bpo-31582: Created a new section describing sys.path initialization

Doc/library/importlib.rst

@@ -51,6 +51,9 @@ managing aspects of Python packages:
     The :func:`.__import__` function
         The :keyword:`import` statement is syntactic sugar for this function.
 
+    :ref:`sys-path-init`
+        The initialization of :data:`sys.path`.
+
     :pep:`235`
         Import on Case-Insensitive Platforms
 

Doc/library/modules.rst

@@ -19,3 +19,4 @@ The full list of modules described in this chapter is:
    importlib.rst
    importlib.resources.rst
    importlib.metadata.rst
+   sys_path_init.rst

Doc/library/site.rst

@@ -276,4 +276,6 @@ value greater than 2 if there is an error.
 
 .. seealso::
 
-   :pep:`370` -- Per user site-packages directory
+   * :pep:`370` -- Per user site-packages directory
+   * :ref:`sys-path-init` -- The initialization of :data:`sys.path`.
+

Doc/library/sys.rst

@@ -1126,15 +1126,16 @@ always available.
    current directory first.  Notice that the script directory is inserted *before*
    the entries inserted as a result of :envvar:`PYTHONPATH`.
 
+   The initialization of :data:`sys.path` is documented at :ref:`sys-path-init`.
+
    A program is free to modify this list for its own purposes.  Only strings
    and bytes should be added to :data:`sys.path`; all other data types are
    ignored during import.
 
 
    .. seealso::
-      Module :mod:`site` This describes how to use .pth files to extend
-      :data:`sys.path`.
-
+      * Module :mod:`site` This describes how to use .pth files to
+        extend :data:`sys.path`.
 
 .. data:: path_hooks
 

Doc/library/sys_path_init.rst

@@ -0,0 +1,108 @@
+.. _sys-path-init:
+
+The initialization of the :data:`sys.path` module search path
+=============================================================
+
+A module search path is initialized when Python starts. This module search path
+may be accessed at :data:`sys.path`.
+
+The first entry in the module search path is the directory that contains the
+input script, if there is one. Otherwise, the first entry is the current
+directory, which is the case when executing the interactive shell, a :option:`-c`
+command, or :option:`-m` module.
+
+The :envvar:`PYTHONPATH` environment variable is often used to add directories
+to the search path. If this environment variable is found then the contents are
+added to the module search path.
+
+.. note::
+
+   :envvar:`PYTHONPATH` will affect all installed Python versions/environments.
+   Be wary of setting this in your shell profile or global environment variables.
+   The :mod:`site` module offers more nuanced techniques as mentioned below.
+
+The next items added are the directories containing standard Python modules as
+well as any :term:`extension module`\s that these modules depend on. Extension
+modules are ``.pyd`` files on Windows and ``.so`` files on other platforms. The
+directory with the platform-independent Python modules is called ``prefix``.
+The directory with the extension modules is called ``exec_prefix``.
+
+The :envvar:`PYTHONHOME` environment variable may be used to set the ``prefix``
+and ``exec_prefix`` locations. Otherwise these directories are found by using
+the Python executable as a starting point and then looking for various 'landmark'
+files and directories. Note that any symbolic links are followed so the real
+Python executable location is used as the search starting point. The Python
+executable location is called ``home``.
+
+Once ``home`` is determined, the ``prefix`` directory is found by first looking
+for :file:`python{majorversion}{minorversion}.zip` (``python311.zip``). On Windows
+the zip archive is searched for in ``home`` and on Unix the archive is expected
+to be in :file:`lib`. Note that the expected zip archive location is added to the
+module search path even if the archive does not exist. If no archive was found,
+Python on Windows will continue the search for ``prefix`` by looking for :file:`Lib\\os.py`.
+Python on Unix will look for :file:`lib/python{majorversion}.{minorversion}/os.py`
+(``lib/python3.11/os.py``). On Windows ``prefix`` and ``exec_prefix`` are the same,
+however on other platforms :file:`lib/python{majorversion}.{minorversion}/lib-dynload`
+(``lib/python3.11/lib-dynload``) is searched for and used as an anchor for
+``exec_prefix``. On some platforms :file:`lib` may be :file:`lib64` or another value,
+see :data:`sys.platlibdir` and :envvar:`PYTHONPLATLIBDIR`.
+
+Once found, ``prefix`` and ``exec_prefix`` are available at :data:`sys.prefix` and
+:data:`sys.exec_prefix` respectively.
+
+Finally, the :mod:`site` module is processed and :file:`site-packages` directories
+are added to the module search path. A common way to customize the search path is
+to create :mod:`sitecustomize` or :mod:`usercustomize` modules as described in
+the :mod:`site` module documentation.
+
+.. note::
+
+   Certain command line options may further affect path calculations.
+   See :option:`-E`, :option:`-I`, :option:`-s` and :option:`-S` for further details.
+
+Virtual environments
+--------------------
+
+If Python is run in a virtual environment (as described at :ref:`tut-venv`)
+then ``prefix`` and ``exec_prefix`` are specific to the virtual environment.
+
+If a ``pyvenv.cfg`` file is found alongside the main executable, or in the
+directory one level above the executable, the following variations apply:
+
+* If ``home`` is an absolute path and :envvar:`PYTHONHOME` is not set, this
+  path is used instead of the path to the main executable when deducing ``prefix``
+  and ``exec_prefix``.
+
+_pth files
+----------
+
+To completely override :data:`sys.path` create a ``._pth`` file with the same
+name as the shared library or executable (``python._pth`` or ``python311._pth``).
+The shared library path is always known on Windows, however it may not be
+available on other platforms. In the ``._pth`` file specify one line for each path
+to add to :data:`sys.path`. The file based on the shared library name overrides
+the one based on the executable, which allows paths to be restricted for any
+program loading the runtime if desired.
+
+When the file exists, all registry and environment variables are ignored,
+isolated mode is enabled, and :mod:`site` is not imported unless one line in the
+file specifies ``import site``. Blank paths and lines starting with ``#`` are
+ignored. Each path may be absolute or relative to the location of the file.
+Import statements other than to ``site`` are not permitted, and arbitrary code
+cannot be specified.
+
+Note that ``.pth`` files (without leading underscore) will be processed normally
+by the :mod:`site` module when ``import site`` has been specified.
+
+Embedded Python
+---------------
+
+If Python is embedded within another application :c:func:`Py_InitializeFromConfig` and
+the :c:type:`PyConfig` structure can be used to initialize Python. The path specific
+details are described at :ref:`init-path-config`. Alternatively the older :c:func:`Py_SetPath`
+can be used to bypass the initialization of the module search path.
+
+.. seealso::
+
+   * :ref:`windows_finding_modules` for detailed Windows notes.
+   * :ref:`using-on-unix` for Unix details.

Doc/tutorial/modules.rst

@@ -194,6 +194,8 @@ named :file:`spam.py` in a list of directories given by the variable
 * The installation-dependent default (by convention including a
   ``site-packages`` directory, handled by the :mod:`site` module).
 
+More details are at :ref:`sys-path-init`.
+
 .. note::
    On file systems which support symlinks, the directory containing the input
    script is calculated after the symlink is followed. In other words the

Doc/using/windows.rst

@@ -946,32 +946,13 @@ target Python.
 
 
 
-.. _finding_modules:
+.. _windows_finding_modules:
 
 Finding modules
 ===============
 
-Python usually stores its library (and thereby your site-packages folder) in the
-installation directory.  So, if you had installed Python to
-:file:`C:\\Python\\`, the default library would reside in
-:file:`C:\\Python\\Lib\\` and third-party modules should be stored in
-:file:`C:\\Python\\Lib\\site-packages\\`.
-
-To completely override :data:`sys.path`, create a ``._pth`` file with the same
-name as the DLL (``python37._pth``) or the executable (``python._pth``) and
-specify one line for each path to add to :data:`sys.path`. The file based on the
-DLL name overrides the one based on the executable, which allows paths to be
-restricted for any program loading the runtime if desired.
-
-When the file exists, all registry and environment variables are ignored,
-isolated mode is enabled, and :mod:`site` is not imported unless one line in the
-file specifies ``import site``. Blank paths and lines starting with ``#`` are
-ignored. Each path may be absolute or relative to the location of the file.
-Import statements other than to ``site`` are not permitted, and arbitrary code
-cannot be specified.
-
-Note that ``.pth`` files (without leading underscore) will be processed normally
-by the :mod:`site` module when ``import site`` has been specified.
+These notes supplement the description at :ref:`sys-path-init` with
+detailed Windows notes.
 
 When no ``._pth`` file is found, this is how :data:`sys.path` is populated on
 Windows:

Doc/whatsnew/3.6.rst

@@ -162,10 +162,10 @@ Windows improvements:
 
 * A ``._pth`` file can be added to force isolated mode and fully specify
   all search paths to avoid registry and environment lookup. See
-  :ref:`the documentation <finding_modules>` for more information.
+  :ref:`the documentation <windows_finding_modules>` for more information.
 
 * A ``python36.zip`` file now works as a landmark to infer
-  :envvar:`PYTHONHOME`. See :ref:`the documentation <finding_modules>` for
+  :envvar:`PYTHONHOME`. See :ref:`the documentation <windows_finding_modules>` for
   more information.
 
 

Doc/whatsnew/3.7.rst

@@ -2474,7 +2474,7 @@ Windows-only Changes
 
 The file used to override :data:`sys.path` is now called
 ``<python-executable>._pth`` instead of ``'sys.path'``.
-See :ref:`finding_modules` for more information.
+See :ref:`windows_finding_modules` for more information.
 (Contributed by Steve Dower in :issue:`28137`.)
 
 

Misc/ACKS

@@ -1890,6 +1890,7 @@ Colin Watson
 David Watson
 Aaron Watters
 Alex Waygood
+Russel Webber
 Henrik Weber
 Leon Weber
 Steve Weber