + sys.argv[1] + " -p blastp -d swissprot" handle = os.popen(cmd, r, 1) for line in handle: print line, handle.close()

What if the entry name does not have a corresponding entry in the database? Let us try the following code:
import os import sys import string cmd="golden swissprot:" + sys.argv[1] handle = os.popen(cmd, r) print string.join(handle.readlines()) status = handle.close() if status is not None: print "An error occured: ", status

Takes the entry name from the Python command line arguments, by using the sys module argv variable. If the provided entry name is invalid, the program returns a non zero value, that is returned by the close function. If you wish to get the complete error message from the program, use the popen3 function:
import os import sys cmd="golden swissprot:" + sys.argv[1] tochild, fromchild, childerror = os.popen3(cmd, r) err = childerror.readlines() if len(err) > 0: print err else: print fromchild.readlines()

In this script, the call returns three objects: one to get results: fromchild (standard output), one to write to the program - on its standard input - (for instance when the program is prompting for a value (see Example 16.4): tochild, and one to access to the standard error le descriptor of the program.

122

Chapter 16. Scripting

Example 16.4. Running a program (3)

The next example shows how to run an interactive program, both reading the output and writing on the standard input to answer programs questions. The Phylip dnapars program, once having read the input le containing an alignment, that is always called infile, waits for the user to enter y and Return to proceed. The following script runs dnapars after some le cleaning.
import popen2 import os.path cmd = "dnapars" if os.path.exists(treefile): os.unlink(treefile) if os.path.exists(outfile): os.unlink(outfile) child = popen2.Popen3(cmd) print "PID: ", child.pid child.tochild.write("y\n") child.tochild.close() child.wait() print "".join(child.fromchild.readlines()) status = child.fromchild.close() if status is not None: print "status: ", status

Removes old dnapars output les. Use of the class Popen3 that stores the information about the run program. The return value: child, is an object representing the "child" processus, that has attributes to get the channels to communicate with the processus: a handle to write to the standard input of the processus (child.tochild), and a handle to read its output (child.fromchild). See Chapter 17 for more informations on classes. Answers to program prompt. This statements helps in cleaning the processus after completion. Reads results.

16.3. Parsing command line options with getopt

The getopt module helps script to parse a complex command line options. Example 16.5 shows an example.

Example 16.5. Getopt example

This example shows a piece of code for handling options in a program called, say, lteralig, taking an alignment and ltering some sites according to specic criteria. Options are of the form: -o1 value1 -o2 value2 where o1 and o2 are the names of the options and value1, value2 are their argument, which might be optional for some of the parameters. You can use this script this way, for instance:

123

Chapter 16. Scripting

filteralig -t0.7 -f2,3 align_file

import sys import getopt def usage(prog="filtersites"): print """ filteralig : filter sites in alignments filteralig [-ch] [-t <threshold>] [-f <frames>] [-i <cols>] <alignment> -h -c -t <threshold> -f <frames> print this message print colum numbers of the original alignment filter all colums with a conservation above <threshold> filter all codonpositions of frames possible values 1, 2, 3 for more than one use syntaxe: 1,2 filter this columns syntaxe: give a string with the column numbers separated by

-i <cols> , <alignment> """

the file has to be in clustalw format

o, a = getopt.getopt(sys.argv[1:], ct:f:i:h) opts = {} for k,v in o: opts[k] = v if opts.has_key(-h): usage(); sys.exit(0) if len(a) < 1: usage(); sys.exit("alignment file missing")

A usage function is very useful to help the user in case of an error on the command line. The rst parameter for the getopt function should be a string containing the actual arguments the script has been called with, not including the script name, available in sys.argv[0]. The second parameter is a string describing the expected options. The options string which is passed to getopt is here: ct:f:i:h. This means that the following options are available: c, t, f, i and h. When a : is added just after, this means that the option expects a value. For instance, the -t option requires a threshold value. See the usage!

124

Chapter 16. Scripting

The getopt function returns tuple, whose rst element is a list of (option, value) pairs. The second element is the list of program arguments left after the option list was stripped. Here, a lename for an alignment le is expected. Storing (option, value) pairs in a dictionary. If the user has entered a -h, help is printed. Has the user provided a lename ? If so, it is available in a[0].

16.4. Parsing
In Bioinformatics, parsing is very important, since it enables to extract informations from data les or to extract results produced by various analysis programs, and to make them available in your programs. For instance, a Blast parser will transform a text output into a list of hits and their alignment, that can be made available as a data structure, such as, for example, Biopython Bio.Blast.Record objects, that you can use in a Python program. The purpose of this section is not to present everything about parsing, but just to introduce some basic notions. Parsing means analyzing a text and producing structured data in a form that is useful for programs. It can be a list of strings, a set of classes instances, or just a boolean result: this depends on the needs and the parsing system you are using. An important aspect of parsing is the architecture that is used to process the text that you want to analyze. Parsing can be done by just reading a le or a stream line by line, and by looking for the occurrence(s) of a word, or a pattern. In Figure 16.1, lines are searched for a header pattern, or a hit pattern, and processed accordingly.

Figure 16.1. Manual parsing blast_report

Python program for line in file: if found header in line: process_header(line) elif hit(line): process_hit(line) elif hsp(line): process_hsp(line)

125

Chapter 16. Scripting

You can tell an external component which set of words you are interested in, by providing a description of set of words to be found. You then feed this component a document to parse, and wait him to tell you when these words are found as well as sending you the occurrences. This kind of system is said to be event-driven, and XML provides tools for such type of parsing.

Figure 16.2. Event-based parsing blast_report

header_handler(data) Engine hit_handler(data) description of required events or elements

hsp_handler(data)

You can describe the whole document by set of hierarchical subparts, and associate to each sub-part actions to be automatically executed by a parsing engine. Figure 16.3 shows such a system to parse a Blast report. A Blast report is described as a header followed by a list of hits. A hit is described as a score and a list of HSP, and a HSP is described as a score and a list of alignments. You dene these subparts by a set of rules, sometimes using patterns, in what is usually called a grammar, or even a decorated grammar, since you decorate each sub-part with an associated action. The lex and yacc system is such a parsing engine.

Figure 16.3. Parsing: decorated grammar

126

Chapter 16. Scripting

blast_report

Engine

decorated grammar blast: header, list of hits header: actions for header hit: score, list of hsps actions for hit hsp: score, alignment actions for hsp alignment: actions for alignment
You can have a parsing engine process your data according to a grammar, and returns a hierarchical data structure in a form that your program can understand (for instance, in Python, as a set of objects). The XML/DOM engine behaves like this.

Figure 16.4. Parsing result as a hierarchical document blast_report

Engine

data structure representing document

grammar

127

Chapter 16. Scripting

So, in all the cases, there is an engine driving the whole process, be it a simple loop or a specialized component. In this chapter, we will just do some "manual" parsing with patterns that are introduced in Section 16.5, as well as some event-driven parsing will be done as a practical work on abstract frameworks (see Exercise 18.2), and during the Web/XML course.

16.5. Searching for patterns.

16.5.1. Introduction to regular expressions
Regular expression is a term used in computer science to refer to the language that is used to describe patterns to be searched for in a text. The term comes from the computing languages theory where regular expressions are used to denote regular languages. In turn, regular languages are dened as the languages that can be recognized by a nite-state automaton (a topic that will be introduced in the algorithmic course). The aim of a pattern is to dene not only one word to be searched for, but a set of words. This denition is provided in a given language, depending on the system you are working with; the set of corresponding words also depends on the system. In the Unix shell, for instance:
ls s*

means list all the les beginning by s, and potentially followed by anything. The command:
ls s[ie]n*

means list all the les beginning by s, followed by either i or e, and followed by anything, including nothing. So, in the context of the ls function within a Unix shell, the set of words is dened as the les existing on the lesytem (as opposed, for instance, to the les stored on another computer, not being made available through a distributed lesystem). In the Prosite database, patterns describing protein domains are described, such as:
H-C-H-x(3)-H-x(3)-[AG]-[LM]

which represent the following set of amino-acid sequences: sequences begining by HCH, followed by 3 positions containing any amino-acid letter, followed by H, followed again by 3 free positions, followed by either A or G, followed by either L or M. As you can notice, the language to dene patterns is different from the Unix shell, as well as the set of corresponding words. Here, they are just sequences of amino-acids letters. In the grep Unix command, a command to search for patterns in les, although similar to the shell pattern syntax, there is a slight difference. Say that a le contains:

128

Chapter 16. Scripting

science s another

the following command:

grep s* file

will return ... all the lines, because s* means all the words beginning by 0 or any number of s. In the grep command, the set of words is composed of the lines of the le. So the term "set of words" must be understood in a broad sense. It must be understood also that the set is not actually generated, of course: it can be innite! Instead, an operational representation of this set is built, through a nite-state automaton. In SQL languages, you can also provide patterns:
select clone_name from CLONE where clone_id like [^A]%[02468]

means that you restrict the query to the clones whose identier does not begin with a A, is followed by any character N times, and ends with an even number (Sybase). While the set of words corresponding to a pattern is described by the given expression and the semantics of the system being used, the set of found words, also called occurrences, depends on data. So, occurrences are the words from the set of words that were actually found within data. Figure 16.5 summarizes the concepts.

Figure 16.5. Pattern searching

System semantics

expression

set of words (representation of)

occurrences of words in data

data

129

Chapter 16. Scripting

16.5.2. Regular expressions in Python

A detailed presentation of Python regular expressions is available here: Regular Expression HOWTO [http://pyhowto.sourceforge.net/regex/regex.html]. To get information about the re module, see pydoc, but also the sre module (Support for regular expressions), for which re is a wrapper. In Python, regular expressions are handled by a module: re:
>>> import re

Before searching for a pattern, you must rst compile it (this builds the "set of words", or rather the representation of this set):
>>> expression = [AP]{1,2}D >>> pattern = re.compile(expression)

pattern is a pattern object. You then issue a search, for instance in the small sequence seq, by a request to the pattern object:
>>> seq = "RPAD" >>> match = pattern.search(seq)

This establishes the matching, or correspondances, between the set of possible words and data. match is called a match object. To get the occurrences, you can ask the match object for the start and end of the match in the searched text:
>>> print match.start(), match.end(), seq[match.start():match.end()] 1 4 PAD

or the group:
>>> match.group(0) PAD

130

Chapter 16. Scripting

Figure 16.6 summarizes this system.

Figure 16.6. Python regular expressions

re.compile expression pattern object

match object pattern.search() pattern.match()

match.start() match.end() match.group(0)

data

Example 16.6. Searching for the occurrence of PS00079 and PS00080 Prosite patterns in the Human Ferroxidase protein
import sys import re from Bio.SwissProt import SProt sp = open(sys.argv[1]) iterator = SProt.Iterator(sp, SProt.SequenceParser()) seq = iterator.next().seq sp.close() PS00079 = G.[FYW].[LIVMFYW].[CST].{8,8}G[LM]...[LIVMFYW] pattern = re.compile(PS00079) match = pattern.search(seq.tostring()) print PS00079 print match.start(), match.end(), seq[match.start():match.end()]

The regular expression is stored in a string. The regular expression is compiled in a pattern. The compiled pattern is searched in the sequence. The result of the search is printed.

131

Chapter 16. Scripting

There are several methods to search: search and match, the difference being that match looks for a match at the beginning of the string. So, back to the example, the following statement:
match = pattern.match(seq.tostring())

would return a positive result only if the sequence begins by an occurrence of PS00079. A convenient feature enables to associate a name to sub-parts of the matched text:
import sys import re from Bio.SwissProt import SProt sp = open(sys.argv[1]) iterator = SProt.Iterator(sp, SProt.SequenceParser()) seq = iterator.next().seq sp.close() PS00080 = (?P<copper3>H)CH...H...[AG](?P<copper1>[LM]) pattern = re.compile(PS00080) match = pattern.search(seq.tostring()) print PS00080 print match.start(), match.end(), seq[match.start():match.end()] print copper type 3 binding residue: , match.group(copper3) print copper type 1 binding residue: , match.group(copper1)

The regular expression now contains 2 identiers: copper1 and copper3. You can print the sub-parts of the result identied by variables: copper1 and copper3. Shortcuts. The re module provides shortcuts to directly search for an expression without compiling the pattern into a pattern object:
>>> match = re.search([AP]{1,2}D, "RPAD") >>> match.group(0) PAD

You can also directly get the occurrences of a pattern object in a string:
>>> pattern = re.compile([AP]{1,2}D, ) >>> pattern.findall("RPAD") [PAD]

or even the occurences of an expression, without compiling the pattern:

132

Chapter 16. Scripting

>>> re.findall([AP]{1,2}D, "RPAD") [PAD]

Figure 16.7 summarizes re module objects and methods to perform pattern searching.

Figure 16.7. Python regular expressions: classes and methods summary

Expression Pattern object Match object Occurrences

compile() match(), search() split(), findall(), sub() match(), search() split(), findall(), sub() start(), end(), span()

pattern

group() re

Search modes. Text substitutions.

16.5.3. Prosite
This section presents the Prosite classes in Biopython, which have a common interface with the Python pattern module.

16.5.3.1. Prosite Dictionary

Biopython denes several dictionaries to access biological databases. Having a dictionary means that you can fetch an entry by:

133

Chapter 16. Scripting

entry = prosite[PS00079]

For this to work, you rst need to create the dictionary:

prosite = Bio.Prosite.ExPASyDictionary()

As you can guess by the name of the module, you actually fetch the Prosite entry on the Web. You could also fetch the Prosite entry from a local database with the golden program (see ???). The entry fetched above is actually a string. In order to have the dictionary return a record, you must rather create it like this:
prosite = Bio.Prosite.ExPASyDictionary(parser=Bio.Prosite.RecordParser())

16.5.3.2. Prosite patterns

The Bio.Prosite package denes a Pattern class that enables to create patterns which may be searched for in sequences objects, as in the re Python module for regular expressions. The result of a search is a PrositeMatch, that behaves in a way similar to a regular expression match.

16.5.4. Searching for patterns and parsing

As a conclusion, let us summarize how pattern searching and parsing interact when analyzing text. Basic text analysis is performed by searching for patterns, that are extracted (or scanned) from a text. Then patterns occurrences may be analyzed (or parsed) as a more general structure. In more complex parsing architectures, as the ones that have been introduced in Section 16.4, it is the parser engine which drives the process, by asking a scanner to feed him with basic token found in the text. In such systems, regular expressions are dened and used in the scanner.

134

Chapter 17. Object-oriented programming

Chapter 17. Object-oriented programming

17.1. Introduction
This chapter introduces objects, that are a way to organize code and data, and classes, that are a mechanism to describe new kinds of objects. We start by the description of what objects, classes and methods are, based on an example (Section 17.2). We then show its class denition (Section 17.3). We then explain how to combine objects in order to build an application (Section 17.4). Finally, technical aspects of classes and objects in Python are presented (Section 17.5). In the next chapter (Chapter 18), concepts related to object-oriented design will be developped.

17.2. What are objects and classes? An example

Our example is a class that handles DNA sequences.

17.2.1. Objects description

Let us rst start by describing our DNA object. Figure 17.1 shows an internal and schematic representation of such an object named s1. The object is actually represented in this diagram as a box enclosing attributes: the name of the sequence (for instance seq1) is stored in a name attribute, and the sequence characters (e.g aaacaacttcgtaagtata) is stored in a seq attribute. Attributes are the same as variables, except that they are associated to an object.

Figure 17.1. A DNA object

s1

name seq

17.2.2. Methods
As it is described so far, the DNA object is just a kind of record to group data. There are other ways of grouping data in Python: you can either use a list, or a dictionary:
>>> seq_record = {name: seq1, seq: aaacaacttcgtaagtata} >>> seq_record[name] seq1

135

Chapter 17. Object-oriented programming

but an object is actually more than just that. As shown in Figure 17.2 (and as stated in Section 6.1), an object is indeed more than just a value or a record of values: it can "do" things, and it "knows" what it can do, or. In other words, it has an associated behaviour. This behaviour is described by functions that are associated to it, and these functions are called methods. For instance, our DNA object can perform operations: it can translate its sequence or it can compute its GC percent by calling the gc function:
>>> s1.gc() 0.26

So, available operations are an important aspect of objects, that we need to add in our representation. In Figure 17.2, methods are represented as "counters", as at the post ofce.

Figure 17.2. Representation showing objects methods as counters

s1

name seq

gc translate revcompl

So, with objects, not only do we have a convenient way to group data that are related, but we also have a way to group data and their related operations in the same place.

Object
An object is a construction to group values and operations on these values.

17.2.3. Classes
Now, how do you dene objects, such as our DNA object, with their attributes and methods? In object-oriented programming, you dene objects by dening a class for them.

Class
A class is an object maker: it contains all the statements needed to create an object, its attributes, as well as the statements to describe the operations that the object will be able to perform. The term class can be somewhat misleading, for it also refers to something belonging to a classication. In objectoriented programming, a class can indeed also be a member of a set of classes organized in a hierarchy. This aspect

136

Chapter 17. Object-oriented programming

will be introduced later (see Section 18.4), through the concept of inheritance. But for now, we will use the term class as just a maker of objects, or an object factory. Having a class dened, you can create as many objects as needed. These objects are called instances of this class. All the instances created with a given class will have the same structure and behaviour. They will only differ regarding their state, i.e regarding the value of their attributes. Namely, names and sequences will not be the same for all sequences objects that will be created.

Instance
Instances are objects created with a class. The behaviour of the instances, i.e, the operations that are available for them either to modify them or to ask them for services, is described by functions associated to the class. These functions are called methods.

Method
A method is a function dened for a class, specifying the behaviour of the class instances.

Important
Classes and instances are not the same: the class is the maker, whereas the instance is the object being made, according to the model dened by the class.

17.2.4. Creating objects

How are we going to create DNA objects? You have actually already created objects throughout this course: strings, lists, etc... However, most of the time the objects you have manipulated were not directly created by your own code, but rather by other components: for instance the [] operator creates a list, the "" operator creates a string, the open function creates a le handle, etc... But the actual direct syntax to create objects or to instantiate a class, i.e to create an instance of class, is by calling a function with the same name as the class. So, in order to create an instance of the DNA class we have just dened, i.e an object which handles a DNA sequence, you do:
>>> s1 = DNA()

After the DNA() function is called, the instance of the class DNA is referred to by the s1 variable, just like with any Python value. You could also provide more information at instantiation. Lets say that we can create the name and seq attributes, we can do :
>>> s2 = DNA(seq2, acaagatgccattgtcccccggcctcctgctgctgctgctctccggggcca)

137

Chapter 17. Object-oriented programming

This time, the new DNA instance has been created with values for name and seq attributes. You can also use keyword arguments (especially if you are no longer sure in what order they should be entered) :
>>> s2 = DNA(name=seq2, seq=acaagatgccattgtcccccggcctcctgctgctgctgctctccggggcca)

So, this creates the object and makes it available from the s2 variable. You can now use this object to call DNA class methods, by using the dot operator:
>>> s2.gc() 0.66

The interpretor can nd which class s2 belongs to, and calls the gc method that is dened for the DNA class. You can also actually access to objects attributes by using the dot operator:
>>> s2.name seq2 >>> s2.seq acaagatgccattgtcccccggcctcctgctgctgctgctctccggggcca

17.3. Dening classes in Python

So, how do we actually dene classes? For instance, let us describe the DNA class. The program listed in Example 17.1 denes a DNA class as dened in Figure 17.2. The denition of the class is composed of two main parts: a header, providing the name of the class, and a body, that is composed of a list of denitions, mainly method denitions, but sometimes also assignments (see Section 17.5.1).

Example 17.1. DNA, a class for DNA sequences

class DNA: def __init__(self, name=None, seq=None): self.name = name self.seq = lower(seq) def gc(self): count_c = self.seq.count(c) count_g = self.seq.count(g) return float(count_c + count_g) / len(self.seq) def revcompl(self): revseq = for c in self.seq:

138

Chapter 17. Object-oriented programming

revseq = c + revseq revcompseq = for base in revseq: if base == a: revcompseq += elif base == t: revcompseq += elif base == c: revcompseq += elif base == g: revcompseq += return revcompseq def translate(self, frame=0): """ frame: 0, 1, 2, -1, -2, -3 """ if frame < 0 : seq = self.revcompl() frame = abs(frame) - 1 else: seq = self.seq if frame > 2: return protseq = for i in range(frame,len(seq) - 2,3): codon = seq[i:i+3] protseq += Standard_Genetic_Code[codon] return protseq

t a g c

This statement declares and creates DNA as a class. The __init__ method is automatically called at instance creation (see below). Initialization of instances attributes (name and seq). This method denes how to compute the GC percent of the sequence. This method denes how to compute the reverse complement of the sequence. This method denes how to translate the DNA sequence into a protein sequence. The Standard_Genetic_Code dictionary is dened elsewhere. The self parameter represents the object itself. You could of course use any other word like carrot or ego, but this would not help the reading of your code by others... So self is present in the Class and methods denitions each time the reference to the object instance is needed.

139

Chapter 17. Object-oriented programming

Let us rst look at one of these methods denitions, the gc method, which computes the GC percent of the sequence:

def gc(self): count_c = self.seq.count(c) count_g = self.seq.count(g) return float(count_c + count_g) / len(self.seq)

Method denitions follow exactly the same pattern as standard function denitions, except that they must have declared a rst parameter (here: self) for referring to the instance. Indeed, an instance identicator, actually a reference to the instance, is required in order for the statements within methods to access to the current instance attributes: here, the access to the seq attribute is needed in order perform the count. In fact, Python automatically passes the instance reference as the rst argument of a method. Hence, it is associated to the rst parameter which is the self parameter. You dont need to specify the argument for the reference to the instance. This is automatically done by Python. In fact, calling:
>>> s2.gc() 0.66

is equivalent to:
>>> DNA.gc(s2) 0.66

The interpretor can nd which class s2 belongs to, and handles the passing of the instance reference argument automatically. How does the method computes its result? For this, it needs to access to the character sequence of the DNA object. This is done by using the seq attribute, that was dened at instantiation (i.e by the __init__ method, see below). Within the method, the attribute is available through the object by the dot operator: self.seq. This shows that the object attributes are always available, at least as long as the object itself still exists. Attributes are thus accessible from all the methods of the class. They are a way to share data among methods. The method also use local variables: count_c and count_g to compute intermediate results. These variables have a local scope, restricted to the methods namespace, exactly like local variables that are dened in functions. Let us now look at the translate method.
def translate(self, frame=0): """ frame: 0, 1, 2, -1, -2, -3 """ if frame < 0 : seq = self.revcompl() frame = abs(frame) - 1 else:

140

Chapter 17. Object-oriented programming

seq = self.seq if frame > 2: return protseq = nb_codons = len(seq)/3 for i in range(frame,len(seq) - 2,3): codon = seq[i:i+3] protseq += Standard_Genetic_Code[codon] return protseq

To call this method, you must provide the frame:

>>> s2.translate(0) TRCHCPPASCCCCSPGP

Indeed, this method has declared a frame parameter. It thus takes two arguments: self and frame. As for the gc method, the rst parameter does not have to be specied at calling time, only the remaining ones. Let us now nally look at the __init__ method.
def __init__(self, name=None, seq=None): self.name = name self.seq = lower(seq)

This is a special method, which, when dened, is called at class instantiation, e.g when you run the following statement:
>>> s2 = DNA(name=seq2, seq=acaagatgccattgtcccccggcctcctgctgctgctgctctccggggcca)

the __init__ method dened for the DNA class is in fact called with 3 arguments. As for the other methods, the self argument is automatically provided by Python as a reference to the newly created instance. You dont have to provide an __init__ method, but it is usually the good place to put initialization statements. Initial values for attributes can be passed as arguments and associated to attributes. A good practice is to assign them default values, such as None. You can also notice that the seq attribute is initialized with the lower string function: the other methods will thus not have to check for this in order to perform their computation.

17.4. Combining objects

A program is generally not built on a single object, but rather on a combination of several objects that interact together. In the DNA example, we can have several other kinds of objects: protein sequences, motifs, ... The protein object could be either created from an initial sequence of protein amino-acids, or computed by the

141

Chapter 17. Object-oriented programming

DNA translate method. Protein objects could have specic methods for protein sequences, such as hydrophobicity or molecular weight. These objects will be instances of the class Protein. First of all, let us look at how we could represent a simplied Protein object. Figure Figure 17.3 shows an object having one method: mw, and 2 attributes: name and seq.

Figure 17.3. A Protein object.

p1

name seq

mw

The denition of the Protein class follows:

class Protein:

weight = {"A":71.08,"C":103.14 ,"D":115.09 ,"E":129.12 ,"F":147.18 ,"G":57.06 ,"H":137.15 ,"I":11

default_prosite_file = prosite.dat def __init__(self, name=None, seq=None): self.name = name self.seq = upper(seq) def mw(self): molW = 0 for aa in self.seq: molW += Protein.weight[aa] #add water at the end of protein molW += 18.02 #convert in Kda molW = molW / 1000 return molW

You can notice that the class starts by the denition of weight and default_prosite_file variables. As we will see later, this class variable is available to all instances of the class. Now, the DNA objects knows how to be translated, right? So it would be more clever for the DNA class translate method to return a Protein object... The new denition of the translate method is:
def translate(self, frame=0):

142

Chapter 17. Object-oriented programming

""" frame: 0, 1, 2, -1, -2, -3 """ if frame < 0 : seq = self.revcompl() frame = abs(frame) - 1 else: seq = self.seq if frame > 2: return protseq = nb_codons = len(seq)/3 for i in range(frame,len(seq) - 2,3): codon = seq[i:i+3] protseq += Standard_Genetic_Code[codon] new_protein = Protein(name=self.name + " translation", seq=protseq) return new_protein

Look at the returned value: it is now a Protein object. The argument for the seq parameter of the Protein classs __init__ method is the value of the newly computed protseq, and the argument for the name is constructed from the Protein objects name. In the Protein object, we might also be interested in keeping the reference to the initial DNA object. This can help to analyze the protein sequence later. Figure 17.4 shows the DNA and Protein objects, and the link between

143

Chapter 17. Object-oriented programming

them. Now, the Protein object has 3 attributes: name, seq and dna.

Figure 17.4. Protein and DNA objects.

s1

name seq

gc translate revcompl
p1

name seq dna mw

The __init__ method of the Protein class is now:

def __init__(self, name=None, seq=None, dna=None): self.name = name self.seq = upper(seq) self.dna = dna

The nal code for Protein object instantiation in the translate method is now:
new_protein = Protein(name=self.name + " translation", return new_protein seq=protseq, dna=self)

Look at the value provided for the dna parameter of the Protein __init__ method. It is a reference to the DNA object, i.e: self. Notice that none of the parameters is mandatory (except self of course). In particular, the dna parameter does not have to be provided when the Protein is directly created from a le, as opposed to translated from a DNA object.

17.5. Classes and objects in Python: technical aspects

The aim of this section is to clarify technical aspects of classes and objects in Python.

144

Chapter 17. Object-oriented programming

17.5.1. Namespaces
Classes and instances have their own namespaces, that is accessible with the dot (.) operator. As illustrated by Figure 17.5, these namespaces are implemented by dictionaries, one for each instance, and one for the class (see also [Martelli2002]).

Figure 17.5. Classes and instances namespaces.

DNA gc translate revcompl
class

class

name
namespace

"seq1"

namespace
"aaacaa..."

name seq

"seq2" "aatag.."

seq

s1

s2

Instances attributes. As we have learnt, a class may dene attributes for its instances. For example, attributes of s1, such as the name, are directly available through the dot operator:
>>> s1.name seq1

The dictionary for the instance attributes is also accessible by its __dict__ variable, or the vars() function:
>>> s1.__dict__ {seq: aaacaacttcgtaagtata, name: seq1} >>> vars(s1) {seq: aaacaacttcgtaagtata, name: seq1}

The dir() command lists more attributes:

>>> dir(s1) [__doc__, __init__, __module__, gc, translate, name, seq]

because it is not limited to the dictionary of the instance. It actually also displays its class attributes, and recursively the attributes of its class base classes (see Section 18.4). You can add attributes to an instance that were not dened by the class, such as the annotation in the following:
>>> s1.annotation = an annotation

145

Chapter 17. Object-oriented programming

>>> s1.__dict__ {seq: aaacaacttcgtaagtata, name: seq1, annotation: an annotation}

Adding attributes on-the-y is not something that is available in many object-oriented programming languages! Be aware that this type of programming should be used carefully, since by doing this, you start to have instances that have different behaviour, at least if you consider that the list of attributes denes a behaviour. This is not the same as having a different state by having different values for the same attribute. But this matter is probably a topic of discussion. Class attributes. It is also possible to dene attributes at the class level. These attributes will be shared by all the instances (Figure 17.6). You dene such attributes in the class body part, usually at the top, for legibility:

class Protein: ... weight = {"A":71.08,"C":103.14 ,"D":115.09 ,"E":129.12 ,"F":147.18 ,"G":57.06 ,"H":1 default_prosite_file = prosite.dat ...

To access this attribute, you use the dot notation:

>>> Protein.default_prosite_file prosite.dat >>> Protein.weight {"A":71.08,"C":103.14 ,"D":115.09 ,"E":129.12 ,"F":147.18 ,"G":57.06 ,"H":137.15 ,"I":113.17 ,"K":128

146

Chapter 17. Object-oriented programming

Figure 17.6. Class attributes in class dictionary

Protein mw weight default_prosite_file
class class
{"A":71.08,"C":103.":71.08, ..} "prosite.dat"

namespace

p1

name seq dna

"seq1 translation" "TRCHCPPA..."

namespace

p2

name seq dna

"p2" "MASVQL..."

DNA

gc translate revcompl
class

class

name
namespace

"seq1"

namespace
"aaacaa..."

name seq

"seq2" "aatag.."

seq

s1

s2

You can also access to this attribute through an instance:

>>> p1.default_prosite_file prosite.dat

You cannot change the attribute through the instance, though:

>>> p1.default_prosite_file = myfile.dat >>> Protein.default_prosite_file prosite.dat

This just creates a new default_prosite_file attribute for the p1 instance, which masks the class attribute, by the way.

147

Chapter 17. Object-oriented programming

The class attributes are displayed by the pydoc command, as opposed to the instance attributes (see Section 17.5.5). Class methods are referenced in the class dictionary: but what is their value actually? As shown in Figure 17.6, the class dictionary entries for methods are pointing to standard Python functions. When accessing to a method through an instance name, such as in p1.mw, there is an intermediate data structure, that itself points to the class and the instance. Objects in this data structure are called bound methods:
>>> s1.gc <bound method DNA.gc> of <__main__.DNA instance at 0x4016a56c>

They are said to be bound, because they are bound to a particular instance, and know about it. This is really important, for it is the only way to know what to do, and on which object to operate.

Figure 17.7. Classes methods and bound methods

DNA gc translate revcompl

class

name s1 seq

"seq1" "aaacaa..."

gc translate revcompl

bound methods : s1.gc s1.translate s1.revcompl

17.5.2. Objects lifespan

Once it is created, an objects lifespan depends on the fact that there are references on it. Namely, as opposed to variables present within functions, an object can still exist after exiting the function or method where it has been created, as long as there is a valid reference to it, as shown in the following example:
class C1: pass class C2: def show(self): print "I am an instance of class ", self.__class__ def create_C2_ref_in(p): p.c2 = C2()

148

Chapter 17. Object-oriented programming

c1 = C1() create_C2_ref_in(c1) c1.c2.show()

This function creates an instance of class C2 and stores its reference in an attribute of p, an instance of class C1. Creation of the C2 instance by calling create_C2_ref_in This statement displays: "I am an instance of class __main__.C2" As you can observe, the C2 instance exists after exiting the create_C2_ref_in function. It will exist as long as its reference remains in the c1.c2 attribute. If you issue:
c1.c2 = None

There will be no reference left to our C2 instance, and it will be automatically deleted. The same would happen if you would issue an additional call to the create_C2_ref_in function:
create_C2_ref_in(c1)

it would overwrite the preceeding reference to the former C2 instance, and delete it. You can check this by asking the c1.c2 reference for its identity:
id(c1.c2)

Of course, another way to delete an object is to use the del function:

del c1.c2

17.5.3. Objects equality

Instances equality cannot be tested by the == operator.If we come back to our DNA class:
>>> a = DNA(seq1, acaagatgccattgtc) >>> b = DNA(seq1, acaagatgccattgtc) >>> a.__dict__ {name: seq1, seq: acaagatgccattgtc} >>> b.__dict__ {name: seq1, seq: acaagatgccattgtc} >>> a == b False >>> a.__dict__ == b.__dict__

149

Chapter 17. Object-oriented programming

True

This means that the equality operator must be dened by the programmer. We will see the __eq__ special method later in Section 18.3.3. Instances identity means that two objects are in fact the same object, or more exactly, that two variables refer to the same object.
>>> a >>> b >>> b True >>> b True = DNA(seq1, acaagatgccattgtc) = a == a is a

As for all Python objects, identity implies equality.

17.5.4. Classes and types

Types in Python include integer,oating-point numbers, strings, lists,dictionaries, etc... Basically, types and classes are very similar. There is a general difference between them, however, lying in the fact that there are literals for built-in types, such as:
34 a nice string 0.006 [7, a, 45]

whereas there is no literal for a class. The reason for this difference between types and classes is that you can dene a predicate for recognizing expressions of a type [Wegner89], while, with class, you cannot, you can only dene collections of objects after a template. As shown in Figure 17.8, the Python type() can be used to know whether a variable is a class or an instance. It will very basically answer ClassType or InstanceType, as dened in module types, but it will not tell you which class an instance belongs to.

Figure 17.8. Types of classes and objects.

DNA __class__ s1 type()

ClassType

type()

InstanceType

150

Chapter 17. Object-oriented programming

17.5.5. Getting information on classes and instances

It is important to know what classes are available, what methods are dened for a class, and what arguments can be passed to them. First, classes are generally dened in modules, and the modules you want to use should have some documentation to explain how to use them. Then, you have the pydoc command that lists the methods of the class, and describes their parameters. The following command displays information on the DNA class, provided it is in the sequence.py le:
pydoc sequence.DNA

See also the embedding module, which might bring additional documentation about related components. This may be important when several classes work together, as is described in Section 17.4. When you consult the documentation of a class with the pydoc command, you get most of the time a strange list of method names, such as __str__ or __getitem__. These methods are special methods to redene operators, and will be explained in the next chapter on object-oriented design (Section 18.3.3). Caution: the dened instances attributes will not be listed by pydoc, since they belong to the instances rather than to the class. That is why they should be described in the documentation string of the class. If they are not, which sometimes happens..., run the Python interpretor and create an instance, then ask for its dictionary or use the dir() command:
>>> s1 = DNA() >>> dir(s1) [__doc__, __init__, __module__, gc, revcompl, translate, name, seq]

Information on instances. There are some mechanisms to know about the class of a given instance. You can use the special attribute __class__:
>>> s1 = DNA() >>> s1.__class__ class __main__.DNA at 0x81d1d64>

You can even use this information to create other instances:

>>> s2=s1.__class__() >>> s2 __main__.DNA instance at 0x8194ca4>

This can be useful if you need to create an object of the same class of another source object, without knowing the class of the source object. You can also ask whether an object belongs to a given class:
>>> isinstance(s1,DNA)

151

Chapter 17. Object-oriented programming

True

As mentionned above, the Python type() will not provide the class of an instance, but just: InstanceType (see Figure 17.8).

152

Chapter 18. Object-oriented design

Chapter 18. Object-oriented design

18.1. Introduction
We rst describe general software engineering guidelines to design components (Section 18.2.5). We next move on to the general question about why and how to design classes, and more specically, how to design new abstract data types (Section 18.3). We will then re-discuss the issue of exibility (Section 18.5) and describe inheritance (Section 18.4). Last, we will introduce the design patterns catalog (Section 18.6), which are well-known design solutions to standard problems in object-oriented programming.

18.2. Components
18.2.1. Software quality factors
The topics introduced in this section address some of the issues of software quality, and how Python can help on this matter. Before entering into details, let us just summarize some important concepts (you can nd a good and more exhaustive description in [Meyer97]). There is no absolute quality in software: depending on the context, scale, scope and goals of the program being developped, you might very well either write on-the-y short pieces of code to solve a temporary problem, or spend a signicant effort to have your application follow an industrial design process. So, rather than only a global so-called standard that should be applied for each program you write, there are a few quality factors to be evaluated according to the actual needs of the project. Among these factors, one usually distinguish between internal and external factors. External quality factors are the ones that corresponds to visible requirements, directly important for the user of the software, such as validity, robustness or efciency. Internal quality factors are properties of the code itself, such as legibility or modularity. In fact, internal factors often indirectly help to get external quality. Some factors, such as reusability, extensibility and compatibility, that we will study more thoroughly here, belong to external quality factors in the sense that they can be required by the users of a software library or the programmers of a shared source code; they are also important internal quality factors in the sense that they make the internal quality of the source code better. The aim of this chapter is mainly to describe these factors, as well as internal quality factors.

18.2.2. Large scale programming

Theoretically, in order to get a program that performs a given task and solves the problem you have specied, a basic set of instructions such as: branching, repetitions, expressions and data structures can be sufcient. Now, the programs that you produce can become a problem by themselves, for several reasons: They can become very large, resulting in thousand lines of code where it is becoming difcult to make even a slight change (extensibility problem).

153

Chapter 18. Object-oriented design

When an application is developped within a team, it is important for different people to be able to share the code and combine parts developped by different people (compatibility) ; having big and complex source les can become a problem. During your programmers life, or within a team, you will very often have to re-use the same kind of instructions set: searching for an item in a collection, organizing a hierarchical data structure, converting data formats, ...; moreover, such typical blocks of code have certainly already been done elsewhere (reusability). Generally, source code not well designed for re-use can thus be a problem. So, depending on the context of the project, there are some issues which are just related to the management of source code by humans, as opposed to the specication of the task to perform. And if you think about it, you probably tend to use variable names that are relevant to your data and problem, arent you? So, why? This is probably not for the computer, but, of course, rather for the human reader. So, in order to handle source structure and management issues, several conceptual and technical solutions have been designed in modern programming languages. This is the topic of this chapter. Let us say that we have to develop source code for a big application and that we want this source code to be spread and shared among several team members, to be easy to maintain and evolve (extensible), and to be useful outside of the project for other people (reusable). What are the properties a source code should have for these purpose? it should be divided in small manageable chunks these chunks should be logically divided they should be easy to understand and use they should be as independant as possible: you should not have to use chunk A each time you need to use chunk B they should have a clear interface dening what they can do The most important concept to obtain these properties is called modularity, or how to build modular software components. The general idea related to software components is that an application program can be built by combining small logical building blocks. In this approach, as shown in gure Figure 18.1, building blocks form a kind of high-level language.

154

Chapter 18. Object-oriented design

Figure 18.1. Components as a language

Application

Components Python

18.2.3. Modularity
The simplest form of modularity is actually something that you already know: writing a function to encapsulate a block of statements within a logical unit, with some form of generalization, or abstraction, through the denition of some parameters. But there are more general and elaborated forms of components, namely: modules and packages. So, what is modularity? As developped in [Meyer97], modularity is again not a general single property, but is rather described by a few principles: A few interfaces: a component must communicate with as few other components as possible. The graph of dependencies between components should be rather loosely coupled. Small interfaces: whenever two components communicate, there should be as few communication as possible between them. Explicit interfaces: interfaces should be explicit. Indirect coupling, in particular through shared variables, should be made explicitly public. Information hiding: information in a component should generally remain private, except for elements explicitly belonging to the interface. This means that it should not be necessary to use non public attributes elements of a component in order to use it. In languages such as Python, as we will see later, it is technically difcult to hide a components attributes. So, some care must be taken in order to document public and private attributes.

155

Chapter 18. Object-oriented design

Syntactic units: Components must correspond to syntactic units of the language. In Python, this means that components should correspond to known elements such as modules, packages, classes, or functions that you use in Python statements:
import dna from Bio.Seq import Seq

dna, Bio, Bio.Seq and Seq are syntactic units, not only les, directories or block of statements. In fact, Python really helps in dening components: almost everything that you dene in a module is a syntactic unit. You can view this approach as though not only the user of the application would be taken into account, but also the programmer, as the user of an intermediate level product. This is why there is a need for interfaces design also at the component level.

18.2.4. Methodology
These properties may be easier to obtain by choosing an appropriate design methodogy. A design methodoly should indeed: help in dening components by successive decomposition; help in dening components that are easy to combine; help in designing self-understandable components: a programmer should be able to understand how to use a component by looking only at this component; help in dening extensible components; the more independant components are, the more they are easy to evolve; for instance, components sharing an internal data structure representation are difcult to change, because you have to modify all of them whenever you change the data structure.

18.2.5. Reusability
Programming is by denition a very repetitive task, and programmers have dreamed a lot of being able to pick offthe-shelves general purpose components, relieving them from this burden of programming the same code again and again. However, this objective has, by far, not really been reached so far. There are several both non technical and technical reasons for this. Non-technical reasons encompass organisational and psychological obstacles: although this has probably been reduced by the wide access to the Web, being aware of existing software, taking the time to learn it, and accepting to use something you dont have built yourself are common difculties in reusing components. On the technical side, there are some conditions for modules to be reusable. 1. Flexibility: One of the main difculty for making reusable components lies in the fact that, while having the impression that you are again programming the same stereotyped code, one does not really repeat exactly the same code. There are indeed slight variations that typically concern to following aspects (for instance, in a standard table lookup):

156

Chapter 18. Object-oriented design

types: the exact data type being used may vary: the table might contain integers, strings, ... data structures and algorithms may vary: the table might be implemented with an array, a dictionary, a binary search tree, ... ; the comparison function in the sort procedure may also vary according to the type of the items. So, as you can understand from these remarks, the more exible the component is, the more reusable it is. Flexibility can be partly obtained by modularity, as long as modules are well designed. However, in order to get real exibility, other techniques are required, such as genericity, polymorphism, or inheritance, that are described in Section 18.5. 2. Independancy towards the internal representation: by providing a interface that does not imply any specic internal data structure, the module can be used more safely. The client program will be able to use the same interface, even if the internal representation is modied. 3. Group of related objects: it is easier to use components when all objects that should be used together (the data structures, data and algorithms) are actually grouped in the same component. 4. Common features: common features or similar templates among different modules should be made shareable, thus making the whole set of modules more consistent.

18.3. Abstract Data Types

18.3.1. Denition
What is an abstract data type? Well, a data type is what we have just presented in chapter Chapter 17: in Python, as in many object-oriented language, it is a class. So, why abstract? One of the main objectives in component building, as we have seen in section Section 18.2, is to provide components that a programmer (you, your colleagues, or any programmer downloading your code) can be condent in. In order to get this programmer, who is a client of your code, condent into your class, you have to make it as stable as possible, and the best method to get a stable class is to dene it at a level where no implementation decision is visible. In other words, dening a class should consist in dening a data type providing some abstract services, whose interface are clearly dened in term of parameters and return value and potential side effects. The advantages of this approach are the following: The implementor of the class can change the implementation for maintenance, bug xes or optimization reasons, without disturbing the client code. The data type is dened as a set of high-level services with a semantic contract dening both the output that is provided and the input that is required from the client. This actually correspond to the general denition of a type in programming: a type is dened as a set of possible values and a set of operations available on these values. Among the methods dened in the interface of a data type, there are methods that build or change the state of the corresponding object, which are called constructors and modicators, and there are methods to access the object, which are called accessors.

157

Chapter 18. Object-oriented design

Example 18.1. A Stack

Our example is a stack. It is directly inspired from [Meyer97], chapter 6. Stacks serve to pile up objects of any kind, and take them out according to the rule "last in, rst out" (LIFO). Imagine a pile of plates at home, in your cupboard, the safest way to take a plate out is to take the last one put in (Figure 17.8). Such structures are omni-present in programing, they serve to move along graphs, to compile programs, etc...

Figure 18.2. A stack

put remove

Stacks can be implemented in different ways: you can use a list, and either add the new item at the beginning or at the end of the list. Conceiving an abstract data type (ADT) for the stack consists in describing the functions needed making abstraction of the implementation choice that will be made at the end (and that may change). Indeed in all cases, the basic services are the same, and can be given common names, that do not "betray" how the list is used "inside" of the code. The set of services is listed in Table 18.1.

Table 18.1. Stack class interface

Name put remove item Input a stack and an item a stack a stack Output a stack a stack an item Description places the item on top of the stack if stack is not empty, removes the last item if stack is not empty, returns the item on top of the pile, without removing it tells whether the stack is empty creates a new stack

empty make

a stack

a Boolean a stack

The description provided by Table 18.1 should sufce for the client to use the class. What is more important here, is that the client does not have to know anything about the internal representation of the stack. In this ADT accessors are: item and empty, constructor is make , and modicators are: put and remove. The next step,once this description of the ADT is made for the client, consists in specifying the ADT, that is, describing in formal or mathematical language what the functions are doing. This involves four steps:

158

Chapter 18. Object-oriented design

Types: this corresponds in Python to the classes that are used for the ADT. In our example there is one class, Stack. Functions: That is their names and the input and output they have, as shown in Table 18.1, rst three columns. Axioms: The rules to which the functions obey, and that are sufcient to describe them. For the stack, these axioms are, for any stack s and element e: item(put(s, e))=e remove(put(s, e))=s empty(make) is True empty(put(s, e) is False Pre-conditions: The functions we need will not operate under all conditions. For example, you cannot remove an element from an empty stack. These pre-conditions can be enumerated as follows for the stack: remove(s) requires : not empty (s) item(s) requires not empty(s) Axioms and pre-conditions formulate more precisely what is said inthe fourth column of Table 18.1

18.3.2. Information hiding

It is a good design practice to protect an instance local data. Why is that? The rst reason is that it makes the interface of the class more easy to grasp. The client just knows what needs to be known in order for the component to be used properly. A second reason is that class data should rather be handled by the class methods: there might be some coherence to be maintained among different attributes, that a method can check. Another important reason is to avoid the risk of collisions, that could happen between the class and the potential base classes, as we will explain in Section 18.4, devoted to inheritance. The problem is that there is no real mechanism in Python to prevent a client code from accessing to an instance attributes, as there are in other languages (such as Java or C++), where you can declare private and public attributes. Python provides just the following: if you name a variable with a __ prex and no _ sufx, such as:
__list = []

within a class named, e.g: Stack, Python automatically changes it to: _Stack__list. You can still access it, but you are aware that you are probably doing something odd, and it can help to avoid name collisions. Another way to distinguish public and private attributes and methods is to prex the private ones with a single _ character. This does not provoke any automatic addition by Python, but it is warning the reader that this part of the

159

Chapter 18. Object-oriented design

code is private. Another style matter is the naming of classes, where it is a good idea to capitalize the rst letter of class names, to distinguish them from variable or function names. Be aware nally that some methods in Python are framed by double __ on both side, they are called special methods and described in the next p aragraph. The lesson to be learnt is that attributes you want to be accessible should be mentionned explicitely, and documented as such. Now that the ADT is specied, that we know which attributes and methods are public or private, We can turn to the implementation part of our example. As shown in Example 18.2, we have decided to use a Python list, and to add new elements to the end of this list.

Example 18.2. Stack class

class Stack: """ A class to handle stacks """ def __init__(self): self._list=[] def put(self, elem): """ places an item on top of the stack """ self._list.append(elem) def remove(self): """ if not empty, removes the last item placed on the stack else does not do anything """ if self._list: self._list.pop() def item(self): """ if not empty, returns the item on top of the stack without removing it else does not do anything """ if self._list: return self._list[-1] def empty(self): """ tells if the stack is empty """

160

Chapter 18. Object-oriented design

if len(self._list)==0: return True return False

18.3.3. Using special methods within classes

The usual Python operators, such as +, the [ ] operator to access an item or a slice, the assignment = operator, the equality == operator, but also print etc..., are dened with special methods that have different denitions for different types. Indeed adding two integers is not the same as adding two strings. It is very convenient to redene these special methods inside the classes created with Python. The very basic special methods encountered in most classes is the one allowing to use print. Such special methods have obligate names, and are framed by __, such as __str__ for print. A set a very precious pages in the book Python. Essential reference concerns all the tables (3.7 to 3.10) listing these obligate names. Here below in Table 18.2, we will list a small number of examples.

Table 18.2. Some of the special methods to redene Python operators

Operator [] == . . and = len() print del Special method __getitem__ __eq__ __getattr__ __setattr__ __len__ __str__ __del__ Parameters self, k self, other self, name self, name, value self self self Description access to item or slice specied by k test for equality between 2 objects access to a non existing attribute named name modication of an attribute length computation string form for instance value deletion of an instance

Example 18.3. Dening operators for the DNA class

Remember what we said in Section 17.5.3: different objects, even with the same state, i.e with the same attributes values, are considered as not equal by Python. There is a way however to specify how the == operator should behave for your class. The following code redenes the == operator as returning True whenever the name and seq attributes are equal.
class DNA: ... def __eq__(self, other):

161

Chapter 18. Object-oriented design

if self.name == other.name and self.seq == other.seq: return True return False >>> s1 = DNA(name=seq2, seq=acaagatgccattgtcccccggcctcctgctgctgctgctctccggggcca) >>> s2 = DNA(name=s1.name, seq=s1.seq) >>> s1 == s2 True >>> s1 is s2 False

Another important operator that could be useful for sequences is the [] operator. for strings, this operator enables to access a character at a specic position. This operator is dened by the __getitem__ special method.
class DNA: ... def __getitem__(self, i): return self.seq[i] >>> print s1[3] a

Exercise 18.1. Operators for the DNA class

Which additional operators could you dene for DNA class? Implement 1 or 2 of them.

18.4. Inheritance: sharing code among classes

18.4.1. Introduction
What we have seen so far, object-based programming, consists in designing programs with objects, that are built with classes. In most object-oriented programming languages, you also have a mechanism that enables classes to share code. The idea is very simple: whenever there are some commonalities between classes, why having them repeat the same code, thus leading to maintenance complexity and potential inconsistency? So the principle of this mechanism is to dene some classes as being the same as other ones, with some differences.

Example 18.4. Inheritance example (1): sequences

So far, we have design two different classes for different types of sequences: indeed, having all the methods dened only for protein sequences also available for DNA sequences can be considered a problem. At the same

162

Chapter 18. Object-oriented design

time, some functionalities are common to both types of sequences: cleaning the sequence, producing a Fasta format to name a few. So we keep our two classes to deal with DNA and protein sequences, and we add a new class: Sequence, which will be a common class to deal with general sequence functions. In order to specify that a DNA (or a Protein) is a Sequence in Python is:
class DNA(Sequence):

The DNA class is called a subclass, and the Sequence is called a superclass or a base class. Following the class statement, there are only the denitions specic to the DNA or Protein classes: you do not need to re-specify the other ones. The following code shows how to specify the Sequence superclass.
class Sequence: def __init__(self, name=None, seq=None): self.name = name self.seq = seq self.clean() def clean(self): seq = "" for c in self.seq: if c in self.alphabet: seq += c self.seq = seq def __str__(self): return ">"+self.name+"\n"+self.seq def __getitem__(self, i): return self.seq[i] def getname(self): return self.seq def getseq(self): return self.seq

class DNA(Sequence): alphabet = "atcg" def gc(self): """GC percent""" ... def revcompl(self): """reverse complement"""

163

Chapter 18. Object-oriented design

... def translate(self): """translation into a protein""" ...

class Protein(Sequence):

weight = {"A":71.08,"C":103.14 ,"D":115.09 ,"E":129.12 ,"F":147.18 ,"G":57.06 ,"H":137.15 ,"I":11

alphabet = weight.keys() def mw(self): molW = 0 for aa in self.seq: molW += self.weight[aa] molW += 18.02 molW = molW / 1000 return molW

Look at how we use these denitions:

>>> s = DNA("s1", "attgccctt") >>> s.gc() 66.66 >>> print s >s1 attgccctt

The __str__ method that is called when issuing a print statement is not dened for the DNA class. So python looks up one level further, nds a denition at the Sequence class denition level and calls it. Provided with the self reference to the object, exactly the same way as for DNA class methods, it can access to the attributes in order to produce a printable string of the object. The clean method is a common method, except that it uses a specic alphabet for each type of sequence. This is why each subclass must dene an alphabet class attribute. When referring to self.alphabet, python looks up in the objects namespace rst, then one level up in the class namespace and nds it. Now look carefuly at the __init__ method: it is now dened in the superclass. There are indeed common things to do for both DNA and Protein classes: attributes initialization, sequence cleaning, ... So it is enough to dene the __init__ method at the superclass level: whenever the __init__ method, or any method is not dened at the subclass level, it is the method at the superclass level which is called (see below Section 18.4.1.2). But, remember, the DNA sequence is supposed to be in lowercase, whereas the protein sequence is in uppercase.

164

Chapter 18. Object-oriented design

How can we do that? The idea is to dene an __init__ method at the subclass level that calls the __init__ method of the superclass, and then proceed to actions specic to the DNA or the Protein sequences.
class Sequence: def __init__(self, name=None, seq=None): self.name = name self.seq = seq self.clean() ... class DNA(Sequence): def __init__(self, name=None, seq=None): Sequence.__init__(self, name, lower(seq)) ... class Protein(Sequence): def __init__(self, name=None, seq=None, dna=None): Sequence.__init__(self, name, upper(seq)) self.dna = dna # when called by DNA.translate() ...

18.4.1.1. Overloading
Let us decide that a generic mw method can be generalized and put at the Sequence class level. This way, a potential RNA subclas of Sequence will have a pre-dened mw method, common to all the molecular sequences. We do not need any mw method at the Protein level anymore. But, since the computation of the molecular weight differs slightly for DNA sequences, we dene a specic mw method in the DNA class.
class Sequence: def mw(self): molW = 0 for c in self.seq: molW += self.weight[] return molW ... class DNA(Sequence):

165

Chapter 18. Object-oriented design

def mw(self): """ Computes the molecular weight of the 2 strands of a DNA sequence. """ molW = 0 for c in self.seq: molW += self.weight[] for c in self.revcompl(): molW += self.weight[] return molW ...

When a method is redened (or overriden) in the subclass, which is the case of the mw method, it is said that the method is overloaded. This means that, according to the actual class of a sequence instance, which can be either DNA or Protein, the method actually called can vary.

Overloading
Overloading a method is to redene at a subclass level a method that exists in upper classes of a class hierarchy. Another term that is used in object-oriented programming is the term: "polymorphism". "Polymorphism" litteraly means: several forms. In other words, the name mw has several meanings, depending on the context of the call.

Polymorphism
A method is said to be polymorphic if it has several denitions at different levels of a class hierarchy. The term "polymorphism" is also used about operators. The + operator, for instance, is polymorphic in the sense that it refers to different operations when used with strings and integers.

18.4.1.2. How does it work? Dynamic binding.

As we said above, by declaring DNA as a subclass of Sequence, you do not need to re-specify in DNA all the methods that are already dened in Sequence. The way it works is that Python looks up methods starting from the actual class of the object, for instance DNa and, if not found, by looking up further in Sequence super class for the needed method or attribute. So, when calling:
>>> my_dna.gc()

166

Chapter 18. Object-oriented design

python looks up for the gc method (Figure 18.3). Python nds it in the current class. When looking up for other methods, such as the __str__ method, which is not dened in the DNA class, Python follows the graph of base classes (Figure 18.4). Here, we only have one: Sequence, where __str__ is dened.

Figure 18.3. Dynamic binding (1)

__dict__ Sequence __bases__ DNA __dict__ gc translate clean __str__

__class__

s1

s1.gc()

167

Chapter 18. Object-oriented design

Figure 18.4. Dynamic binding (2)

__dict__ Sequence __bases__ DNA __dict__ gc translate clean __str__

__class__

s1

print s1

Forms of inheritance. There are two forms of inheritance: extension and specialisation. In other words, inheritance can be used to extend a base class adding functionalities, or to specialise a base class. In the case of the DNA class, it is rather an extension. Sometimes, the term subclass is criticized because, in the case of extension, you actually provide more services, rather than a subset. The term subclass ts better to the case of specialization, where the subclass addresses a specic subset of the base class potential objects (for instance, dogs are a subset of mammals).

168

Chapter 18. Object-oriented design

UML diagrams. Classes relationships (as well as instances relationships, not represented here) can be represented by so-called UML diagrams, as illustrated in Figure 18.5.

Figure 18.5. UML diagram for inheritance

Sequence

DNA

Protein

18.4.1.3. Multiple inheritance

In Python, as in several programming language, you can have a class inherit from several base classes. Normally, this happens when you need to mix very different functionalities. For instance, you want to wrap your class with a class that provides tools for testing (Figure 18.6), or services for persistence. Inheriting from classes that come from the same hierarchy can be tricky, in particular if methods are dened everywhere: you will have to know how the interpretor choose a path in the classes graph. But this case is more like a theoretical limit case, that should not happen in a well designed program.

Figure 18.6. Multiple Inheritance

Testing A

18.4.2. Discussion
169

Chapter 18. Object-oriented design

Benets of inheritance. Regarding exibility, the inheritance mechanism provides a convenient way to dene variants at the method level: with inheritance indeed, the methods become parameters, since you can redene any of them. So you get more than just the possibility of changing a parameter value. Regarding reusability, inheritance is very efcient, since the objective is to reuse a code which is already dened. Components designed with the idea of reuse in mind very often have a clean inheritance hierarchy in order to provide a way for programmers to adapt the component to their own need. Inheritance also provides an elegant extensibility mechanism, by denition. It lets you extend a class without changing the code of this class, or the code of the module containing the class. Combining class or combining objects? Using inheritance is not mandatory. The main risk of using it too much is to get a complex set of classes having a lot of dependencies and that are difcult to understand. There are generally two possibilities to get things done in object-oriented programming: Inheritance: you combine classes in order to get a "rich" class providing a set of services coming from each combined class. Composition: you combine objects from different classes. The use of composition instead of inheritance is also illustrated by the design patterns from [Gamma95], that are introduced in Section 18.6. Problem with inheritance for extension. When using inheritance to extend a base class, you might want to have a method in the subclass not just overloading the method in the base class, but as a complement. In this case, one usually rst calls the base class, and then performs the complementary stuff. In Python, you have to know explicitely the name of the superclass to perform this call (see for instance method __init__). Be aware, that this can become rather tricky sometimes, for you have to design a real protocol describing the sequence of calls that have to be done among classes, when not only one method is involved. When using inheritance or composition: summary. The question of choosing between inheritance and composition to combine objects A and B results in deciding whether A is-a B or whether A has-a B. Unfortunately, it is not always possible to decide about this, only on the basis of the nature of A and B. There a few guidelines, though (see [Harrison97], chapter 2). The main advantage of inheritance over composition is that method binding, e.g lookup and actual method call, is done automatically: you do not have to perform the method lookup and call, whereas, when combining objects, you have to know which one has the appropriate method. For instance, a Protein instance may have a dna if created by translate. Use composition when you catch yourself making exceptions to the is-a rule and redening several inherited methods or willing to protect access to the base class attributes and/or methods. In such a case, the advantage described in the previous item of having automatic attribute and method access becomes a problem.

170

Chapter 18. Object-oriented design

Use composition when the relationships between objects are dynamic. For instance, a different way to design sequence tools, such as clean, __str__, etc... could be to design a SeqTools class:
class SeqTools: def __init__(self, seqobject): self.seqobject = seqobject del clean(self): seq = "" for c in self.seqobject.seq: if c in self.seqobject.alphabet: seq = seq + c self.seqobject.seq = seq ...

The DNA or Protein instances could then have a reference to a SeqTools instance, created at instantiation:
class DNA: def __init__(self, name=None, seq=None): self.toolbox = build_seqtools(self) self.name = name self.seq = seq self.toolbox.clean() def build_toolbox(seqobject): return SeqTools(seqobject)

The main advantage is exibility: you could more easily redene which sequence toolbox to use, provided the public interface is the same, than with inheritance. Indeed, inheritance relationships are xed at denition time:
class DNA(Sequence): ...

Use composition when a single object must have more than one part of a given type. Use composition to avoid deep inheritance hierarchies. If you cant decide between inheritance and composition, take composition. Do not use inheritance when you get too many combined method calls between base class and sub-classes, which can happen when using inheritance for extension.

171

Chapter 18. Object-oriented design

Use inheritance when you want to build an abstract framework, which purpose is to be specialized for various uses (see Exercise 18.2). A good example is the parsing framework in Biopython, that lets you create as many new parsers as needed. Inheritance also provides a good mechanism to design several layers of abstraction, that dene interfaces that programmers must follow when developping their components. Bioperl modules, although not developped in a true object-oriented language, are a good example of this.

Exercise 18.2. Example of an abstract framework: Enzyme parser

Design a parser for the Enzyme database, using Biopython parsing framework. See ??? for more details.

18.5. Flexibility
18.5.1. Summary of mechanisms for exibility in Python
Earlier, in Section 18.2.5, we have seen that one of the powerful properties a software piece can have is to be reusable. We also concluded that the more exible it is, the more general and thus reusable it will be. Let us now summarize some mechanisms in object-oriented programming that help in achieving more exibility. Genericity: genericity (as available in the ADA programming language), is a technique to dene parameters for a module, exactly as you dene parameters for a function, thus making the module more general. For instance, you can dene the type of an element of a Table module, or the function to move to the next item, as being generic. There is no specic syntax in Python to dene generic components, but in a way, it is not really necessary because Python is dynamically typed, and can handle functions as parameters for other functions. Inheritance: as we have just seen in Section 18.4 this describes the ability, in object-oriented programming, to derive a class from another, either to extend it or to specialize it. Overloading, which refers to the possibility for an operator or a method to behave differently according to the actual data types of their operands or arguments. This feature does not directly address the problem of exibility: it is rather an elegant syntactic mean not to have different operators or names to perform similar tasks on different objects or sets of objects. In this sense, it is actually a form of polymorphism. In objectoriented programming, overloading moreover exactly means being able to redene a method in a derived class, thus enabling polymorphism of instances: given an instance, it is possible that you do not know exactly to which class in a hierarchy it belongs (e.g DNA or Protein). In other programming languages, such as Java, there is another kind of overloading: it is possible, within the same class, to have several denitions of the same method, but with different signatures, i.e a different set of parameters. This does not exist in Python, and as described below (see Section 18.5.2), you have to handle it manually. Polymorphism: refers to the possibility for something to be of several forms, e.g, for an object to be of any type. In Python, for instance, lists are polymorphic, since you can put items of any type in a list. Overloading is one aspect of polymorphism, polymorphism of methods. Polymorphism brings a lot of exibility, just only because you do not have to dene as many classes as potential data type for it, as you would have to do in statically typed languages. In this sense, Python helps in designing more simple classes frameworks.

172

Chapter 18. Object-oriented design

Some of these techniques, mainly inheritance, are available in Python, other, such as genericity and overloading, are not, or in a restricted way.

18.5.2. Manual overloading

If you want a method to behave differently according to the parameters, you must either create subclasses or write the code to analyze the parameters yourself. In Example 18.5, there is an example a method dealing with different types of parameters, that are manually handled.

Example 18.5. Curve class: manual overloading

def __getitem__(self, key): """ Extended access to the curve. c[3] c[toto] c[x<100] c[x==y] """ _x = [] _y = [] _result = None if type(key) is ListType: _x = key _x.sort() _y = self.y(_x) _result = Curve(zip(_x, _y)) elif type(key) is SliceType: _x = self._range_x(key.start, key.stop) _y = apply(self.y, _x) if _x and _y: _result = Curve(zip(_x, _y)) elif self._curve.has_key(key): _result = self._curve[key] elif type(key) is StringType: _x = self._tag_x(key) if len(_x) > 0: _x.sort() _y = self.y(_x) _result = Curve(zip(_x, _y)) else: _result = self._getexpr(key) if _result is not None: return _result

Recognizes the argument as a list.

173

Chapter 18. Object-oriented design

Recognizes the argument as a slice. Recognizes the argument as an index. Recognizes the argument as a string, that is going to be further analyzed as a tag name or an expression. Recognizes the argument as a tag name. Recognizes the argument as an expression, such as x< y or y>10. Python is not able to do this analysis automatically, because it is dynamically typed. You do not have any mean to specify the type of the parameter. In a language providing static typing and full method overloading, i.e also within the same class, you could have several __getitem__ denitions that would look like:
def __getitem__(self, key: List): ... def __getitem__(self, key: Slice): ... def __getitem__(self, key: String): ...

As you can notice, this is of course not valid Python code, since there is no possibility to dene the type of a parameter in Python.

18.6. Object-oriented design patterns

Presenting Python classes and various related mechanisms without any description of their common uses would be like dening the for and while statements without any description of the typical uses of these constructions, and how they adapt to various programming needs. It is the purpose of this section to present common uses of classes and objects, the so-called design patterns, that have been used for a long time by programmers since object-oriented programming was born. Computer science is full of design patterns, and there is no exception for object-oriented programming. A catalog of object-oriented has been published by [Gamma95], and [Christopher2002] has a chapter dedicated to a few of them, with examples implemented in Python. Design patterns are not programs: they are widely used design choices to build programs. As such, they form a kind of conceptual catalog that you are encouraged to reuse to build your application. Not only are they useful in order not to reinvent the wheel and to save your time, but also because they provide a comprehension framework for programmers who intend to reuse your code and who need to understand it. This part does not aim at an exhaustive presentation of object-oriented design patterns, which are very well described elsewhere. Its main purpose is to give a taste of it, with examples in bioinformatics, and to introduce the main related ideas. There are three categories of object-oriented patterns: Creational patterns: patterns that can be used to create objects. Structural patterns: patterns that can be used to combine objects and classes in order to build structured objects. Behavioral patterns: patterns that can be used to build a computation and to control data ows.

174

Chapter 18. Object-oriented design

Creational patterns. There are two main categories of creational patterns: those for creating objects without having to know the class name, that you could call "abstract object makers" (abstract factory and factory method), and those to ensure a certain property regarding object creation, such as prohibiting more than one instance for a class (singleton), building a set of instances from different classes in a consistent way (builder), or creating an instance with a specic state (prototype). Abstract factory: an abstract factory is an object maker, where, instead of specifying a class name, you specify the kind of object you want. For instance, say that you want to create an agent to run analysis programs, you can ask a factory to do it for you:
clustalw = AnalysisFactory.program(clustalw) result = clustalw.run(seqfile = myseqs) print result.alig

The clustalw object is an instance of, say, the AnalysisAgent.Clustalw class, but you do not have to know about it at creation time. The only thing you know is the name of the program you want (clustalw), and the factory will do the rest for you. Factory method: a factory method is very similar to an abstract factory: just, instead of being a class, it is a method. For instance, you can create a sequence object (Bio.Seq.Seq in Biopython) by asking the get_seq_by_num method of an alignment object (Bio.Align.Generic.Alignment):
first_seq = align.get_seq_by_num(0)

The method which creates this instance of Bio.Seq.Seq is a factory method. The difference with a factory class is also that the factory method is often more than an object maker: it sometimes incorporates much more knowledge about the way to create the object than a factory would. A more simple factory method would be a new method dened in a class to create new instances of the same class:
other_seq = seq.new()

In this case, notice that in order to create my_scoring, you really do not have to know the actual class of scoring: the only thing you know is that you will get the same one, even if there is a whole hierarchy of different classes of scoring.

175

Chapter 18. Object-oriented design

Singleton: ensures that you cannot create more than one instance. For example, if you can dene a class to contain operations and data for genetic code: you need only one instance of this class to perform the task. Actually, this pattern would not be implemented with a class in Python, but rather with a module, at least if you can dene it statically (a dynamic singleton could not be a module, for a module has to be a le):
>>> import genetic_code >>> genetic_code.aa(TTT) F

Prototype: this pattern also lets you create a new object without knowing its class, but here, the new target object is created with the same state as the source object:
another_seq = seq.copy()

The interest is that you do not get an "empty" object here, but an object identical to seq. You can thus play with another_seq, change its attributes, etc... without breaking the original object. Builder: you sometimes need to create a complex object composed of several parts. This is the role of the builder. For instance, a builder is needed to build the whole set of nodes and leafs of a tree. The Blast parser in Biopython simultaneously instantiates several classes that are all component parts of of hit: Description, Alignment and Header. Structural patterns. Structural patterns address issues regarding how to combine and structure objects. For this reason, several structural patterns provide alternative solutions to design problems that would else involve inheritance relationships between classes. Decorator, proxy, adapter: these patterns all enable to combine two (or more) components, as shown in Figure 18.7. There is one component, A, "in front" of another one, B. A is the visible object a client will see. The role of A is either to extend or restrict B, or help in using B. So, this pattern is similar to subclassing, except that, where a sub-class inherits a method from a base class, the decorator delegates to its decoratee when it does not have the required method. The advantage is exibility (see Section 18.4.2): you can combine several of these components in any order at run time without having to create a big and complex hierarchy of subclasses.

176

Chapter 18. Object-oriented design

Figure 18.7. Delegation

A B

Generally, the Python code of the A class looks like:

class A: def __init__(self, b): """storing of the decoratee b (b is an instance of class B)""" self.b = b def __getattr__(self, name): """ methods/attributes A does not know about are delegated to b """ return getattr(self.b, name)

At use time, an instance of class A is created by providing a b instance.

b = B() a = A(b) print a.f()

Everything that class A cannot perform is forwarded to b (providing that class B knows about it). The decorator enables to add functionalities to another object. Example 18.6 shows a very simple decorator that prints a sequence in uppercase.

Example 18.6. An uppercase sequence class

import string class UpSeq: def __init__(self, seq): self.seq = seq def __str__(self): return string.upper(self.seq) def __getattr__(self,name):

177

Chapter 18. Object-oriented design

return getattr(self.seq,name)

The way to use it is for instance:

>>> s=UpSeq(DNA(name=name1, seq=atcgctgtc)) >>> print s ATCGCTGTC >>> s[0:3] atc >>> len(s) 9

Exercise 18.3. An analyzed sequence class

How would you design sequence classes where all sequence analyses of type f would be available as:
seq.f()

without actually dening all the possible f methods within the sequence class? The proxy rather handles the access to an object. There are several kinds of proxy: protection proxy: to protect the access to an object.

Exercise 18.4. A partially editable sequence

How would you design a sequence class where the only allowed editions would be:
del seq[i]

on a gap character, and:

seq[i] = A

on a N character?

178

Chapter 18. Object-oriented design

virtual proxy: to physically fetch data only when needed. Database dictionaries in Biopython work this way:
prosite = Bio.Prosite.ExPASyDictionary() entry = prosite[PS00079]

Data are fetched only when an access to an entry is actually requested. remote proxy: to simulate a local access for a remotely activated procedure. The adapter (or wrapper) helps in connecting two components that have been developped independantly and that have a different interface. For instance, the Pise package transforms Unix programs interfaces in standardized interfaces, either Web interfaces, or API. For instance, the golden Unix command has the following interface:
bash> golden embl:MMVASP

But the Pise wrapper enables to run it and get the result by a Python program, having an interface dened in the Python language:
factory = PiseFactory() golden = factory.program("golden",db="embl",query="MMVASP") job = golden.run() print job->content()

Composite: this pattern is often used to handle complex composite recursive structures. Example 18.7 shows a set of classes for a tree structure, illustrated in Figure 18.8. The main idea of the composite design pattern is to provide an uniform interface to instances from different classes in the same hierarchy, where instances are all components of the same composite complex object. In Example 18.7, you have two types of nodes: Node and Leaf, but you want a similar interface for them, that is at least dened by a common base class, AbstractNode, with two operations: print subtree. These operations should be callable on any node instance, without knowing its actual sub-class.
>>> t1 = Leaf( A, 0.71399) >>> t2 = Node (Leaf(B, -0.00804), Leaf(C, 0.07470)) >>> t3 = Node ( Leaf ( A, 0.71399), Node ( Node ( Leaf(B, -0.00804), Leaf(C, 0.07470), 0.15685), Leaf (D, -0.04732), 0.0666), )

179

Chapter 18. Object-oriented design

>>> print t3 (A: 0.71399, ((B: -0.00804, C: 0.0747): 0.15685, D: -0.04732): 0.0666) >>> t4 = t3.right.subtree() >>> print t4 ((B: -0.00804, C: 0.0747): 0.15685, D: -0.04732) >>> t5 = t3.left.subtree() >>> print t5 A: 0.71399

Figure 18.8. A composite tree

AbstractNode __str__ subtree

Leaf __str__ subtree

Node __str__ subtree

left right

Example 18.7. A composite tree

class AbstractNode: def __str__(self): pass def subtree(self): pass class Node(AbstractNode): def __init__(self, left=None, right=None, length=None): self.left=left self.right=right self.length=length

def __str__(self): if self.length: return "(" + self.left.__str__() + ", " + self.right.__str__() + ")" + ": " + str(self.length else:

180

Chapter 18. Object-oriented design

return "(" + self.left.__str__() + ", " + self.right.__str__() + ")" def subtree(self): return Node(self.left, self.right) class Leaf(AbstractNode): def __init__(self, name, length=None): self.name = name self.length=length self.left = None self.right = None def __str__(self): return self.name + ": " + str(self.length) def subtree(self): return Leaf(self.name, self.length)

if __name__ == "__main__": t1 = Leaf( A, 0.71399) print t1 t2 = Node (Leaf(B, -0.00804), Leaf(C, 0.07470)) print t2 t3 = Node ( Leaf ( A, 0.71399), Node ( Node ( Leaf(B, -0.00804), Leaf(C, 0.07470), 0.15685), Leaf (D, -0.04732), 0.0666), ) print t3

Abstract class AbstractNode, base class for both Node and Leaf Internal nodes are instances of Node class. Leafs are instances of Leaf class. Behavioral patterns. Patterns of this category are very useful in sequence analysis, where you often have to combine algorithms and to analyze complex data structure in a exible way.

181

Chapter 18. Object-oriented design

Template: this pattern consists in separating the invariant part of an algorithm from the variant part. In a sorting procedure, you can generally separate the function which compares items from the main body of the algorithm. The template method, in this case, is the sort() method, whereas the compare() can be dened by each subclass depending on its implementation or data types. Strategy: it is the object-oriented equivalent of passing a function as an argument. In the following example, the f_test method is a strategy:
class MotifDB: def __init__(self, fh=None, db=None): self._db = [] if fh is not None: self._load_fh(fh) elif db is not None: self._load_db(db) def filter(self, f_test) : return self.__class__(db=[motif for motif in self._db if f_test(motif)])

Applied to all elements of a set, this function is provided as a parameter to lter the elements. It can be used the following way:
new_db = db.filter(lambda(motif): kinase in motif.get_desc())

where only elements of this motifs databse containing kinase in their description will be returned. Iterator: an iterator is an object that let you browse a sequence of items from the beginning to the end. Generally, it provides: a method to start iteration a method to get the next item a method to test for the end of the iteration

182

Chapter 18. Object-oriented design

Usually, one distinguishes internal versus external iterators. An external iterator is an iterator which enables to do a for or a while loop on a range of values that are returned by the iterator:
for e in l.elements(): f(e)

or:
i = l.iterator() e = i.next() while e: f(e) e = i.next()

In the above examples, you control the loop. On the other hand, an internal iterator just lets you dene a function or a method (say, f, called a visitor, see below) to apply to all elements:
l.iterate(f)

In the Biopython package, les and databases are generally available through an iterator.
handle = open(...) iter = Bio.Fasta.Iterator(handle) seq = iter.next() while seq: print seq.name print seq.seq seq = iter.next() handle.close()

Starting the iterator. Getting the next element. Testing for the end of the iteration. Getting the next element.

183

Chapter 18. Object-oriented design

Visitor: this pattern is useful to specify a function that will be applied on each item of a collection. The Python map function provides a way to use visitors, such as the f function, which visits each item of the l list in turn:
>>> def f(x): return x + 1 >>> l=[0,1,2] >>> map(f,l) [1, 2, 3]

The map is an example of an internal iterator (with the f function as a visitor). The f_test function in the MotifDB class above is also a visitor. A visitor is also a strategy that applies to all the elements of a set. Observer The observer pattern provides a framework to maintain a consistent distributed state between loosely coupled components. One agent, the observer, is in charge of maintaining a list of subscribers, e.g components that have subscribed to be informed about changes in a given state. Whenever a change occurs in a state, the observer has to inform each subscriber about it. A well-known example is the Model-View-Controller [Krasner88] framework. The view components, the ones who actually display data, subscribe to "edit events" in order to be able to refresh and redisplay them whenever a change occurs.

184

Chapter 18. Object-oriented design

Bibliography
Python Books
[Beazley2001] Python. Essential Reference. 2. David M. Beazley. 0-7357-1091-0. New Riders. 2001. [Christopher2002] Python Programming Patterns. Thomas W Christopher. 0-13-040956-1. Prentice Hall. 2002. [Deitel2002] Python. How to program. H. M. Deitel, P. J. Deitel, J. P. Liperi, and B. A. Wiederman. 0-13-0923613. Prentice Hall. 2002. [Downey2002] How to Think Like a Computer Scientist. Learining with Python. Allen Downey, Jeffrey Elkner, and Chris Meyers. 0-97167775-0-6. Green Tea Press. 2002. www.thinkpython.com [http://www.thinkpython.com]. [Grayson2000] Python and Tkinter Programming. John E. Grayson. 1-884777-81-3. Manning Publications Co.. 2000. [LutzAscher2001] Learning Python. Mark Lutz and David Ascher. 1-56592-464-9. OReilly. 2001. [Lutz2001] Programming Python. Mark Lutz. 0-596-00085-5. OReilly. 2001. [Martelli2002] Python Cookbook. Alex Martelli and David Ascher. 0-596-00167-3. OReilly. 2002.

Other Books
[Meyer97] Object Oriented software Construction. Bertrand Meyer. 0-13-629155-4. Prentice Hall Professional Technical Reference. 1997. [Booch94] Object-Oriented Analysis and Design with Applications. Grady Booch. Addison-Wesley Object Technology Series. 1994. [Gamma95] Design Patterns, Elements of Reusable Object-Oriented Software. Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides. Addison Wesley. 1995. [Harrison97] Tcl/Tk Tools. Mark Harrison. OReilly. 1997.

Articles
[Wegner89] BYTE Magazine. Learning the language. Peter Wegner. 245-253. march, 1989. [Mangalam2002] The Bio* toolkits - a brief overview. Briengs in Bioinformatics. Mangalam Harry. 296-302. 2002. [Rossum99] Computer Programming for Everybody. Guido van Rossum. 1999. [Wang2000] A Qualitative Analysis of the Usability of Perl, Python, and Tcl. International Python Conference. Lingyun Wang and Phil Pfeiffer. 2002. Proceedings of The Tenth

[Krasner88] Journal of Object-Oriented Programming (JOOP). 1. 3. A Cookbook for Using the Model-ViewController User Interface Paradigm in Smalltalk-80. G Krasner and S Pope. 26-49. 1988.

185

Chapter 18. Object-oriented design

186