4. Building C and C++ Extensions on Windows

This chapter briefly explains how to create a Windows extension module for Python using Microsoft Visual C++, and follows with more detailed background information on how it works. The explanatory material is useful for both the Windows programmer learning to build Python extensions and the Unix programmer interested in producing software which can be successfully built on both Unix and Windows.

Module authors are encouraged to use the distutils approach for building extension modules, instead of the one described in this section. You will still need the C compiler that was used to build Python; typically Microsoft Visual C++.


This chapter mentions a number of filenames that include an encoded Python version number. These filenames are represented with the version number shown as XY; in practice, 'X' will be the major version number and 'Y' will be the minor version number of the Python release you’re working with. For example, if you are using Python 2.2.1, XY will actually be 22.

4.1. A Cookbook Approach

There are two approaches to building extension modules on Windows, just as there are on Unix: use the distutils package to control the build process, or do things manually. The distutils approach works well for most extensions; documentation on using distutils to build and package extension modules is available in Distributing Python Modules. This section describes the manual approach to building Python extensions written in C or C++.

To build extensions using these instructions, you need to have a copy of the Python sources of the same version as your installed Python. You will need Microsoft Visual C++ “Developer Studio”; project files are supplied for VC++ version 7.1, but you can use older versions of VC++. Notice that you should use the same version of VC++that was used to build Python itself. The example files described here are distributed with the Python sources in the PC\example_nt\ directory.

  1. Copy the example files — The example_nt directory is a subdirectory of the PC directory, in order to keep all the PC-specific files under the same directory in the source distribution. However, the example_nt directory can’t actually be used from this location. You first need to copy or move it up one level, so that example_nt is a sibling of the PC and Include directories. Do all your work from within this new location.

  2. Open the project — From VC++, use the File ‣ Open Solution dialog (not File ‣ Open!). Navigate to and select the file example.sln, in the copy of the example_nt directory you made above. Click Open.

  3. Build the example DLL — In order to check that everything is set up right, try building:

  4. Select a configuration. This step is optional. Choose Build ‣ Configuration Manager ‣ Active Solution Configuration and select either Release or Debug. If you skip this step, VC++ will use the Debug configuration by default.

  5. Build the DLL. Choose Build ‣ Build Solution. This creates all intermediate and result files in a subdirectory called either Debug or Release, depending on which configuration you selected in the preceding step.

  6. Testing the debug-mode DLL — Once the Debug build has succeeded, bring up a DOS box, and change to the example_nt\Debug directory. You should now be able to repeat the following session (C> is the DOS prompt, >>> is the Python prompt; note that build information and various debug output from Python may not match this screen dump exactly):

    Adding parser accelerators ...
    Python 2.2 (#28, Dec 19 2001, 23:26:37) [MSC 32 bit (Intel)] on win32
    Type "copyright", "credits" or "license" for more information.
    >>> import example
    [4897 refs]
    >>> example.foo()
    Hello, world
    [4903 refs]

    Congratulations! You’ve successfully built your first Python extension module.

  7. Creating your own project — Choose a name and create a directory for it. Copy your C sources into it. Note that the module source file name does not necessarily have to match the module name, but the name of the initialization function should match the module name — you can only import a module spam if its initialization function is called initspam(), and it should call Py_InitModule() with the string "spam" as its first argument (use the minimal example.c in this directory as a guide). By convention, it lives in a file called spam.c or spammodule.c. The output file should be called spam.pyd (in Release mode) or spam_d.pyd (in Debug mode). The extension .pyd was chosen to avoid confusion with a system library spam.dll to which your module could be a Python interface.

    Changed in version 2.5: Previously, file names like spam.dll (in release mode) or spam_d.dll (in debug mode) were also recognized.

    Now your options are:

  8. Copy example.sln and example.vcproj, rename them to spam.*, and edit them by hand, or

  9. Create a brand new project; instructions are below.

    In either case, copy example_nt\example.def to spam\spam.def, and edit the new spam.def so its second line contains the string ‘initspam‘. If you created a new project yourself, add the file spam.def to the project now. (This is an annoying little file with only two lines. An alternative approach is to forget about the .def file, and add the option /export:initspam somewhere to the Link settings, by manually editing the setting in Project Properties dialog).

  10. Creating a brand new project — Use the File ‣ New ‣ Project dialog to create a new Project Workspace. Select Visual C++ Projects/Win32/ Win32 Project, enter the name (spam), and make sure the Location is set to parent of the spam directory you have created (which should be a direct subdirectory of the Python build tree, a sibling of Include and PC). Select Win32 as the platform (in my version, this is the only choice). Make sure the Create new workspace radio button is selected. Click OK.

    You should now create the file spam.def as instructed in the previous section. Add the source files to the project, using Project ‣ Add Existing Item. Set the pattern to *.* and select both spam.c and spam.def and click OK. (Inserting them one by one is fine too.)

    Now open the Project ‣ spam properties dialog. You only need to change a few settings. Make sure All Configurations is selected from the Settings for: dropdown list. Select the C/C++ tab. Choose the General category in the popup menu at the top. Type the following text in the entry box labeled Additional Include Directories:


    Then, choose the General category in the Linker tab, and enter


    in the text box labelled Additional library Directories.

    Now you need to add some mode-specific settings:

    Select Release in the Configuration dropdown list. Choose the Link tab, choose the Input category, and append pythonXY.lib to the list in the Additional Dependencies box.

    Select Debug in the Configuration dropdown list, and append pythonXY_d.lib to the list in the Additional Dependencies box. Then click the C/C++ tab, select Code Generation, and select Multi-threaded Debug DLL from the Runtime library dropdown list.

    Select Release again from the Configuration dropdown list. Select Multi-threaded DLL from the Runtime library dropdown list.

If your module creates a new type, you may have trouble with this line:


Static type object initializers in extension modules may cause compiles to fail with an error message like “initializer not a constant”. This shows up when building DLL under MSVC. Change it to:


and add the following to the module initialization function:

if (PyType_Ready(&MyObject_Type) < 0)
     return NULL;

4.2. Differences Between Unix and Windows

Unix and Windows use completely different paradigms for run-time loading of code. Before you try to build a module that can be dynamically loaded, be aware of how your system works.

In Unix, a shared object (.so) file contains code to be used by the program, and also the names of functions and data that it expects to find in the program. When the file is joined to the program, all references to those functions and data in the file’s code are changed to point to the actual locations in the program where the functions and data are placed in memory. This is basically a link operation.

In Windows, a dynamic-link library (.dll) file has no dangling references. Instead, an access to functions or data goes through a lookup table. So the DLL code does not have to be fixed up at runtime to refer to the program’s memory; instead, the code already uses the DLL’s lookup table, and the lookup table is modified at runtime to point to the functions and data.

In Unix, there is only one type of library file (.a) which contains code from several object files (.o). During the link step to create a shared object file (.so), the linker may find that it doesn’t know where an identifier is defined. The linker will look for it in the object files in the libraries; if it finds it, it will include all the code from that object file.

In Windows, there are two types of library, a static library and an import library (both called .lib). A static library is like a Unix .a file; it contains code to be included as necessary. An import library is basically used only to reassure the linker that a certain identifier is legal, and will be present in the program when the DLL is loaded. So the linker uses the information from the import library to build the lookup table for using identifiers that are not included in the DLL. When an application or a DLL is linked, an import library may be generated, which will need to be used for all future DLLs that depend on the symbols in the application or DLL.

Suppose you are building two dynamic-load modules, B and C, which should share another block of code A. On Unix, you would not pass A.a to the linker for B.so and C.so; that would cause it to be included twice, so that B and C would each have their own copy. In Windows, building A.dll will also build A.lib. You do pass A.lib to the linker for B and C. A.lib does not contain code; it just contains information which will be used at runtime to access A’s code.

In Windows, using an import library is sort of like using import spam; it gives you access to spam’s names, but does not create a separate copy. On Unix, linking with a library is more like from spam import *; it does create a separate copy.

4.3. Using DLLs in Practice

Windows Python is built in Microsoft Visual C++; using other compilers may or may not work (though Borland seems to). The rest of this section is MSVC++ specific.

When creating DLLs in Windows, you must pass pythonXY.lib to the linker. To build two DLLs, spam and ni (which uses C functions found in spam), you could use these commands:

cl /LD /I/python/include spam.c ../libs/pythonXY.lib
cl /LD /I/python/include ni.c spam.lib ../libs/pythonXY.lib

The first command created three files: spam.obj, spam.dll and spam.lib. Spam.dll does not contain any Python functions (such as PyArg_ParseTuple()), but it does know how to find the Python code thanks to pythonXY.lib.

The second command created ni.dll (and .obj and .lib), which knows how to find the necessary functions from spam, and also from the Python executable.

Not every identifier is exported to the lookup table. If you want any other modules (including Python) to be able to see your identifiers, you have to say _declspec(dllexport), as in void _declspec(dllexport) initspam(void) or PyObject _declspec(dllexport) *NiGetSpamData(void).

Developer Studio will throw in a lot of import libraries that you do not really need, adding about 100K to your executable. To get rid of them, use the Project Settings dialog, Link tab, to specify ignore default libraries. Add the correct msvcrtxx.lib to the list of libraries.