Python for Delphi Developers - Part 2
0 likes733 views
The document discusses a webinar focused on integrating Python into Delphi applications, presented by Kiriakos Vlahos and Jim McKeeth. It covers the use of Python libraries such as NumPy and TensorFlow, techniques for running Python scripts without blocking the GUI, and the creation of Python extension modules through Delphi. The content emphasizes leveraging Python's capabilities for data analytics, machine learning, and GUI development within Delphi environments.
![VarPyth Demoprocedure TForm1.FormCreate(Sender: TObject);
begin
var np := Import('numpy');
var np_array: Variant :=
np.array(VarPythonCreate([1,2,3,4,5,6,7,8,9,10]));
PythonModule.SetVar('np_array’,
ExtractPythonObjectFrom(np_array));
end;
Python code:
from delphi_module import np_array
print("type(np_array) = ", type(np_array))
print("len(np_array) = ", len(np_array))
print("np_array = ", np_array)
res_array = np_array.copy()
for i in range(len(np_array)):
res_array[i] *= np_array[i]
print("res_array = ", res_array)
Output:
type(np_array) = <class 'numpy.ndarray'>
len(np_array) = 10
np_array = [ 1 2 3 4 5 6 7 8 9 10]
res_array = [ 1 4 9 16 25 36 49 64 81 100]
procedure TForm1.btnRunClick(Sender: TObject);
begin
GetPythonEngine.ExecString(UTF8Encode(sePythonCode.Text));
for var V in VarPyIterate(MainModule.res_array) do
ListBox.Items.Add(V);
end;
To learn more, explore Demo25 and the VarPyth unit tests](https://image.slidesharecdn.com/pythonfordelphidevelopersii-201020010139/85/Python-for-Delphi-Developers-Part-2-4-320.jpg)
Python for Delphi Developers - Part 2
- 1. Python for Delphi Developers (Part II) Webinar by Kiriakos Vlahos (aka PyScripter) and Jim McKeeth (Embarcadero)
- 2. Contents •VarPyth Using python libraries and objects in Delphi code •Core libraries •Visualization •Machine Learning Python based data analytics in Delphi applications •TPythonThread Running python scripts without blocking your GUI Creating Python extension modules using Delphi •WrapDelphi and WrapDelphiVCL units Python GUI development using the VCL
- 3. VarPyth • High-level access to python objects from Delphi code • It also allows you to create common python objects (lists, tuples etc.) • Uses custom variants • No need to use the low-level python API or deal with python object reference counting • Small performance penalty • Example • ShowMessage(SysModule.version) • Explore Demo25 and the VarPyth unit tests
- 4. VarPyth Demoprocedure TForm1.FormCreate(Sender: TObject); begin var np := Import('numpy'); var np_array: Variant := np.array(VarPythonCreate([1,2,3,4,5,6,7,8,9,10])); PythonModule.SetVar('np_array’, ExtractPythonObjectFrom(np_array)); end; Python code: from delphi_module import np_array print("type(np_array) = ", type(np_array)) print("len(np_array) = ", len(np_array)) print("np_array = ", np_array) res_array = np_array.copy() for i in range(len(np_array)): res_array[i] *= np_array[i] print("res_array = ", res_array) Output: type(np_array) = <class 'numpy.ndarray'> len(np_array) = 10 np_array = [ 1 2 3 4 5 6 7 8 9 10] res_array = [ 1 4 9 16 25 36 49 64 81 100] procedure TForm1.btnRunClick(Sender: TObject); begin GetPythonEngine.ExecString(UTF8Encode(sePythonCode.Text)); for var V in VarPyIterate(MainModule.res_array) do ListBox.Items.Add(V); end; To learn more, explore Demo25 and the VarPyth unit tests
- 5. Core Python Libraries • Core libraries • numpy – arrays and matrix algebra • scipy – math, science and engineering • pandas – data structure and analysis, R-like dataframes • Data visualization • matplotlib – matlab-like plotting • seaborn – statistical data visualization • mpld3 – turn matplotlib plots into interactive web pages • bokeh - interactive visualization library for modern browsers • plotly – interactive browser-based graphing library • altair – based on the visualization grammar vega-light
- 6. Data visualization Demos - pyVIZsvg • Create charts with python libraries, save them in svg format and plot them in Delphi • Uses matplotlib and seaborn python libraries • Uses TSVGIconImage from EtheaDev SVGIconImageList components
- 7. Interactive Data Visualization Demo - PychartHtml • Create interactive charts in python save them to html and show them inside Delphi applications • Showcases the new TEdgeBrowser • Uses the matplotlib with mpld3, altair and bohek python libraries
- 8. Machine Learning Python Libraries • tensorflow (Google) • keras – high level pythonic interface • PyTorch (Facebook) • CNTK - The Microsoft Cognitive Toolkit (Microsoft) • Spark MLlib and mxnet (Apache Foundation) • scikit-learn • pure-python, general library, wide coverage, good choice for newcomers to ML/AI
- 9. Data Analytics Demo: COVID-19 • Demonstrates the combined use of numpy, pandas, matplotlib and scikit-learn • Use pandas to download Covid-19 confirmed cases data from Web and aggregate it • Use scikit-learn to • Split the data into training and test sets • Transform the data • Define model (Bayesian Ridge Regression) • Optimize model parameters • Generate predictions • Use matplotlib to compare actual vrs fitted test data.
- 10. Running Python Scripts Without Blocking Your Main Thread • Python uses the infamous Global Interpreter Lock (GIL) • This means that only one thread can run python code at any given time • Python switches between python created threads automatically • Applications using python need to acquire the GIL before running python code and release it soon afterwards • Welcome to TPythonThread • TThread descendent • Automatically acquires and releases the GIL
- 11. Python threads – demo33 • Shows you how to • use TPythonThread • use new sub-interpreters • synchronize with the main thread • interrupt running threads with a keyboard interrupt exception
- 12. Technical Aside I • FPU Exception mask • Delphi’s default FPU exception mask is different from most other Windows apps • Incompatible with python libraries written in C or C++ • If you use numpy, scipy, tensorflow etc. you need to match the FPU mask they expect to operate with • PythonEngine.pas provides a function for doing that: MaskFPUExceptions • Call MaskFPUExceptions(True) before python is loaded • e.g. the initialization section of your main form • See the P4D Wiki page for details
- 13. Technical Aside II • Working with different python distributions: • P4D makes a good effort to discover registered python distributions automatically • But sometimes you want to use python distributions which are not registered • One such case is when want to deploy your application bundled with python. Python.org offers such a minimal distribution for applications embedding python. • Additional considerations apply when using an Anaconda distribution • Read the Finding Python P4D Wiki page carefully
- 14. P4D Python Extension Modules • Python extension modules are dynamic link libraries that can be used by python in a way similar to modules developed in python. • They have ‘pyd’ extension. • P4D makes it very easy to create extension modules that contain functions and types developed in Delphi. • These extension modules can be used by Python, independently of Delphi, and can be packaged with setuptools and distributed through PyPi.
- 15. Python extension modules Demo function PyInit_DemoModule: PPyObject; //function exported by dll – initializes the module begin try gEngine := TPythonEngine.Create(nil); gEngine.AutoFinalize := False; gEngine.UseLastKnownVersion := False; // Adapt to the desired python version gEngine.RegVersion := '3.8'; gEngine.DllName := 'python38.dll'; gModule := TPythonModule.Create(nil); gModule.Engine := gEngine; gModule.ModuleName := 'DemoModule'; gModule.AddMethod('is_prime', delphi_is_prime, 'is_prime(n) -> bool' ); gEngine.LoadDll; except end; Result := gModule.Module; end; Python test module: test.py from DemoModule import is_prime from timeit import Timer def count_primes(max_n): res = 0 for i in range(2, max_n + 1): if is_prime(i): res += 1 return res Command line: > C:PythonPython38python test.py Number of primes between 0 and 1000000 = 78498 Elapsed time: 0.29756570000000004 secs
- 16. Wrapping VCL as a Python Extension Module I • We can use the same approach to create an extension module wrapping VCL • Uses the WrapDelphi and WrapDelphiVCL units • The whole of VCL (almost) is wrapped in 40 lines of code • The generated Delphi.pyd file is only 5Mbs • It can be uses to create VCL-based GUIs in python without needing Delphi • A similar approach could be used to wrap FMX and create a multi-platform python GUI library • Unclear whether there are licensing restrictions in distributing such a file
- 17. Wrapping VCL as a Python Extension Module II unit uMain; interface uses PythonEngine; function PyInit_Delphi: PPyObject; cdecl; implementation uses WrapDelphi, WrapDelphiVCL; // similar to the previous extension module TestApp.py from Delphi import * class MainForm(Form): def __init__(self, Owner): self.Caption = "A Delphi Form..." self.SetBounds(10, 10, 500, 400) self.lblHello = Label(self) self.lblHello.SetProps(Parent=self, Caption="Hello World") self.lblHello.SetBounds(10, 10, 300, 24) self.OnClose = self.MainFormClose def MainFormClose(self, Sender, Action): Action.Value = caFree def main(): Application.Initialize() Application.Title = "MyDelphiApp" f = MainForm(Application) f.Show() FreeConsole() Application.Run() main() Command line: > C:PythonPython38python .TestApp.py
- 18. Summary • With Python for Delphi you can get the best of both worlds • P4D makes it very easy to integrate Python into Delphi applications in RAD way, thus providing access to a vast range of python libraries • Expose Delphi function, objects, records and types to Python using low or high-level interfaces (WrapDelphi) • Create/Access/Use Python objects/modules in your Delphi code using a high- level interface (VarPyth) • Run python code in threads • Create python extensions modules • Wrap Vcl as a Python extension module to create GUIs with python github.com/pyscripter/python4delphi/tree/master/Tutorials/Webinar%20II