Boost C++ Libraries

...one of the most highly regarded and expertly designed C++ library projects in the world. — Herb Sutter and Andrei Alexandrescu, C++ Coding Standards

Boost Library Requirements and Guidelines

Introduction

This page describes requirements and guidelines for the content of a library submitted to Boost.

See the Boost Library Submission Process page for a description of the process involved.

Requirements

To avoid the frustration and wasted time of a proposed library being rejected, it must meet these requirements:

There's no requirement that an author read the mailing list for a time before making a submission. It has been noted, however, that submissions which begin "I just started to read this mailing list ..." seem to fail, often embarrassingly.

License requirements

The preferred way to meet the license requirements is to use the Boost Software License. See license information. If for any reason you do not intend to use the Boost Software License, please discuss the issues on the Boost developers mailing list first.

The license requirements:

  • Must be simple to read and understand.
  • Must grant permission without fee to copy, use and modify the software for any use (commercial and non-commercial).
  • Must require that the license appear on all copies of the software source code.
  • Must not require that the license appear with executables or other binary uses of the library.
  • Must not require that the source code be available for execution or other binary uses of the library.
  • May restrict the use of the name and description of the library to the standard version found on the Boost web site.

Portability requirements

  • A library's interface must portable and not restricted to a particular compiler or operating system.
  • A library's implementation must if possible be portable and not restricted to a particular compiler or operating system. If a portable implementation is not possible, non-portable constructions are acceptable if reasonably easy to port to other environments, and implementations are provided for at least two popular operating systems (such as UNIX and Windows).
  • A library runs on at least two C++ compilers implementing the latest ISO C++ standard.
  • There is no requirement that a library run on C++ compilers which do not conform to the ISO standard.
  • There is no requirement that a library run on any particular C++ compiler. Boost contributors often try to ensure their libraries work with popular compilers. The boost/config.hpp configuration header is the preferred mechanism for working around compiler deficiencies.

Since there is no absolute way to prove portability, many boost submissions demonstrate practical portability by compiling and executing correctly with two different C++ compilers, often under different operating systems. Otherwise reviewers may disbelieve that porting is in fact practical.

Ownership

Are you sure you own the library you are thinking of submitting? "How to Copyright Software" by MJ Salone, Nolo Press, 1990 says:

Doing work on your own time that is very similar to programming you do for your employer on company time can raise nasty legal problems. In this situation, it's best to get a written release from your employer in advance.

Place a copyright notice in all the important files you submit. Boost won't accept libraries without clear copyright information.

Organization

The quality of the Boost libraries is not just about the APIs and code design. But also about presenting a consistent view to users of the libraries as a whole. Upon acceptance libraries must adhere to this directory and file structure:

Boost standard library organization
Sub-directory or file Contents Required
build Library build files such as a Jamfile, IDE projects, Makefiles, Cmake files, etc. Required if the library has sources to build.
config Files used for build-time configuration checks. This directory may contain source files and build system scripts to be used when building the library, tests or examples to check if the target system satisfies certain conditions. For example, a check may test if the compiler implements a certain feature, or if the target system supports a certain API. Optional.
doc Sources to build with and built documentation for the library. If the library needs to build documentation from non-HTML files this location must be buildable with Boost Build. Required for all libraries.
doc/html Documentation (HTML) files. Required for all libraries with pregenerated documentation. And generated documentation must be generated here.
example Sample program files. Required if library has sample files. Which is highly recommended.
index.html Redirection to HTML documentation. See "Redirection" for a template for this file. Required for all libraries.
include/boost/library Header files for the library. Required for all libraries.
meta Meta-data about the library. Required for all libraries.
meta/libraries.json A JSON file containing information about the library used to generate website and documentation for the Boost C++ Libraries collection. Required for all libraries.
meta/explicit-failures-markup.xml XML file describing expected test failures, used to generate the test report. Optional
src Source files which must be compiled to build the library. Required if the library has source files to build.
test Regression or other test programs or scripts. This is the only location considered for automated testing. If you have additional locations that need to be part of automated testing it is required that this location refer to the additional test locations. Required for all libraries.
tools Tools used, or offered, by the library. The structure within this is up to the library, but it's recommended to use similar structure as a regular Boost library or tool. Required for libraries that have runable tools.

Integration

Once a library is accepted as part of the Boost C++ Libraries it is required that it integrate properly into the development, testing, documentation, and release processes. This integration increases the eventual quality of all the libraries and is integral to the expected quality of the whole of the Boost C++ Libraries from users. In addition to the organization requirements above the following integration is required:

Building Sources

The library needs to provide a Boost Build project that the user, and the top level Boost project, can use to build the library if it has sources to build. The Jamfile for the source build needs to minimally declare the project, the library target(s), and register the target(s) for installation. For example:

project boost/my_lib ;

lib boost_my_lib : a.cpp ;

boost-install boost_my_lib ;

Testing

The library needs to provide a Boost Build project that the user, and the root Boost test script, can use to build and run the tests for the library. The testing build project must reside in the project-root/test directory and must be buildable from this or another directory (for example, b2 libs/library/test from the Boost root must work.)

An example test/Jamfile is given below:

import testing ;

run default_constructor.cpp ;
run copy_constructor.cpp ;
compile nested_value_type.cpp ;
compile-fail invalid_conversion_1.cpp ;

WARNING: This is the only location considered for testing by the top level testing script. If you want to test additional locations you must declare such that they are built as dependencies or by using build-project.

If the library requires a level of C++ conformance that precludes certain compilers or configurations from working, it's possible (and recommended) to declare these requirements in the test Jamfile so that the tests aren't run, to conserve test resources, as given in the example below:

import testing ;
import ../../config/checks/config : requires ;

project : requirements [ requires cxx11_variadic_templates cxx11_template_aliases ] ;

run cpp11_test.cpp ;

For more information, see the documentation of Boost.Config.

Building Documentation

The library needs to provide a Boost Build project for building the documentation for the library. The project-root/doc project is the only location refered to by the top level documentation build scripts and the release building scripts. The documentation build project must have the following two features:

  1. Define a boostdoc target. This target should likely be an alias that looks roughly like:
    alias boostdoc : my_boostbook_target
    : : : <implicit-dependency>my_boostbook_target ;
    But if your project doesn't integrate into the global documentation book you can use an empty alias like:
    alias boostdoc ;
  2. The project must default to building standalone documentation if it has any. The release scripts build this default so as to guarantee all projects have up to date documentation.

Guidelines

Please use these guidelines as a checklist for preparing the content a library submission. Not every guideline applies to every library, but a reasonable effort to comply is expected.

Backwards Compatibility

Boost libraries generally have a large and diverse user base. To ensure successful transitions from old APIs to newer APIs under those circumstances, library authors are encouraged to follow a few guidelines when introducing breaking changes in their library:
  1. Non-breaking changes can be done without restriction.
  2. Small breaking changes can be made, but users should be given notice a few releases before the change is published. Most breaking changes fall into this category.
  3. For large breaking changes with a migration path from the old API to the new API (for example boost::filesystem v2 to v3), the new API should be introduced in a separate directory/namespace, and users should be noticed and given a few releases to move over. The old API can be removed after some time.
  4. For large breaking changes without a migration path (for example boost::spirit v2 to v3), the new API should be provided in a separate directory/namespace, and the old API should be preserved (because there's no migration path). Removing the API should be considered the same as removing a Boost library, which can be done but needs a more extensive deprecation period.
  5. Large breaking changes that are equivalent to a redesign or rewrite of the library should be treated as a new library and a formal review (or at least a mini review) is encouraged.

Design and Programming

Aim first for clarity and correctness; optimization should be only a secondary concern in most Boost libraries.

Aim for ISO Standard C++. Than means making effective use of the standard features of the language, and avoiding non-standard compiler extensions. It also means using the C++ Standard Library where applicable.

Headers should be good neighbors. See the header policy. See Naming consistency.

Follow quality programming practices. See, for example, "Effective C++" 2nd Edition, and "More Effective C++", both by Scott Meyers, published by Addison Wesley.

Use the C++ Standard Library or other Boost libraries, but only when the benefits outweigh the costs. Do not use libraries other than the C++ Standard Library or Boost. See Library reuse.

Read Implementation Variation to see how to supply performance, platform, or other implementation variations.

Browse through the Best Practices Handbook for ideas and links to source code in existing Boost libraries.

Read the guidelines for libraries with separate source to see how to ensure that compiled link libraries meet user expectations.

Use the naming conventions of the C++ Standard Library (See Naming conventions rationale):

  • Names (except as noted below) should be all lowercase, with words separated by underscores.
  • Acronyms should be treated as ordinary names (e.g. xml_parser instead of XML_parser).
  • Template parameter names begin with an uppercase letter.
  • Macro (gasp!) names all uppercase and begin with BOOST_.

Choose meaningful names - explicit is better than implicit, and readability counts. There is a strong preference for clear and descriptive names, even if lengthy.

Use exceptions to report errors where appropriate, and write code that is safe in the face of exceptions.

Avoid exception-specifications. See exception-specification rationale.

Provide sample programs or confidence tests so potential users can see how to use your library.

Provide a regression test program or programs which follow the Test Policies and Protocols.

Although some boost members use proportional fonts, tabs, and unrestricted line lengths in their own code, boost's widely distributed source code should follow more conservative guidelines:

End all documentation files (HTML or otherwise) with a copyright message and a licensing message. See the license information page for the preferred form.

Begin all source files (including programs, headers, scripts, etc.) with:

  • A comment line describing the contents of the file.
  • Comments describing copyright and licensing: again, the preferred form is indicated in the license information page
  • Note that developers are allowed to provide a copy of the license text in LICENSE_1_0.txt, LICENSE.txt or LICENSE file within repositories of their libraries.
  • A comment line referencing your library on the Boost web site. For example: 
    // See https://www.boost.org/libs/foo for library home page.

    Where foo is the directory name (see below) for the library. As well as aiding users who come across a Boost file detached from its documentation, some of Boost's automatic tools depend on this comment to identify which library header files belong to.

Assertions: If you want to add runtime assertions to your code (you should!), avoid C's assert macro and use Boost's BOOST_ASSERT macro (in boost/assert.hpp ) instead. It is more configurable. Use BOOST_ASSERT in public headers and in library source code (for separately compiled libraries). Use of C's assert macro is ok in examples and in documentation.

Make sure your code compiles in the presence of the min() and max() macros. Some platform headers define min() and max() macros which cause some common C++ constructs to fail to compile. Some simple tricks can protect your code from inappropriate macro substitution:

  • If you want to call std::min() or std::max():
    • If you do not require argument-dependent look-up, use (std::min)(a,b).
    • If you do require argument-dependent look-up, you should:
      • #include <boost/config.hpp>
      • Use BOOST_USING_STD_MIN(); to bring std::min() into the current scope.
      • Use min BOOST_PREVENT_MACRO_SUBSTITUTION (a,b); to make an argument-dependent call to min(a,b).
  • If you want to call std::numeric_limits<int>::max(), use (std::numeric_limits<int>::max)() instead.
  • If you want to call a min() or max() member function, instead to doing obj.min(), use (obj.min)().
  • If you want to declare or define a function or a member function named min or max, then you must use the BOOST_PREVENT_MACRO_SUBSTITUTION macro. Instead of writing int min() { return 0; } you should write int min BOOST_PREVENT_MACRO_SUBSTITUTION () { return 0; } This is true regardless if the function is a free (namespace scope) function, a member function or a static member function, and it applies for the function declaration as well as for the function definition.

Filenames

Naming requirements ensure that file and directory names are relatively portable, including to ISO 9660:1999 (with extensions) and other relatively limited file systems. Superscript links are provided to detailed rationale for each choice.

  • Names must contain only lowercase1 ASCII letters ('a'-'z'), numbers ('0'-'9'), underscores ('_'), hyphens ('-'), and periods ('.'). Spaces are not allowed2.
  • Directory names must not contain periods ('.')3.
  • The first and last character of a file name must not be a period ('.')4.
  • The first character of names must not be a hyphen ('-')5.
  • The maximum length of directory and file names is 31 characters6.
  • The total path length must not exceed 207 characters7.

Other conventions ease communication:

  • Files intended to be processed by a C++ compiler as part of a translation unit should have a three-letter filename extension ending in "pp". Other files should not use extensions ending in "pp". This convention makes it easy to identify all of the C++ source in Boost.
  • All libraries have at their highest level a primary directory named for the particular library. See Naming consistency. The primary directory may have sub-directories.

Redirection

The primary directory should always contain a file named index.html. Authors have requested this so that they can publish URL's in the form https://www.boost.org/libs/lib-name with the assurance a documentation reorganization won't invalidate the URL. Boost's internal tools are also simplified by knowing that a library's documentation is always reachable via the simplified URL.

The primary directory index.html file should just do an automatic redirection to the doc/html subdirectory:

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
    "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">

<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
<head>
  <title>Boost.Name Documentation</title>
  <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
  <meta http-equiv="refresh" content="0; URL=doc/html/index.html" />
</head>

<body>
  Automatic redirection failed, please go to <a href=
  "doc/index.html">doc/index.html</a>
</body>
</html>

Naming consistency

As library developers and users have gained experience with Boost, the following consistent naming approach has come to be viewed as very helpful, particularly for larger libraries that need their own header subdirectories and namespaces.

Here is how it works. The library is given a name that describes the contents of the library. Cryptic abbreviations are strongly discouraged. Following the practice of the C++ Standard Library, names are usually singular rather than plural. For example, a library dealing with file systems might chose the name "filesystem", but not "filesystems", "fs" or "nicecode".

  • The library's primary directory (in parent boost-root/libs) is given that same name. For example, boost-root/libs/filesystem.
  • The library's primary header directory (in boost-root/libs/name/include) is given that same name. For example, boost-root/libs/filesystem/boost/filesystem.
  • The library's primary namespace (in parent ::boost) is given that same name, except when there's a component with that name (e.g., boost::tuple), in which case the namespace name is pluralized. For example, ::boost::filesystem.

When documenting Boost libraries, follow these conventions (see also the following section of this document):

  • The library name is set in roman type.
  • The library name is capitalized.
  • A period between "Boost" and the library name (e.g., Boost.Bind) is used if and only if the library name is not followed by the word "library".
  • The word "library" is not part of the library name and is therefore lowercased.

Here are a few examples of how to apply these conventions:

  • Boost.Bind was written by Peter Dimov.
  • The Boost Bind library was written by Peter Dimov.
  • I regularly use Bind, a Boost library written by Peter Dimov.

Documentation

Even the simplest library needs some documentation; the amount should be proportional to the need. The documentation should assume the readers have a basic knowledge of C++, but are not necessarily experts.

The format for documentation should be HTML, and should not require an advanced browser or server-side extensions. Style sheets are acceptable. ECMAScript/JavaScript is discouraged. The documentation entry point should always be a file named index.html; see Redirection.

There is no single right way to do documentation. HTML documentation is often organized quite differently from traditional printed documents. Task-oriented styles differ from reference oriented styles. In the end, it comes down to the question: Is the documentation sufficient for the mythical "average" C++ programmer to use the library successfully?

Appropriate topics for documentation often include:

  • General introduction to the library. The introduction particularly needs to include:
    • A very high-level overview of what the library is good for, and perhaps what it isn't good for, understandable even by those with no prior knowledge of the problem domain.
    • The simplest possible ("hello world") example of using the library.
  • Tutorial covering basic use cases.
  • Reference documentation:
    • Description of each class.
    • Relationship between classes.
    • For each function, as applicable, description, requirements (preconditions), effects, post-conditions, returns, and throws.
    • Discussion of error detection and recovery strategy.
  • How to compile and link.
  • How to test.
  • Version or revision history.
  • Rationale for design decisions. See Rationale rationale.
  • Acknowledgements. See Acknowledgments rationale.

If you need more help with how to write documentation you can check out the article on Writing Documentation for Boost.

Rationale

Rationale for some of the requirements and guidelines follows.

Exception-specification rationale

Exception specifications [ISO 15.4] are sometimes coded to indicate what exceptions may be thrown, or because the programmer hopes they will improve performance. But consider the following member from a smart pointer:

T& operator*() const throw()  { return *ptr; }

This function calls no other functions; it only manipulates fundamental data types like pointers Therefore, no runtime behavior of the exception-specification can ever be invoked. The function is completely exposed to the compiler; indeed it is declared inline Therefore, a smart compiler can easily deduce that the functions are incapable of throwing exceptions, and make the same optimizations it would have made based on the empty exception-specification. A "dumb" compiler, however, may make all kinds of pessimizations.

For example, some compilers turn off inlining if there is an exception-specification. Some compilers add try/catch blocks. Such pessimizations can be a performance disaster which makes the code unusable in practical applications.

Although initially appealing, an exception-specification tends to have consequences that require very careful thought to understand. The biggest problem with exception-specifications is that programmers use them as though they have the effect the programmer would like, instead of the effect they actually have.

A non-inline function is the one place a "throws nothing" exception-specification may have some benefit with some compilers.

Naming conventions rationale

The C++ standard committee's Library Working Group discussed this issue in detail, and over a long period of time. The discussion was repeated again in early boost postings. A short summary:

  • Naming conventions are contentious, and although several are widely used, no one style predominates.
  • Given the intent to propose portions of boost for the next revision of the C++ standard library, boost decided to follow the standard library's conventions.
  • Once a library settles on a particular convention, a vast majority of stakeholders want that style to be consistently used.

Source code fonts rationale

Dave Abrahams comments: An important purpose (I daresay the primary purpose) of source code is communication: the documentation of intent. This is a doubly important goal for boost, I think. Using a fixed-width font allows us to communicate with more people, in more ways (diagrams are possible) right there in the source. Code written for fixed-width fonts using spaces will read reasonably well when viewed with a variable-width font, and as far as I can tell every editor supporting variable-width fonts also supports fixed width. I don't think the converse is true.

Tabs rationale

Tabs are banned because of the practical problems caused by tabs in multi-developer projects like Boost, rather than any dislike in principle. See mailing list archives. Problems include maintenance of a single source file by programmers using tabs and programmers using spaces, and the difficulty of enforcing a consistent tab policy other than just "no tabs". Discussions concluded that Boost files should either all use tabs, or all use spaces, and thus the decision to stick with spaces for indentation.

Directory and File Names rationale

1. Some legacy file systems require single-case names. Single-case names eliminate casing mistakes when moving from case-insensitive to case-sensitive file systems.

2. This is the lowercase portion of the POSIX portable filename character set. To quote the POSIX standard, "Filenames should be constructed from the portable filename character set because the use of other characters can be confusing or ambiguous in certain contexts."

3. Strict implementations of ISO 9660:1999 and some legacy operating systems prohibit dots in directory names. The need for this restriction is fading, and it will probably be removed fairly soon.

4. POSIX has special rules for names beginning with a period. Windows prohibits names ending in a period.

5. Would be too confusing or ambiguous in certain contexts.

6. We had to draw the line somewhere, and so the limit imposed by a now obsolete Apple file system was chosen years ago. It still seems a reasonable limit to aid human comprehension.

7. ISO 9660:1999.

ECMAScript/JavaScript rationale

Before the 1.29.0 release, two Boost libraries added ECMAScript/JavaScript documentation. Controversy followed (see mailing list archives), and the developers were asked to remove the ECMAScript/JavaScript. Reasons given for banning included:

  • Incompatible with some older browsers and some text based browsers.
  • Makes printing docs pages difficult.
  • Often results in really bad user interface design.
  • "It's just annoying in general."
  • Would require Boost to test web pages for ECMAScript/JavaScript compliance.
  • Makes docs maintenance by other than the original developer more difficult.

Please conside those reasons if you decide that JavaScript is something you must use. In particular please keep in mind that the Boost community is not responsible for testing your use of JavaScript. And hence it is up to you to ensure that the above issues are fully resolved in your use case.

ECMAScript/JavaScript use is allowed but discouraged for the reasons above.

Rationale rationale

Rationale is defined as "The fundamental reasons for something; basis" by the American Heritage Dictionary.

Beman Dawes comments: Failure to supply contemporaneous rationale for design decisions is a major defect in many software projects. Lack of accurate rationale causes issues to be revisited endlessly, causes maintenance bugs when a maintainer changes something without realizing it was done a certain way for some purpose, and shortens the useful lifetime of software.

Rationale is fairly easy to provide at the time decisions are made, but very hard to accurately recover even a short time later.

Acknowledgements rationale

As a library matures, it almost always accumulates improvements suggested to the authors by other boost members. It is a part of the culture of boost.org to acknowledge such contributions, identifying the person making the suggestion. Major contributions are usually acknowledged in the documentation, while minor fixes are often mentioned in comments within the code itself.