Initialization, Finalization, and Threads¶

Before Python Initialization¶

In an application embedding Python, the Py_Initialize() function must be called before using any other Python/C API functions; with the exception of a few functions and the global configuration variables.

The following functions can be safely called before Python is initialized:

Note

The following functions should not be called before Py_Initialize(): Py_EncodeLocale(), Py_GetPath(), Py_GetPrefix(), Py_GetExecPrefix(), Py_GetProgramFullPath(), Py_GetPythonHome(), Py_GetProgramName() and PyEval_InitThreads().

Global configuration variables¶

Python has variables for the global configuration to control different features and options. By default, these flags are controlled by command line options.

When a flag is set by an option, the value of the flag is the number of times that the option was set. For example, -b sets Py_BytesWarningFlag to 1 and -bb sets Py_BytesWarningFlag to 2.

int Py_BytesWarningFlag¶

Issue a warning when comparing bytes or bytearray with str or bytes with int. Issue an error if greater or equal to 2.

Set by the -b option.

int Py_DebugFlag¶

Turn on parser debugging output (for expert only, depending on compilation options).

Set by the -d option and the PYTHONDEBUG environment variable.

int Py_DontWriteBytecodeFlag¶

If set to non-zero, Python won’t try to write .pyc files on the import of source modules.

Set by the -B option and the PYTHONDONTWRITEBYTECODE environment variable.

int Py_FrozenFlag¶

Suppress error messages when calculating the module search path in Py_GetPath().

Private flag used by _freeze_importlib and frozenmain programs.

int Py_HashRandomizationFlag¶

Set to 1 if the PYTHONHASHSEED environment variable is set to a non-empty string.

If the flag is non-zero, read the PYTHONHASHSEED environment variable to initialize the secret hash seed.

int Py_IgnoreEnvironmentFlag¶

Ignore all PYTHON* environment variables, e.g. PYTHONPATH and PYTHONHOME, that might be set.

Set by the -E and -I options.

int Py_InspectFlag¶

When a script is passed as first argument or the -c option is used, enter interactive mode after executing the script or the command, even when sys.stdin does not appear to be a terminal.

Set by the -i option and the PYTHONINSPECT environment variable.

int Py_InteractiveFlag¶: Set by the -i option.

int Py_IsolatedFlag¶

Run Python in isolated mode. In isolated mode sys.path contains neither the script’s directory nor the user’s site-packages directory.

Set by the -I option.

New in version 3.4.

int Py_LegacyWindowsFSEncodingFlag¶

If the flag is non-zero, use the mbcs encoding instead of the UTF-8 encoding for the filesystem encoding.

Set to 1 if the PYTHONLEGACYWINDOWSFSENCODING environment variable is set to a non-empty string.

See PEP 529 for more details.

Availability: Windows.

int Py_LegacyWindowsStdioFlag¶

If the flag is non-zero, use io.FileIO instead of WindowsConsoleIO for sys standard streams.

Set to 1 if the PYTHONLEGACYWINDOWSSTDIO environment variable is set to a non-empty string.

See PEP 528 for more details.

Availability: Windows.

int Py_NoSiteFlag¶

Disable the import of the module site and the site-dependent manipulations of sys.path that it entails. Also disable these manipulations if site is explicitly imported later (call site.main() if you want them to be triggered).

Set by the -S option.

int Py_NoUserSiteDirectory¶

Don’t add the user site-packages directory to sys.path.

Set by the -s and -I options, and the PYTHONNOUSERSITE environment variable.

int Py_OptimizeFlag¶: Set by the -O option and the PYTHONOPTIMIZE environment variable.

int Py_QuietFlag¶

Don’t display the copyright and version messages even in interactive mode.

Set by the -q option.

New in version 3.2.

int Py_UnbufferedStdioFlag¶

Force the stdout and stderr streams to be unbuffered.

Set by the -u option and the PYTHONUNBUFFERED environment variable.

int Py_VerboseFlag¶

Print a message each time a module is initialized, showing the place (filename or built-in module) from which it is loaded. If greater or equal to 2, print a message for each file that is checked for when searching for a module. Also provides information on module cleanup at exit.

Set by the -v option and the PYTHONVERBOSE environment variable.

Initializing and finalizing the interpreter¶

void Py_Initialize()¶

Initialize the Python interpreter. In an application embedding Python, this should be called before using any other Python/C API functions; see Before Python Initialization for the few exceptions.

This initializes the table of loaded modules (sys.modules), and creates the fundamental modules builtins, __main__ and sys. It also initializes the module search path (sys.path). It does not set sys.argv; use PySys_SetArgvEx() for that. This is a no-op when called for a second time (without calling Py_FinalizeEx() first). There is no return value; it is a fatal error if the initialization fails.

Note

On Windows, changes the console mode from O_TEXT to O_BINARY, which will also affect non-Python uses of the console using the C Runtime.

void Py_InitializeEx(int initsigs)¶: This function works like Py_Initialize() if initsigs is 1. If initsigs is 0, it skips initialization registration of signal handlers, which might be useful when Python is embedded.

int Py_IsInitialized()¶: Return true (nonzero) when the Python interpreter has been initialized, false (zero) if not. After Py_FinalizeEx() is called, this returns false until Py_Initialize() is called again.

int Py_FinalizeEx()¶

Undo all initializations made by Py_Initialize() and subsequent use of Python/C API functions, and destroy all sub-interpreters (see Py_NewInterpreter() below) that were created and not yet destroyed since the last call to Py_Initialize(). Ideally, this frees all memory allocated by the Python interpreter. This is a no-op when called for a second time (without calling Py_Initialize() again first). Normally the return value is 0. If there were errors during finalization (flushing buffered data), -1 is returned.

This function is provided for a number of reasons. An embedding application might want to restart Python without having to restart the application itself. An application that has loaded the Python interpreter from a dynamically loadable library (or DLL) might want to free all memory allocated by Python before unloading the DLL. During a hunt for memory leaks in an application a developer might want to free all memory allocated by Python before exiting from the application.

Bugs and caveats: The destruction of modules and objects in modules is done in random order; this may cause destructors (__del__() methods) to fail when they depend on other objects (even functions) or modules. Dynamically loaded extension modules loaded by Python are not unloaded. Small amounts of memory allocated by the Python interpreter may not be freed (if you find a leak, please report it). Memory tied up in circular references between objects is not freed. Some memory allocated by extension modules may not be freed. Some extensions may not work properly if their initialization routine is called more than once; this can happen if an application calls Py_Initialize() and Py_FinalizeEx() more than once.

Raises an auditing event cpython._PySys_ClearAuditHooks with no arguments.

New in version 3.6.

void Py_Finalize()¶: This is a backwards-compatible version of Py_FinalizeEx() that disregards the return value.

Process-wide parameters¶

int Py_SetStandardStreamEncoding(const char *encoding, const char *errors)¶

This function should be called before Py_Initialize(), if it is called at all. It specifies which encoding and error handling to use with standard IO, with the same meanings as in str.encode().

It overrides PYTHONIOENCODING values, and allows embedding code to control IO encoding when the environment variable does not work.

encoding and/or errors may be NULL to use PYTHONIOENCODING and/or default values (depending on other settings).

Note that sys.stderr always uses the “backslashreplace” error handler, regardless of this (or any other) setting.

If Py_FinalizeEx() is called, this function will need to be called again in order to affect subsequent calls to Py_Initialize().

Returns 0 if successful, a nonzero value on error (e.g. calling after the interpreter has already been initialized).

New in version 3.4.

void Py_SetProgramName(const wchar_t *name)¶

This function should be called before Py_Initialize() is called for the first time, if it is called at all. It tells the interpreter the value of the argv[0] argument to the main() function of the program (converted to wide characters). This is used by Py_GetPath() and some other functions below to find the Python run-time libraries relative to the interpreter executable. The default value is 'python'. The argument should point to a zero-terminated wide character string in static storage whose contents will not change for the duration of the program’s execution. No code in the Python interpreter will change the contents of this storage.

Use Py_DecodeLocale() to decode a bytes string to get a wchar_* string.

wchar* Py_GetProgramName()¶: Return the program name set with Py_SetProgramName(), or the default. The returned string points into static storage; the caller should not modify its value.

wchar_t* Py_GetPrefix()¶: Return the prefix for installed platform-independent files. This is derived through a number of complicated rules from the program name set with Py_SetProgramName() and some environment variables; for example, if the program name is '/usr/local/bin/python', the prefix is '/usr/local'. The returned string points into static storage; the caller should not modify its value. This corresponds to the prefix variable in the top-level Makefile and the --prefix argument to the configure script at build time. The value is available to Python code as sys.prefix. It is only useful on Unix. See also the next function.

wchar_t* Py_GetExecPrefix()¶

Return the exec-prefix for installed platform-dependent files. This is derived through a number of complicated rules from the program name set with Py_SetProgramName() and some environment variables; for example, if the program name is '/usr/local/bin/python', the exec-prefix is '/usr/local'. The returned string points into static storage; the caller should not modify its value. This corresponds to the exec_prefix variable in the top-level Makefile and the --exec-prefix argument to the configure script at build time. The value is available to Python code as sys.exec_prefix. It is only useful on Unix.

Background: The exec-prefix differs from the prefix when platform dependent files (such as executables and shared libraries) are installed in a different directory tree. In a typical installation, platform dependent files may be installed in the /usr/local/plat subtree while platform independent may be installed in /usr/local.

Generally speaking, a platform is a combination of hardware and software families, e.g. Sparc machines running the Solaris 2.x operating system are considered the same platform, but Intel machines running Solaris 2.x are another platform, and Intel machines running Linux are yet another platform. Different major revisions of the same operating system generally also form different platforms. Non-Unix operating systems are a different story; the installation strategies on those systems are so different that the prefix and exec-prefix are meaningless, and set to the empty string. Note that compiled Python bytecode files are platform independent (but not independent from the Python version by which they were compiled!).

System administrators will know how to configure the mount or automount programs to share /usr/local between platforms while having /usr/local/plat be a different filesystem for each platform.

wchar_t* Py_GetProgramFullPath()¶: Return the full program name of the Python executable; this is computed as a side-effect of deriving the default module search path from the program name (set by Py_SetProgramName() above). The returned string points into static storage; the caller should not modify its value. The value is available to Python code as sys.executable.

wchar_t* Py_GetPath()¶: Return the default module search path; this is computed from the program name (set by Py_SetProgramName() above) and some environment variables. The returned string consists of a series of directory names separated by a platform dependent delimiter character. The delimiter character is ':' on Unix and Mac OS X, ';' on Windows. The returned string points into static storage; the caller should not modify its value. The list sys.path is initialized with this value on interpreter startup; it can be (and usually is) modified later to change the search path for loading modules.

void Py_SetPath(const wchar_t *)¶

Set the default module search path. If this function is called before Py_Initialize(), then Py_GetPath() won’t attempt to compute a default search path but uses the one provided instead. This is useful if Python is embedded by an application that has full knowledge of the location of all modules. The path components should be separated by the platform dependent delimiter character, which is ':' on Unix and Mac OS X, ';' on Windows.

This also causes sys.executable to be set to the program full path (see Py_GetProgramFullPath()) and for sys.prefix and sys.exec_prefix to be empty. It is up to the caller to modify these if required after calling Py_Initialize().

Use Py_DecodeLocale() to decode a bytes string to get a wchar_* string.

The path argument is copied internally, so the caller may free it after the call completes.

Changed in version 3.8: The program full path is now used for sys.executable, instead of the program name.

const char* Py_GetVersion()¶

Return the version of this Python interpreter. This is a string that looks something like

"3.0a5+ (py3k:63103M, May 12 2008, 00:53:55) \n[GCC 4.2.3]"

The first word (up to the first space character) is the current Python version; the first three characters are the major and minor version separated by a period. The returned string points into static storage; the caller should not modify its value. The value is available to Python code as sys.version.

const char* Py_GetPlatform()¶: Return the platform identifier for the current platform. On Unix, this is formed from the “official” name of the operating system, converted to lower case, followed by the major revision number; e.g., for Solaris 2.x, which is also known as SunOS 5.x, the value is 'sunos5'. On Mac OS X, it is 'darwin'. On Windows, it is 'win'. The returned string points into static storage; the caller should not modify its value. The value is available to Python code as sys.platform.

const char* Py_GetCopyright()¶

Return the official copyright string for the current Python version, for example

The returned string points into static storage; the caller should not modify its value. The value is available to Python code as sys.copyright.

const char* Py_GetCompiler()¶

Return an indication of the compiler used to build the current Python version, in square brackets, for example:

"[GCC 2.7.2.2]"

The returned string points into static storage; the caller should not modify its value. The value is available to Python code as part of the variable sys.version.

const char* Py_GetBuildInfo()¶

Return information about the sequence number and build date and time of the current Python interpreter instance, for example

"#67, Aug  1 1997, 22:34:28"

The returned string points into static storage; the caller should not modify its value. The value is available to Python code as part of the variable sys.version.

void PySys_SetArgvEx(int argc, wchar_t **argv, int updatepath)¶

Set sys.argv based on argc and argv. These parameters are similar to those passed to the program’s main() function with the difference that the first entry should refer to the script file to be executed rather than the executable hosting the Python interpreter. If there isn’t a script that will be run, the first entry in argv can be an empty string. If this function fails to initialize sys.argv, a fatal condition is signalled using Py_FatalError().

If updatepath is zero, this is all the function does. If updatepath is non-zero, the function also modifies sys.path according to the following algorithm:

If the name of an existing script is passed in argv[0], the absolute path of the directory where the script is located is prepended to sys.path.
Otherwise (that is, if argc is 0 or argv[0] doesn’t point to an existing file name), an empty string is prepended to sys.path, which is the same as prepending the current working directory (".").

Use Py_DecodeLocale() to decode a bytes string to get a wchar_* string.

Note

It is recommended that applications embedding the Python interpreter for purposes other than executing a single script pass 0 as updatepath, and update sys.path themselves if desired. See CVE-2008-5983.

On versions before 3.1.3, you can achieve the same effect by manually popping the first sys.path element after having called PySys_SetArgv(), for example using:

PyRun_SimpleString("import sys; sys.path.pop(0)\n");

New in version 3.1.3.

void PySys_SetArgv(int argc, wchar_t **argv)¶

This function works like PySys_SetArgvEx() with updatepath set to 1 unless the python interpreter was started with the -I.

Use Py_DecodeLocale() to decode a bytes string to get a wchar_* string.

Changed in version 3.4: The updatepath value depends on -I.

void Py_SetPythonHome(const wchar_t *home)¶

Set the default “home” directory, that is, the location of the standard Python libraries. See PYTHONHOME for the meaning of the argument string.

The argument should point to a zero-terminated character string in static storage whose contents will not change for the duration of the program’s execution. No code in the Python interpreter will change the contents of this storage.

Use Py_DecodeLocale() to decode a bytes string to get a wchar_* string.

w_char* Py_GetPythonHome()¶: Return the default “home”, that is, the value set by a previous call to Py_SetPythonHome(), or the value of the PYTHONHOME environment variable if it is set.

Thread State and the Global Interpreter Lock¶

The Python interpreter is not fully thread-safe. In order to support multi-threaded Python programs, there’s a global lock, called the global interpreter lock or GIL, that must be held by the current thread before it can safely access Python objects. Without the lock, even the simplest operations could cause problems in a multi-threaded program: for example, when two threads simultaneously increment the reference count of the same object, the reference count could end up being incremented only once instead of twice.

Therefore, the rule exists that only the thread that has acquired the GIL may operate on Python objects or call Python/C API functions. In order to emulate concurrency of execution, the interpreter regularly tries to switch threads (see sys.setswitchinterval()). The lock is also released around potentially blocking I/O operations like reading or writing a file, so that other Python threads can run in the meantime.

The Python interpreter keeps some thread-specific bookkeeping information inside a data structure called PyThreadState. There’s also one global variable pointing to the current PyThreadState: it can be retrieved using PyThreadState_Get().

Releasing the GIL from extension code¶

Most extension code manipulating the GIL has the following simple structure:

Save the thread state in a local variable.
Release the global interpreter lock.
... Do some blocking I/O operation ...
Reacquire the global interpreter lock.
Restore the thread state from the local variable.

This is so common that a pair of macros exists to simplify it:

Py_BEGIN_ALLOW_THREADS
... Do some blocking I/O operation ...
Py_END_ALLOW_THREADS

The Py_BEGIN_ALLOW_THREADS macro opens a new block and declares a hidden local variable; the Py_END_ALLOW_THREADS macro closes the block.

The block above expands to the following code:

PyThreadState *_save;

_save = PyEval_SaveThread();
... Do some blocking I/O operation ...
PyEval_RestoreThread(_save);

Here is how these functions work: the global interpreter lock is used to protect the pointer to the current thread state. When releasing the lock and saving the thread state, the current thread state pointer must be retrieved before the lock is released (since another thread could immediately acquire the lock and store its own thread state in the global variable). Conversely, when acquiring the lock and restoring the thread state, the lock must be acquired before storing the thread state pointer.

Note

Calling system I/O functions is the most common use case for releasing the GIL, but it can also be useful before calling long-running computations which don’t need access to Python objects, such as compression or cryptographic functions operating over memory buffers. For example, the standard zlib and hashlib modules release the GIL when compressing or hashing data.

Non-Python created threads¶

When threads are created using the dedicated Python APIs (such as the threading module), a thread state is automatically associated to them and the code showed above is therefore correct. However, when threads are created from C (for example by a third-party library with its own thread management), they don’t hold the GIL, nor is there a thread state structure for them.

If you need to call Python code from these threads (often this will be part of a callback API provided by the aforementioned third-party library), you must first register these threads with the interpreter by creating a thread state data structure, then acquiring the GIL, and finally storing their thread state pointer, before you can start using the Python/C API. When you are done, you should reset the thread state pointer, release the GIL, and finally free the thread state data structure.

The PyGILState_Ensure() and PyGILState_Release() functions do all of the above automatically. The typical idiom for calling into Python from a C thread is:

PyGILState_STATE gstate;
gstate = PyGILState_Ensure();

/* Perform Python actions here. */
result = CallSomeFunction();
/* evaluate result or handle exception */

/* Release the thread. No Python API allowed beyond this point. */
PyGILState_Release(gstate);

Note that the PyGILState_*() functions assume there is only one global interpreter (created automatically by Py_Initialize()). Python supports the creation of additional interpreters (using Py_NewInterpreter()), but mixing multiple interpreters and the PyGILState_*() API is unsupported.

Cautions about fork()¶

Another important thing to note about threads is their behaviour in the face of the C fork() call. On most systems with fork(), after a process forks only the thread that issued the fork will exist. This has a concrete impact both on how locks must be handled and on all stored state in CPython’s runtime.

The fact that only the “current” thread remains means any locks held by other threads will never be released. Python solves this for os.fork() by acquiring the locks it uses internally before the fork, and releasing them afterwards. In addition, it resets any Lock Objects in the child. When extending or embedding Python, there is no way to inform Python of additional (non-Python) locks that need to be acquired before or reset after a fork. OS facilities such as pthread_atfork() would need to be used to accomplish the same thing. Additionally, when extending or embedding Python, calling fork() directly rather than through os.fork() (and returning to or calling into Python) may result in a deadlock by one of Python’s internal locks being held by a thread that is defunct after the fork. PyOS_AfterFork_Child() tries to reset the necessary locks, but is not always able to.

The fact that all other threads go away also means that CPython’s runtime state there must be cleaned up properly, which os.fork() does. This means finalizing all other PyThreadState objects belonging to the current interpreter and all other PyInterpreterState objects. Due to this and the special nature of the “main” interpreter, fork() should only be called in that interpreter’s “main” thread, where the CPython global runtime was originally initialized. The only exception is if exec() will be called immediately after.

High-level API¶

These are the most commonly used types and functions when writing C extension code, or when embedding the Python interpreter:

PyInterpreterState¶

This data structure represents the state shared by a number of cooperating threads. Threads belonging to the same interpreter share their module administration and a few other internal items. There are no public members in this structure.

Threads belonging to different interpreters initially share nothing, except process state like available memory, open file descriptors and such. The global interpreter lock is also shared by all threads, regardless of to which interpreter they belong.

PyThreadState¶: This data structure represents the state of a single thread. The only public data member is interp (PyInterpreterState *), which points to this thread’s interpreter state.

void PyEval_InitThreads()¶

Deprecated function which does nothing.

In Python 3.6 and older, this function created the GIL if it didn’t exist.

Changed in version 3.9: The function now does nothing.

Changed in version 3.7: This function is now called by Py_Initialize(), so you don’t have to call it yourself anymore.

Changed in version 3.2: This function cannot be called before Py_Initialize() anymore.

Deprecated since version 3.9, will be removed in version 3.11.

int PyEval_ThreadsInitialized()¶: Returns a non-zero value if PyEval_InitThreads() has been called. This function can be called without holding the GIL, and therefore can be used to avoid calls to the locking API when running single-threaded.

Changed in version 3.7: The GIL is now initialized by Py_Initialize().

Deprecated since version 3.9, will be removed in version 3.11.

PyThreadState* PyEval_SaveThread()¶: Release the global interpreter lock (if it has been created) and reset the thread state to NULL, returning the previous thread state (which is not NULL). If the lock has been created, the current thread must have acquired it.

void PyEval_RestoreThread(PyThreadState *tstate)¶: Acquire the global interpreter lock (if it has been created) and set the thread state to tstate, which must not be NULL. If the lock has been created, the current thread must not have acquired it, otherwise deadlock ensues.

Note

Calling this function from a thread when the runtime is finalizing will terminate the thread, even if the thread was not created by Python. You can use _Py_IsFinalizing() or sys.is_finalizing() to check if the interpreter is in process of being finalized before calling this function to avoid unwanted termination.

PyThreadState* PyThreadState_Get()¶: Return the current thread state. The global interpreter lock must be held. When the current thread state is NULL, this issues a fatal error (so that the caller needn’t check for NULL).

PyThreadState* PyThreadState_Swap(PyThreadState *tstate)¶: Swap the current thread state with the thread state given by the argument tstate, which may be NULL. The global interpreter lock must be held and is not released.

The following functions use thread-local storage, and are not compatible with sub-interpreters:

PyGILState_STATE PyGILState_Ensure()¶

Ensure that the current thread is ready to call the Python C API regardless of the current state of Python, or of the global interpreter lock. This may be called as many times as desired by a thread as long as each call is matched with a call to PyGILState_Release(). In general, other thread-related APIs may be used between PyGILState_Ensure() and PyGILState_Release() calls as long as the thread state is restored to its previous state before the Release(). For example, normal usage of the Py_BEGIN_ALLOW_THREADS and Py_END_ALLOW_THREADS macros is acceptable.

The return value is an opaque “handle” to the thread state when PyGILState_Ensure() was called, and must be passed to PyGILState_Release() to ensure Python is left in the same state. Even though recursive calls are allowed, these handles cannot be shared - each unique call to PyGILState_Ensure() must save the handle for its call to PyGILState_Release().

When the function returns, the current thread will hold the GIL and be able to call arbitrary Python code. Failure is a fatal error.

Note

Calling this function from a thread when the runtime is finalizing will terminate the thread, even if the thread was not created by Python. You can use _Py_IsFinalizing() or sys.is_finalizing() to check if the interpreter is in process of being finalized before calling this function to avoid unwanted termination.

void PyGILState_Release(PyGILState_STATE)¶

Release any resources previously acquired. After this call, Python’s state will be the same as it was prior to the corresponding PyGILState_Ensure() call (but generally this state will be unknown to the caller, hence the use of the GILState API).

Every call to PyGILState_Ensure() must be matched by a call to PyGILState_Release() on the same thread.

PyThreadState* PyGILState_GetThisThreadState()¶: Get the current thread state for this thread. May return NULL if no GILState API has been used on the current thread. Note that the main thread always has such a thread-state, even if no auto-thread-state call has been made on the main thread. This is mainly a helper/diagnostic function.

int PyGILState_Check()¶: Return 1 if the current thread is holding the GIL and 0 otherwise. This function can be called from any thread at any time. Only if it has had its Python thread state initialized and currently is holding the GIL will it return 1. This is mainly a helper/diagnostic function. It can be useful for example in callback contexts or memory allocation functions when knowing that the GIL is locked can allow the caller to perform sensitive actions or otherwise behave differently.

New in version 3.4.

The following macros are normally used without a trailing semicolon; look for example usage in the Python source distribution.

Py_BEGIN_ALLOW_THREADS¶: This macro expands to { PyThreadState *_save; _save = PyEval_SaveThread();. Note that it contains an opening brace; it must be matched with a following Py_END_ALLOW_THREADS macro. See above for further discussion of this macro.

Py_END_ALLOW_THREADS¶: This macro expands to PyEval_RestoreThread(_save); }. Note that it contains a closing brace; it must be matched with an earlier Py_BEGIN_ALLOW_THREADS macro. See above for further discussion of this macro.

Py_BLOCK_THREADS¶: This macro expands to PyEval_RestoreThread(_save);: it is equivalent to Py_END_ALLOW_THREADS without the closing brace.

Py_UNBLOCK_THREADS¶: This macro expands to _save = PyEval_SaveThread();: it is equivalent to Py_BEGIN_ALLOW_THREADS without the opening brace and variable declaration.

Low-level API¶

All of the following functions must be called after Py_Initialize().

Changed in version 3.7: Py_Initialize() now initializes the GIL.

PyInterpreterState* PyInterpreterState_New()¶

Create a new interpreter state object. The global interpreter lock need not be held, but may be held if it is necessary to serialize calls to this function.

Raises an auditing event cpython.PyInterpreterState_New with no arguments.

void PyInterpreterState_Clear(PyInterpreterState *interp)¶

Reset all information in an interpreter state object. The global interpreter lock must be held.

Raises an auditing event cpython.PyInterpreterState_Clear with no arguments.

void PyInterpreterState_Delete(PyInterpreterState *interp)¶: Destroy an interpreter state object. The global interpreter lock need not be held. The interpreter state must have been reset with a previous call to PyInterpreterState_Clear().

PyThreadState* PyThreadState_New(PyInterpreterState *interp)¶: Create a new thread state object belonging to the given interpreter object. The global interpreter lock need not be held, but may be held if it is necessary to serialize calls to this function.

void PyThreadState_Clear(PyThreadState *tstate)¶: Reset all information in a thread state object. The global interpreter lock must be held.

Changed in version 3.9: This function now calls the PyThreadState.on_delete callback. Previously, that happened in PyThreadState_Delete().

void PyThreadState_Delete(PyThreadState *tstate)¶: Destroy a thread state object. The global interpreter lock need not be held. The thread state must have been reset with a previous call to PyThreadState_Clear().

void PyThreadState_DeleteCurrent(void)¶: Destroy the current thread state and release the global interpreter lock. Like PyThreadState_Delete(), the global interpreter lock need not be held. The thread state must have been reset with a previous call to PyThreadState_Clear().

PyFrameObject* PyThreadState_GetFrame(PyThreadState *tstate)¶

Get the current frame of the Python thread state tstate.

Return a strong reference. Return NULL if no frame is currently executing.

Sub-interpreter support¶

While in most uses, you will only embed a single Python interpreter, there are cases where you need to create several independent interpreters in the same process and perhaps even in the same thread. Sub-interpreters allow you to do that.

The “main” interpreter is the first one created when the runtime initializes. It is usually the only Python interpreter in a process. Unlike sub-interpreters, the main interpreter has unique process-global responsibilities like signal handling. It is also responsible for execution during runtime initialization and is usually the active interpreter during runtime finalization. The PyInterpreterState_Main() function returns a pointer to its state.

You can switch between sub-interpreters using the PyThreadState_Swap() function. You can create and destroy them using the following functions:

PyThreadState* Py_NewInterpreter()¶

Create a new sub-interpreter. This is an (almost) totally separate environment for the execution of Python code. In particular, the new interpreter has separate, independent versions of all imported modules, including the fundamental modules builtins, __main__ and sys. The table of loaded modules (sys.modules) and the module search path (sys.path) are also separate. The new environment has no sys.argv variable. It has new standard I/O stream file objects sys.stdin, sys.stdout and sys.stderr (however these refer to the same underlying file descriptors).

The return value points to the first thread state created in the new sub-interpreter. This thread state is made in the current thread state. Note that no actual thread is created; see the discussion of thread states below. If creation of the new interpreter is unsuccessful, NULL is returned; no exception is set since the exception state is stored in the current thread state and there may not be a current thread state. (Like all other Python/C API functions, the global interpreter lock must be held before calling this function and is still held when it returns; however, unlike most other Python/C API functions, there needn’t be a current thread state on entry.)

Extension modules are shared between (sub-)interpreters as follows:

For modules using multi-phase initialization, e.g. PyModule_FromDefAndSpec(), a separate module object is created and initialized for each interpreter. Only C-level static and global variables are shared between these module objects.
For modules using single-phase initialization, e.g. PyModule_Create(), the first time a particular extension is imported, it is initialized normally, and a (shallow) copy of its module’s dictionary is squirreled away. When the same extension is imported by another (sub-)interpreter, a new module is initialized and filled with the contents of this copy; the extension’s init function is not called. Objects in the module’s dictionary thus end up shared across (sub-)interpreters, which might cause unwanted behavior (see Bugs and caveats below).

Note that this is different from what happens when an extension is imported after the interpreter has been completely re-initialized by calling Py_FinalizeEx() and Py_Initialize(); in that case, the extension’s initmodule function is called again. As with multi-phase initialization, this means that only C-level static and global variables are shared between these modules.

void Py_EndInterpreter(PyThreadState *tstate)¶: Destroy the (sub-)interpreter represented by the given thread state. The given thread state must be the current thread state. See the discussion of thread states below. When the call returns, the current thread state is NULL. All thread states associated with this interpreter are destroyed. (The global interpreter lock must be held before calling this function and is still held when it returns.) Py_FinalizeEx() will destroy all sub-interpreters that haven’t been explicitly destroyed at that point.

Bugs and caveats¶

Because sub-interpreters (and the main interpreter) are part of the same process, the insulation between them isn’t perfect — for example, using low-level file operations like os.close() they can (accidentally or maliciously) affect each other’s open files. Because of the way extensions are shared between (sub-)interpreters, some extensions may not work properly; this is especially likely when using single-phase initialization or (static) global variables. It is possible to insert objects created in one sub-interpreter into a namespace of another (sub-)interpreter; this should be avoided if possible.

Special care should be taken to avoid sharing user-defined functions, methods, instances or classes between sub-interpreters, since import operations executed by such objects may affect the wrong (sub-)interpreter’s dictionary of loaded modules. It is equally important to avoid sharing objects from which the above are reachable.

Also note that combining this functionality with PyGILState_*() APIs is delicate, because these APIs assume a bijection between Python thread states and OS-level threads, an assumption broken by the presence of sub-interpreters. It is highly recommended that you don’t switch sub-interpreters between a pair of matching PyGILState_Ensure() and PyGILState_Release() calls. Furthermore, extensions (such as ctypes) using these APIs to allow calling of Python code from non-Python created threads will probably be broken when using sub-interpreters.

Asynchronous Notifications¶

A mechanism is provided to make asynchronous notifications to the main interpreter thread. These notifications take the form of a function pointer and a void pointer argument.

int Py_AddPendingCall(int (*func)(void *), void *arg)¶

Schedule a function to be called from the main interpreter thread. On success, 0 is returned and func is queued for being called in the main thread. On failure, -1 is returned without setting any exception.

When successfully queued, func will be eventually called from the main interpreter thread with the argument arg. It will be called asynchronously with respect to normally running Python code, but with both these conditions met:

on a bytecode boundary;
with the main thread holding the global interpreter lock (func can therefore use the full C API).

func must return 0 on success, or -1 on failure with an exception set. func won’t be interrupted to perform another asynchronous notification recursively, but it can still be interrupted to switch threads if the global interpreter lock is released.

This function doesn’t need a current thread state to run, and it doesn’t need the global interpreter lock.

To call this function in a subinterpreter, the caller must hold the GIL. Otherwise, the function func can be scheduled to be called from the wrong interpreter.

Warning

This is a low-level function, only useful for very special cases. There is no guarantee that func will be called as quick as possible. If the main thread is busy executing a system call, func won’t be called before the system call returns. This function is generally not suitable for calling Python code from arbitrary C threads. Instead, use the PyGILState API.

Changed in version 3.9: If this function is called in a subinterpreter, the function func is now scheduled to be called from the subinterpreter, rather than being called from the main interpreter. Each subinterpreter now has its own list of scheduled calls.

New in version 3.1.

Profiling and Tracing¶

The Python interpreter provides some low-level support for attaching profiling and execution tracing facilities. These are used for profiling, debugging, and coverage analysis tools.

This C interface allows the profiling or tracing code to avoid the overhead of calling through Python-level callable objects, making a direct C function call instead. The essential attributes of the facility have not changed; the interface allows trace functions to be installed per-thread, and the basic events reported to the trace function are the same as had been reported to the Python-level trace functions in previous versions.

int (*Py_tracefunc)(PyObject *obj, PyFrameObject *frame, int what, PyObject *arg)¶

The type of the trace function registered using PyEval_SetProfile() and PyEval_SetTrace(). The first parameter is the object passed to the registration function as obj, frame is the frame object to which the event pertains, what is one of the constants PyTrace_CALL, PyTrace_EXCEPTION, PyTrace_LINE, PyTrace_RETURN, PyTrace_C_CALL, PyTrace_C_EXCEPTION, PyTrace_C_RETURN, or PyTrace_OPCODE, and arg depends on the value of what:

Value of what	Meaning of arg
`PyTrace_CALL`	Always `Py_None`.
`PyTrace_EXCEPTION`	Exception information as returned by `sys.exc_info()`.
`PyTrace_LINE`	Always `Py_None`.
`PyTrace_RETURN`	Value being returned to the caller, or `NULL` if caused by an exception.
`PyTrace_C_CALL`	Function object being called.
`PyTrace_C_EXCEPTION`	Function object being called.
`PyTrace_C_RETURN`	Function object being called.
`PyTrace_OPCODE`	Always `Py_None`.

int PyTrace_CALL¶: The value of the what parameter to a Py_tracefunc function when a new call to a function or method is being reported, or a new entry into a generator. Note that the creation of the iterator for a generator function is not reported as there is no control transfer to the Python bytecode in the corresponding frame.

int PyTrace_EXCEPTION¶: The value of the what parameter to a Py_tracefunc function when an exception has been raised. The callback function is called with this value for what when after any bytecode is processed after which the exception becomes set within the frame being executed. The effect of this is that as exception propagation causes the Python stack to unwind, the callback is called upon return to each frame as the exception propagates. Only trace functions receives these events; they are not needed by the profiler.

int PyTrace_LINE¶: The value passed as the what parameter to a Py_tracefunc function (but not a profiling function) when a line-number event is being reported. It may be disabled for a frame by setting f_trace_lines to 0 on that frame.

int PyTrace_RETURN¶: The value for the what parameter to Py_tracefunc functions when a call is about to return.

int PyTrace_C_CALL¶: The value for the what parameter to Py_tracefunc functions when a C function is about to be called.

int PyTrace_C_EXCEPTION¶: The value for the what parameter to Py_tracefunc functions when a C function has raised an exception.

int PyTrace_C_RETURN¶: The value for the what parameter to Py_tracefunc functions when a C function has returned.

int PyTrace_OPCODE¶: The value for the what parameter to Py_tracefunc functions (but not profiling functions) when a new opcode is about to be executed. This event is not emitted by default: it must be explicitly requested by setting f_trace_opcodes to 1 on the frame.

void PyEval_SetProfile(Py_tracefunc func, PyObject *obj)¶

Set the profiler function to func. The obj parameter is passed to the function as its first parameter, and may be any Python object, or NULL. If the profile function needs to maintain state, using a different value for obj for each thread provides a convenient and thread-safe place to store it. The profile function is called for all monitored events except PyTrace_LINE PyTrace_OPCODE and PyTrace_EXCEPTION.

The caller must hold the GIL.

void PyEval_SetTrace(Py_tracefunc func, PyObject *obj)¶

Set the tracing function to func. This is similar to PyEval_SetProfile(), except the tracing function does receive line-number events and per-opcode events, but does not receive any event related to C function objects being called. Any trace function registered using PyEval_SetTrace() will not receive PyTrace_C_CALL, PyTrace_C_EXCEPTION or PyTrace_C_RETURN as a value for the what parameter.

The caller must hold the GIL.

Advanced Debugger Support¶

These functions are only intended to be used by advanced debugging tools.

PyInterpreterState* PyInterpreterState_Head()¶: Return the interpreter state object at the head of the list of all such objects.

PyInterpreterState* PyInterpreterState_Main()¶: Return the main interpreter state object.

PyInterpreterState* PyInterpreterState_Next(PyInterpreterState *interp)¶: Return the next interpreter state object after interp from the list of all such objects.

PyThreadState * PyInterpreterState_ThreadHead(PyInterpreterState *interp)¶: Return the pointer to the first PyThreadState object in the list of threads associated with the interpreter interp.

PyThreadState* PyThreadState_Next(PyThreadState *tstate)¶: Return the next thread state object after tstate from the list of all such objects belonging to the same PyInterpreterState object.

Thread Local Storage Support¶

The Python interpreter provides low-level support for thread-local storage (TLS) which wraps the underlying native TLS implementation to support the Python-level thread local storage API (threading.local). The CPython C level APIs are similar to those offered by pthreads and Windows: use a thread key and functions to associate a void* value per thread.

The GIL does not need to be held when calling these functions; they supply their own locking.

Note that Python.h does not include the declaration of the TLS APIs, you need to include pythread.h to use thread-local storage.

Note

None of these API functions handle memory management on behalf of the void* values. You need to allocate and deallocate them yourself. If the void* values happen to be PyObject*, these functions don’t do refcount operations on them either.

Thread Specific Storage (TSS) API¶

TSS API is introduced to supersede the use of the existing TLS API within the CPython interpreter. This API uses a new type Py_tss_t instead of int to represent thread keys.

New in version 3.7.

Dynamic Allocation¶

Dynamic allocation of the Py_tss_t, required in extension modules built with Py_LIMITED_API, where static allocation of this type is not possible due to its implementation being opaque at build time.

Py_tss_t* PyThread_tss_alloc()¶: Return a value which is the same state as a value initialized with Py_tss_NEEDS_INIT, or NULL in the case of dynamic allocation failure.

void PyThread_tss_free(Py_tss_t *key)¶: Free the given key allocated by PyThread_tss_alloc(), after first calling PyThread_tss_delete() to ensure any associated thread locals have been unassigned. This is a no-op if the key argument is NULL.

Note

A freed key becomes a dangling pointer, you should reset the key to NULL.

Methods¶

The parameter key of these functions must not be NULL. Moreover, the behaviors of PyThread_tss_set() and PyThread_tss_get() are undefined if the given Py_tss_t has not been initialized by PyThread_tss_create().

int PyThread_tss_is_created(Py_tss_t *key)¶: Return a non-zero value if the given Py_tss_t has been initialized by PyThread_tss_create().

int PyThread_tss_create(Py_tss_t *key)¶: Return a zero value on successful initialization of a TSS key. The behavior is undefined if the value pointed to by the key argument is not initialized by Py_tss_NEEDS_INIT. This function can be called repeatedly on the same key – calling it on an already initialized key is a no-op and immediately returns success.

void PyThread_tss_delete(Py_tss_t *key)¶: Destroy a TSS key to forget the values associated with the key across all threads, and change the key’s initialization state to uninitialized. A destroyed key is able to be initialized again by PyThread_tss_create(). This function can be called repeatedly on the same key – calling it on an already destroyed key is a no-op.

int PyThread_tss_set(Py_tss_t *key, void *value)¶: Return a zero value to indicate successfully associating a void* value with a TSS key in the current thread. Each thread has a distinct mapping of the key to a void* value.

void* PyThread_tss_get(Py_tss_t *key)¶: Return the void* value associated with a TSS key in the current thread. This returns NULL if no value is associated with the key in the current thread.

Thread Local Storage (TLS) API¶

Deprecated since version 3.7: This API is superseded by Thread Specific Storage (TSS) API.

Note

This version of the API does not support platforms where the native TLS key is defined in a way that cannot be safely cast to int. On such platforms, PyThread_create_key() will return immediately with a failure status, and the other TLS functions will all be no-ops on such platforms.

Due to the compatibility problem noted above, this version of the API should not be used in new code.

int PyThread_create_key()¶

void PyThread_delete_key(int key)¶

int PyThread_set_key_value(int key, void *value)¶

void* PyThread_get_key_value(int key)¶

void PyThread_delete_key_value(int key)¶

void PyThread_ReInitTLS()¶

Initialization, Finalization, and Threads¶

Before Python Initialization¶

Global configuration variables¶

Initializing and finalizing the interpreter¶

Process-wide parameters¶

Thread State and the Global Interpreter Lock¶

Releasing the GIL from extension code¶

Non-Python created threads¶

Cautions about fork()¶

High-level API¶

Low-level API¶

Sub-interpreter support¶

Bugs and caveats¶

Asynchronous Notifications¶

Profiling and Tracing¶

Advanced Debugger Support¶

Thread Local Storage Support¶

Thread Specific Storage (TSS) API¶

Dynamic Allocation¶

Methods¶

Thread Local Storage (TLS) API¶

Table of Contents

Previous topic

Next topic

This Page