6 *****************************************
7 Initialization, Finalization, and Threads
8 *****************************************
11 .. cfunction:: void Py_Initialize()
14 single: Py_SetProgramName()
15 single: PyEval_InitThreads()
16 single: PyEval_ReleaseLock()
17 single: PyEval_AcquireLock()
18 single: modules (in module sys)
19 single: path (in module sys)
23 triple: module; search; path
24 single: PySys_SetArgv()
27 Initialize the Python interpreter. In an application embedding Python, this
28 should be called before using any other Python/C API functions; with the
29 exception of :cfunc:`Py_SetProgramName`, :cfunc:`PyEval_InitThreads`,
30 :cfunc:`PyEval_ReleaseLock`, and :cfunc:`PyEval_AcquireLock`. This initializes
31 the table of loaded modules (``sys.modules``), and creates the fundamental
32 modules :mod:`__builtin__`, :mod:`__main__` and :mod:`sys`. It also initializes
33 the module search path (``sys.path``). It does not set ``sys.argv``; use
34 :cfunc:`PySys_SetArgv` for that. This is a no-op when called for a second time
35 (without calling :cfunc:`Py_Finalize` first). There is no return value; it is a
36 fatal error if the initialization fails.
39 .. cfunction:: void Py_InitializeEx(int initsigs)
41 This function works like :cfunc:`Py_Initialize` if *initsigs* is 1. If
42 *initsigs* is 0, it skips initialization registration of signal handlers, which
43 might be useful when Python is embedded.
48 .. cfunction:: int Py_IsInitialized()
50 Return true (nonzero) when the Python interpreter has been initialized, false
51 (zero) if not. After :cfunc:`Py_Finalize` is called, this returns false until
52 :cfunc:`Py_Initialize` is called again.
55 .. cfunction:: void Py_Finalize()
57 Undo all initializations made by :cfunc:`Py_Initialize` and subsequent use of
58 Python/C API functions, and destroy all sub-interpreters (see
59 :cfunc:`Py_NewInterpreter` below) that were created and not yet destroyed since
60 the last call to :cfunc:`Py_Initialize`. Ideally, this frees all memory
61 allocated by the Python interpreter. This is a no-op when called for a second
62 time (without calling :cfunc:`Py_Initialize` again first). There is no return
63 value; errors during finalization are ignored.
65 This function is provided for a number of reasons. An embedding application
66 might want to restart Python without having to restart the application itself.
67 An application that has loaded the Python interpreter from a dynamically
68 loadable library (or DLL) might want to free all memory allocated by Python
69 before unloading the DLL. During a hunt for memory leaks in an application a
70 developer might want to free all memory allocated by Python before exiting from
73 **Bugs and caveats:** The destruction of modules and objects in modules is done
74 in random order; this may cause destructors (:meth:`__del__` methods) to fail
75 when they depend on other objects (even functions) or modules. Dynamically
76 loaded extension modules loaded by Python are not unloaded. Small amounts of
77 memory allocated by the Python interpreter may not be freed (if you find a leak,
78 please report it). Memory tied up in circular references between objects is not
79 freed. Some memory allocated by extension modules may not be freed. Some
80 extensions may not work properly if their initialization routine is called more
81 than once; this can happen if an application calls :cfunc:`Py_Initialize` and
82 :cfunc:`Py_Finalize` more than once.
85 .. cfunction:: PyThreadState* Py_NewInterpreter()
91 single: stdout (in module sys)
92 single: stderr (in module sys)
93 single: stdin (in module sys)
95 Create a new sub-interpreter. This is an (almost) totally separate environment
96 for the execution of Python code. In particular, the new interpreter has
97 separate, independent versions of all imported modules, including the
98 fundamental modules :mod:`__builtin__`, :mod:`__main__` and :mod:`sys`. The
99 table of loaded modules (``sys.modules``) and the module search path
100 (``sys.path``) are also separate. The new environment has no ``sys.argv``
101 variable. It has new standard I/O stream file objects ``sys.stdin``,
102 ``sys.stdout`` and ``sys.stderr`` (however these refer to the same underlying
103 :ctype:`FILE` structures in the C library).
105 The return value points to the first thread state created in the new
106 sub-interpreter. This thread state is made in the current thread state.
107 Note that no actual thread is created; see the discussion of thread states
108 below. If creation of the new interpreter is unsuccessful, *NULL* is
109 returned; no exception is set since the exception state is stored in the
110 current thread state and there may not be a current thread state. (Like all
111 other Python/C API functions, the global interpreter lock must be held before
112 calling this function and is still held when it returns; however, unlike most
113 other Python/C API functions, there needn't be a current thread state on
117 single: Py_Finalize()
118 single: Py_Initialize()
120 Extension modules are shared between (sub-)interpreters as follows: the first
121 time a particular extension is imported, it is initialized normally, and a
122 (shallow) copy of its module's dictionary is squirreled away. When the same
123 extension is imported by another (sub-)interpreter, a new module is initialized
124 and filled with the contents of this copy; the extension's ``init`` function is
125 not called. Note that this is different from what happens when an extension is
126 imported after the interpreter has been completely re-initialized by calling
127 :cfunc:`Py_Finalize` and :cfunc:`Py_Initialize`; in that case, the extension's
128 ``initmodule`` function *is* called again.
130 .. index:: single: close() (in module os)
132 **Bugs and caveats:** Because sub-interpreters (and the main interpreter) are
133 part of the same process, the insulation between them isn't perfect --- for
134 example, using low-level file operations like :func:`os.close` they can
135 (accidentally or maliciously) affect each other's open files. Because of the
136 way extensions are shared between (sub-)interpreters, some extensions may not
137 work properly; this is especially likely when the extension makes use of
138 (static) global variables, or when the extension manipulates its module's
139 dictionary after its initialization. It is possible to insert objects created
140 in one sub-interpreter into a namespace of another sub-interpreter; this should
141 be done with great care to avoid sharing user-defined functions, methods,
142 instances or classes between sub-interpreters, since import operations executed
143 by such objects may affect the wrong (sub-)interpreter's dictionary of loaded
144 modules. (XXX This is a hard-to-fix bug that will be addressed in a future
147 Also note that the use of this functionality is incompatible with extension
148 modules such as PyObjC and ctypes that use the :cfunc:`PyGILState_\*` APIs (and
149 this is inherent in the way the :cfunc:`PyGILState_\*` functions work). Simple
150 things may work, but confusing behavior will always be near.
153 .. cfunction:: void Py_EndInterpreter(PyThreadState *tstate)
155 .. index:: single: Py_Finalize()
157 Destroy the (sub-)interpreter represented by the given thread state. The given
158 thread state must be the current thread state. See the discussion of thread
159 states below. When the call returns, the current thread state is *NULL*. All
160 thread states associated with this interpreter are destroyed. (The global
161 interpreter lock must be held before calling this function and is still held
162 when it returns.) :cfunc:`Py_Finalize` will destroy all sub-interpreters that
163 haven't been explicitly destroyed at that point.
166 .. cfunction:: void Py_SetProgramName(char *name)
169 single: Py_Initialize()
173 This function should be called before :cfunc:`Py_Initialize` is called for
174 the first time, if it is called at all. It tells the interpreter the value
175 of the ``argv[0]`` argument to the :cfunc:`main` function of the program.
176 This is used by :cfunc:`Py_GetPath` and some other functions below to find
177 the Python run-time libraries relative to the interpreter executable. The
178 default value is ``'python'``. The argument should point to a
179 zero-terminated character string in static storage whose contents will not
180 change for the duration of the program's execution. No code in the Python
181 interpreter will change the contents of this storage.
184 .. cfunction:: char* Py_GetProgramName()
186 .. index:: single: Py_SetProgramName()
188 Return the program name set with :cfunc:`Py_SetProgramName`, or the default.
189 The returned string points into static storage; the caller should not modify its
193 .. cfunction:: char* Py_GetPrefix()
195 Return the *prefix* for installed platform-independent files. This is derived
196 through a number of complicated rules from the program name set with
197 :cfunc:`Py_SetProgramName` and some environment variables; for example, if the
198 program name is ``'/usr/local/bin/python'``, the prefix is ``'/usr/local'``. The
199 returned string points into static storage; the caller should not modify its
200 value. This corresponds to the :makevar:`prefix` variable in the top-level
201 :file:`Makefile` and the :option:`--prefix` argument to the :program:`configure`
202 script at build time. The value is available to Python code as ``sys.prefix``.
203 It is only useful on Unix. See also the next function.
206 .. cfunction:: char* Py_GetExecPrefix()
208 Return the *exec-prefix* for installed platform-*dependent* files. This is
209 derived through a number of complicated rules from the program name set with
210 :cfunc:`Py_SetProgramName` and some environment variables; for example, if the
211 program name is ``'/usr/local/bin/python'``, the exec-prefix is
212 ``'/usr/local'``. The returned string points into static storage; the caller
213 should not modify its value. This corresponds to the :makevar:`exec_prefix`
214 variable in the top-level :file:`Makefile` and the :option:`--exec-prefix`
215 argument to the :program:`configure` script at build time. The value is
216 available to Python code as ``sys.exec_prefix``. It is only useful on Unix.
218 Background: The exec-prefix differs from the prefix when platform dependent
219 files (such as executables and shared libraries) are installed in a different
220 directory tree. In a typical installation, platform dependent files may be
221 installed in the :file:`/usr/local/plat` subtree while platform independent may
222 be installed in :file:`/usr/local`.
224 Generally speaking, a platform is a combination of hardware and software
225 families, e.g. Sparc machines running the Solaris 2.x operating system are
226 considered the same platform, but Intel machines running Solaris 2.x are another
227 platform, and Intel machines running Linux are yet another platform. Different
228 major revisions of the same operating system generally also form different
229 platforms. Non-Unix operating systems are a different story; the installation
230 strategies on those systems are so different that the prefix and exec-prefix are
231 meaningless, and set to the empty string. Note that compiled Python bytecode
232 files are platform independent (but not independent from the Python version by
233 which they were compiled!).
235 System administrators will know how to configure the :program:`mount` or
236 :program:`automount` programs to share :file:`/usr/local` between platforms
237 while having :file:`/usr/local/plat` be a different filesystem for each
241 .. cfunction:: char* Py_GetProgramFullPath()
244 single: Py_SetProgramName()
245 single: executable (in module sys)
247 Return the full program name of the Python executable; this is computed as a
248 side-effect of deriving the default module search path from the program name
249 (set by :cfunc:`Py_SetProgramName` above). The returned string points into
250 static storage; the caller should not modify its value. The value is available
251 to Python code as ``sys.executable``.
254 .. cfunction:: char* Py_GetPath()
257 triple: module; search; path
258 single: path (in module sys)
260 Return the default module search path; this is computed from the program name
261 (set by :cfunc:`Py_SetProgramName` above) and some environment variables. The
262 returned string consists of a series of directory names separated by a platform
263 dependent delimiter character. The delimiter character is ``':'`` on Unix and
264 Mac OS X, ``';'`` on Windows. The returned string points into static storage;
265 the caller should not modify its value. The value is available to Python code
266 as the list ``sys.path``, which may be modified to change the future search path
269 .. XXX should give the exact rules
272 .. cfunction:: const char* Py_GetVersion()
274 Return the version of this Python interpreter. This is a string that looks
277 "1.5 (#67, Dec 31 1997, 22:34:28) [GCC 2.7.2.2]"
279 .. index:: single: version (in module sys)
281 The first word (up to the first space character) is the current Python version;
282 the first three characters are the major and minor version separated by a
283 period. The returned string points into static storage; the caller should not
284 modify its value. The value is available to Python code as ``sys.version``.
287 .. cfunction:: const char* Py_GetBuildNumber()
289 Return a string representing the Subversion revision that this Python executable
290 was built from. This number is a string because it may contain a trailing 'M'
291 if Python was built from a mixed revision source tree.
293 .. versionadded:: 2.5
296 .. cfunction:: const char* Py_GetPlatform()
298 .. index:: single: platform (in module sys)
300 Return the platform identifier for the current platform. On Unix, this is
301 formed from the "official" name of the operating system, converted to lower
302 case, followed by the major revision number; e.g., for Solaris 2.x, which is
303 also known as SunOS 5.x, the value is ``'sunos5'``. On Mac OS X, it is
304 ``'darwin'``. On Windows, it is ``'win'``. The returned string points into
305 static storage; the caller should not modify its value. The value is available
306 to Python code as ``sys.platform``.
309 .. cfunction:: const char* Py_GetCopyright()
311 Return the official copyright string for the current Python version, for example
313 ``'Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam'``
315 .. index:: single: copyright (in module sys)
317 The returned string points into static storage; the caller should not modify its
318 value. The value is available to Python code as ``sys.copyright``.
321 .. cfunction:: const char* Py_GetCompiler()
323 Return an indication of the compiler used to build the current Python version,
324 in square brackets, for example::
328 .. index:: single: version (in module sys)
330 The returned string points into static storage; the caller should not modify its
331 value. The value is available to Python code as part of the variable
335 .. cfunction:: const char* Py_GetBuildInfo()
337 Return information about the sequence number and build date and time of the
338 current Python interpreter instance, for example ::
340 "#67, Aug 1 1997, 22:34:28"
342 .. index:: single: version (in module sys)
344 The returned string points into static storage; the caller should not modify its
345 value. The value is available to Python code as part of the variable
349 .. cfunction:: void PySys_SetArgv(int argc, char **argv)
353 single: Py_FatalError()
354 single: argv (in module sys)
356 Set :data:`sys.argv` based on *argc* and *argv*. These parameters are
357 similar to those passed to the program's :cfunc:`main` function with the
358 difference that the first entry should refer to the script file to be
359 executed rather than the executable hosting the Python interpreter. If there
360 isn't a script that will be run, the first entry in *argv* can be an empty
361 string. If this function fails to initialize :data:`sys.argv`, a fatal
362 condition is signalled using :cfunc:`Py_FatalError`.
364 This function also prepends the executed script's path to :data:`sys.path`.
365 If no script is executed (in the case of calling ``python -c`` or just the
366 interactive interpreter), the empty string is used instead.
368 .. XXX impl. doesn't seem consistent in allowing 0/NULL for the params;
372 .. cfunction:: void Py_SetPythonHome(char *home)
374 Set the default "home" directory, that is, the location of the standard
375 Python libraries. The libraries are searched in
376 :file:`{home}/lib/python{version}` and :file:`{home}/lib/python{version}`.
377 The argument should point to a zero-terminated character string in static
378 storage whose contents will not change for the duration of the program's
379 execution. No code in the Python interpreter will change the contents of
383 .. cfunction:: char* Py_GetPythonHome()
385 Return the default "home", that is, the value set by a previous call to
386 :cfunc:`Py_SetPythonHome`, or the value of the :envvar:`PYTHONHOME`
387 environment variable if it is set.
392 Thread State and the Global Interpreter Lock
393 ============================================
396 single: global interpreter lock
397 single: interpreter lock
398 single: lock, interpreter
400 The Python interpreter is not fully thread safe. In order to support
401 multi-threaded Python programs, there's a global lock, called the :dfn:`global
402 interpreter lock` or :dfn:`GIL`, that must be held by the current thread before
403 it can safely access Python objects. Without the lock, even the simplest
404 operations could cause problems in a multi-threaded program: for example, when
405 two threads simultaneously increment the reference count of the same object, the
406 reference count could end up being incremented only once instead of twice.
408 .. index:: single: setcheckinterval() (in module sys)
410 Therefore, the rule exists that only the thread that has acquired the global
411 interpreter lock may operate on Python objects or call Python/C API functions.
412 In order to support multi-threaded Python programs, the interpreter regularly
413 releases and reacquires the lock --- by default, every 100 bytecode instructions
414 (this can be changed with :func:`sys.setcheckinterval`). The lock is also
415 released and reacquired around potentially blocking I/O operations like reading
416 or writing a file, so that other threads can run while the thread that requests
417 the I/O is waiting for the I/O operation to complete.
420 single: PyThreadState
421 single: PyThreadState
423 The Python interpreter needs to keep some bookkeeping information separate per
424 thread --- for this it uses a data structure called :ctype:`PyThreadState`.
425 There's one global variable, however: the pointer to the current
426 :ctype:`PyThreadState` structure. Before the addition of :dfn:`thread-local
427 storage` (:dfn:`TLS`) the current thread state had to be manipulated
430 This is easy enough in most cases. Most code manipulating the global
431 interpreter lock has the following simple structure::
433 Save the thread state in a local variable.
434 Release the global interpreter lock.
435 ...Do some blocking I/O operation...
436 Reacquire the global interpreter lock.
437 Restore the thread state from the local variable.
439 This is so common that a pair of macros exists to simplify it::
441 Py_BEGIN_ALLOW_THREADS
442 ...Do some blocking I/O operation...
446 single: Py_BEGIN_ALLOW_THREADS
447 single: Py_END_ALLOW_THREADS
449 The :cmacro:`Py_BEGIN_ALLOW_THREADS` macro opens a new block and declares a
450 hidden local variable; the :cmacro:`Py_END_ALLOW_THREADS` macro closes the
451 block. Another advantage of using these two macros is that when Python is
452 compiled without thread support, they are defined empty, thus saving the thread
453 state and GIL manipulations.
455 When thread support is enabled, the block above expands to the following code::
457 PyThreadState *_save;
459 _save = PyEval_SaveThread();
460 ...Do some blocking I/O operation...
461 PyEval_RestoreThread(_save);
463 Using even lower level primitives, we can get roughly the same effect as
466 PyThreadState *_save;
468 _save = PyThreadState_Swap(NULL);
469 PyEval_ReleaseLock();
470 ...Do some blocking I/O operation...
471 PyEval_AcquireLock();
472 PyThreadState_Swap(_save);
475 single: PyEval_RestoreThread()
477 single: PyEval_SaveThread()
478 single: PyEval_ReleaseLock()
479 single: PyEval_AcquireLock()
481 There are some subtle differences; in particular, :cfunc:`PyEval_RestoreThread`
482 saves and restores the value of the global variable :cdata:`errno`, since the
483 lock manipulation does not guarantee that :cdata:`errno` is left alone. Also,
484 when thread support is disabled, :cfunc:`PyEval_SaveThread` and
485 :cfunc:`PyEval_RestoreThread` don't manipulate the GIL; in this case,
486 :cfunc:`PyEval_ReleaseLock` and :cfunc:`PyEval_AcquireLock` are not available.
487 This is done so that dynamically loaded extensions compiled with thread support
488 enabled can be loaded by an interpreter that was compiled with disabled thread
491 The global interpreter lock is used to protect the pointer to the current thread
492 state. When releasing the lock and saving the thread state, the current thread
493 state pointer must be retrieved before the lock is released (since another
494 thread could immediately acquire the lock and store its own thread state in the
495 global variable). Conversely, when acquiring the lock and restoring the thread
496 state, the lock must be acquired before storing the thread state pointer.
498 It is important to note that when threads are created from C, they don't have
499 the global interpreter lock, nor is there a thread state data structure for
500 them. Such threads must bootstrap themselves into existence, by first
501 creating a thread state data structure, then acquiring the lock, and finally
502 storing their thread state pointer, before they can start using the Python/C
503 API. When they are done, they should reset the thread state pointer, release
504 the lock, and finally free their thread state data structure.
506 Beginning with version 2.3, threads can now take advantage of the
507 :cfunc:`PyGILState_\*` functions to do all of the above automatically. The
508 typical idiom for calling into Python from a C thread is now::
510 PyGILState_STATE gstate;
511 gstate = PyGILState_Ensure();
513 /* Perform Python actions here. */
514 result = CallSomeFunction();
515 /* evaluate result */
517 /* Release the thread. No Python API allowed beyond this point. */
518 PyGILState_Release(gstate);
520 Note that the :cfunc:`PyGILState_\*` functions assume there is only one global
521 interpreter (created automatically by :cfunc:`Py_Initialize`). Python still
522 supports the creation of additional interpreters (using
523 :cfunc:`Py_NewInterpreter`), but mixing multiple interpreters and the
524 :cfunc:`PyGILState_\*` API is unsupported.
526 Another important thing to note about threads is their behaviour in the face
527 of the C :cfunc:`fork` call. On most systems with :cfunc:`fork`, after a
528 process forks only the thread that issued the fork will exist. That also
529 means any locks held by other threads will never be released. Python solves
530 this for :func:`os.fork` by acquiring the locks it uses internally before
531 the fork, and releasing them afterwards. In addition, it resets any
532 :ref:`lock-objects` in the child. When extending or embedding Python, there
533 is no way to inform Python of additional (non-Python) locks that need to be
534 acquired before or reset after a fork. OS facilities such as
535 :cfunc:`posix_atfork` would need to be used to accomplish the same thing.
536 Additionally, when extending or embedding Python, calling :cfunc:`fork`
537 directly rather than through :func:`os.fork` (and returning to or calling
538 into Python) may result in a deadlock by one of Python's internal locks
539 being held by a thread that is defunct after the fork.
540 :cfunc:`PyOS_AfterFork` tries to reset the necessary locks, but is not
543 .. ctype:: PyInterpreterState
545 This data structure represents the state shared by a number of cooperating
546 threads. Threads belonging to the same interpreter share their module
547 administration and a few other internal items. There are no public members in
550 Threads belonging to different interpreters initially share nothing, except
551 process state like available memory, open file descriptors and such. The global
552 interpreter lock is also shared by all threads, regardless of to which
553 interpreter they belong.
556 .. ctype:: PyThreadState
558 This data structure represents the state of a single thread. The only public
559 data member is :ctype:`PyInterpreterState \*`:attr:`interp`, which points to
560 this thread's interpreter state.
563 .. cfunction:: void PyEval_InitThreads()
566 single: PyEval_ReleaseLock()
567 single: PyEval_ReleaseThread()
568 single: PyEval_SaveThread()
569 single: PyEval_RestoreThread()
571 Initialize and acquire the global interpreter lock. It should be called in the
572 main thread before creating a second thread or engaging in any other thread
573 operations such as :cfunc:`PyEval_ReleaseLock` or
574 ``PyEval_ReleaseThread(tstate)``. It is not needed before calling
575 :cfunc:`PyEval_SaveThread` or :cfunc:`PyEval_RestoreThread`.
577 .. index:: single: Py_Initialize()
579 This is a no-op when called for a second time. It is safe to call this function
580 before calling :cfunc:`Py_Initialize`.
582 .. index:: module: thread
584 When only the main thread exists, no GIL operations are needed. This is a
585 common situation (most Python programs do not use threads), and the lock
586 operations slow the interpreter down a bit. Therefore, the lock is not
587 created initially. This situation is equivalent to having acquired the lock:
588 when there is only a single thread, all object accesses are safe. Therefore,
589 when this function initializes the global interpreter lock, it also acquires
590 it. Before the Python :mod:`thread` module creates a new thread, knowing
591 that either it has the lock or the lock hasn't been created yet, it calls
592 :cfunc:`PyEval_InitThreads`. When this call returns, it is guaranteed that
593 the lock has been created and that the calling thread has acquired it.
595 It is **not** safe to call this function when it is unknown which thread (if
596 any) currently has the global interpreter lock.
598 This function is not available when thread support is disabled at compile time.
601 .. cfunction:: int PyEval_ThreadsInitialized()
603 Returns a non-zero value if :cfunc:`PyEval_InitThreads` has been called. This
604 function can be called without holding the GIL, and therefore can be used to
605 avoid calls to the locking API when running single-threaded. This function is
606 not available when thread support is disabled at compile time.
608 .. versionadded:: 2.4
611 .. cfunction:: void PyEval_AcquireLock()
613 Acquire the global interpreter lock. The lock must have been created earlier.
614 If this thread already has the lock, a deadlock ensues. This function is not
615 available when thread support is disabled at compile time.
618 .. cfunction:: void PyEval_ReleaseLock()
620 Release the global interpreter lock. The lock must have been created earlier.
621 This function is not available when thread support is disabled at compile time.
624 .. cfunction:: void PyEval_AcquireThread(PyThreadState *tstate)
626 Acquire the global interpreter lock and set the current thread state to
627 *tstate*, which should not be *NULL*. The lock must have been created earlier.
628 If this thread already has the lock, deadlock ensues. This function is not
629 available when thread support is disabled at compile time.
632 .. cfunction:: void PyEval_ReleaseThread(PyThreadState *tstate)
634 Reset the current thread state to *NULL* and release the global interpreter
635 lock. The lock must have been created earlier and must be held by the current
636 thread. The *tstate* argument, which must not be *NULL*, is only used to check
637 that it represents the current thread state --- if it isn't, a fatal error is
638 reported. This function is not available when thread support is disabled at
642 .. cfunction:: PyThreadState* PyEval_SaveThread()
644 Release the global interpreter lock (if it has been created and thread
645 support is enabled) and reset the thread state to *NULL*, returning the
646 previous thread state (which is not *NULL*). If the lock has been created,
647 the current thread must have acquired it. (This function is available even
648 when thread support is disabled at compile time.)
651 .. cfunction:: void PyEval_RestoreThread(PyThreadState *tstate)
653 Acquire the global interpreter lock (if it has been created and thread
654 support is enabled) and set the thread state to *tstate*, which must not be
655 *NULL*. If the lock has been created, the current thread must not have
656 acquired it, otherwise deadlock ensues. (This function is available even
657 when thread support is disabled at compile time.)
660 .. cfunction:: void PyEval_ReInitThreads()
662 This function is called from :cfunc:`PyOS_AfterFork` to ensure that newly
663 created child processes don't hold locks referring to threads which
664 are not running in the child process.
667 The following macros are normally used without a trailing semicolon; look for
668 example usage in the Python source distribution.
671 .. cmacro:: Py_BEGIN_ALLOW_THREADS
673 This macro expands to ``{ PyThreadState *_save; _save = PyEval_SaveThread();``.
674 Note that it contains an opening brace; it must be matched with a following
675 :cmacro:`Py_END_ALLOW_THREADS` macro. See above for further discussion of this
676 macro. It is a no-op when thread support is disabled at compile time.
679 .. cmacro:: Py_END_ALLOW_THREADS
681 This macro expands to ``PyEval_RestoreThread(_save); }``. Note that it contains
682 a closing brace; it must be matched with an earlier
683 :cmacro:`Py_BEGIN_ALLOW_THREADS` macro. See above for further discussion of
684 this macro. It is a no-op when thread support is disabled at compile time.
687 .. cmacro:: Py_BLOCK_THREADS
689 This macro expands to ``PyEval_RestoreThread(_save);``: it is equivalent to
690 :cmacro:`Py_END_ALLOW_THREADS` without the closing brace. It is a no-op when
691 thread support is disabled at compile time.
694 .. cmacro:: Py_UNBLOCK_THREADS
696 This macro expands to ``_save = PyEval_SaveThread();``: it is equivalent to
697 :cmacro:`Py_BEGIN_ALLOW_THREADS` without the opening brace and variable
698 declaration. It is a no-op when thread support is disabled at compile time.
700 All of the following functions are only available when thread support is enabled
701 at compile time, and must be called only when the global interpreter lock has
705 .. cfunction:: PyInterpreterState* PyInterpreterState_New()
707 Create a new interpreter state object. The global interpreter lock need not
708 be held, but may be held if it is necessary to serialize calls to this
712 .. cfunction:: void PyInterpreterState_Clear(PyInterpreterState *interp)
714 Reset all information in an interpreter state object. The global interpreter
718 .. cfunction:: void PyInterpreterState_Delete(PyInterpreterState *interp)
720 Destroy an interpreter state object. The global interpreter lock need not be
721 held. The interpreter state must have been reset with a previous call to
722 :cfunc:`PyInterpreterState_Clear`.
725 .. cfunction:: PyThreadState* PyThreadState_New(PyInterpreterState *interp)
727 Create a new thread state object belonging to the given interpreter object.
728 The global interpreter lock need not be held, but may be held if it is
729 necessary to serialize calls to this function.
732 .. cfunction:: void PyThreadState_Clear(PyThreadState *tstate)
734 Reset all information in a thread state object. The global interpreter lock
738 .. cfunction:: void PyThreadState_Delete(PyThreadState *tstate)
740 Destroy a thread state object. The global interpreter lock need not be held.
741 The thread state must have been reset with a previous call to
742 :cfunc:`PyThreadState_Clear`.
745 .. cfunction:: PyThreadState* PyThreadState_Get()
747 Return the current thread state. The global interpreter lock must be held.
748 When the current thread state is *NULL*, this issues a fatal error (so that
749 the caller needn't check for *NULL*).
752 .. cfunction:: PyThreadState* PyThreadState_Swap(PyThreadState *tstate)
754 Swap the current thread state with the thread state given by the argument
755 *tstate*, which may be *NULL*. The global interpreter lock must be held.
758 .. cfunction:: PyObject* PyThreadState_GetDict()
760 Return a dictionary in which extensions can store thread-specific state
761 information. Each extension should use a unique key to use to store state in
762 the dictionary. It is okay to call this function when no current thread state
763 is available. If this function returns *NULL*, no exception has been raised and
764 the caller should assume no current thread state is available.
766 .. versionchanged:: 2.3
767 Previously this could only be called when a current thread is active, and *NULL*
768 meant that an exception was raised.
771 .. cfunction:: int PyThreadState_SetAsyncExc(long id, PyObject *exc)
773 Asynchronously raise an exception in a thread. The *id* argument is the thread
774 id of the target thread; *exc* is the exception object to be raised. This
775 function does not steal any references to *exc*. To prevent naive misuse, you
776 must write your own C extension to call this. Must be called with the GIL held.
777 Returns the number of thread states modified; this is normally one, but will be
778 zero if the thread id isn't found. If *exc* is :const:`NULL`, the pending
779 exception (if any) for the thread is cleared. This raises no exceptions.
781 .. versionadded:: 2.3
784 .. cfunction:: PyGILState_STATE PyGILState_Ensure()
786 Ensure that the current thread is ready to call the Python C API regardless
787 of the current state of Python, or of the global interpreter lock. This may
788 be called as many times as desired by a thread as long as each call is
789 matched with a call to :cfunc:`PyGILState_Release`. In general, other
790 thread-related APIs may be used between :cfunc:`PyGILState_Ensure` and
791 :cfunc:`PyGILState_Release` calls as long as the thread state is restored to
792 its previous state before the Release(). For example, normal usage of the
793 :cmacro:`Py_BEGIN_ALLOW_THREADS` and :cmacro:`Py_END_ALLOW_THREADS` macros is
796 The return value is an opaque "handle" to the thread state when
797 :cfunc:`PyGILState_Ensure` was called, and must be passed to
798 :cfunc:`PyGILState_Release` to ensure Python is left in the same state. Even
799 though recursive calls are allowed, these handles *cannot* be shared - each
800 unique call to :cfunc:`PyGILState_Ensure` must save the handle for its call
801 to :cfunc:`PyGILState_Release`.
803 When the function returns, the current thread will hold the GIL. Failure is a
806 .. versionadded:: 2.3
809 .. cfunction:: void PyGILState_Release(PyGILState_STATE)
811 Release any resources previously acquired. After this call, Python's state will
812 be the same as it was prior to the corresponding :cfunc:`PyGILState_Ensure` call
813 (but generally this state will be unknown to the caller, hence the use of the
816 Every call to :cfunc:`PyGILState_Ensure` must be matched by a call to
817 :cfunc:`PyGILState_Release` on the same thread.
819 .. versionadded:: 2.3
823 Asynchronous Notifications
824 ==========================
826 A mechanism is provided to make asynchronous notifications to the main
827 interpreter thread. These notifications take the form of a function
828 pointer and a void argument.
830 .. index:: single: setcheckinterval() (in module sys)
832 Every check interval, when the global interpreter lock is released and
833 reacquired, python will also call any such provided functions. This can be used
834 for example by asynchronous IO handlers. The notification can be scheduled from
835 a worker thread and the actual call than made at the earliest convenience by the
836 main thread where it has possession of the global interpreter lock and can
837 perform any Python API calls.
839 .. cfunction:: void Py_AddPendingCall( int (*func)(void *, void *arg) )
841 .. index:: single: Py_AddPendingCall()
843 Post a notification to the Python main thread. If successful, *func* will be
844 called with the argument *arg* at the earliest convenience. *func* will be
845 called having the global interpreter lock held and can thus use the full
846 Python API and can take any action such as setting object attributes to
847 signal IO completion. It must return 0 on success, or -1 signalling an
848 exception. The notification function won't be interrupted to perform another
849 asynchronous notification recursively, but it can still be interrupted to
850 switch threads if the global interpreter lock is released, for example, if it
851 calls back into python code.
853 This function returns 0 on success in which case the notification has been
854 scheduled. Otherwise, for example if the notification buffer is full, it
855 returns -1 without setting any exception.
857 This function can be called on any thread, be it a Python thread or some
858 other system thread. If it is a Python thread, it doesn't matter if it holds
859 the global interpreter lock or not.
861 .. versionadded:: 2.7
867 Profiling and Tracing
868 =====================
870 .. sectionauthor:: Fred L. Drake, Jr. <fdrake@acm.org>
873 The Python interpreter provides some low-level support for attaching profiling
874 and execution tracing facilities. These are used for profiling, debugging, and
875 coverage analysis tools.
877 Starting with Python 2.2, the implementation of this facility was substantially
878 revised, and an interface from C was added. This C interface allows the
879 profiling or tracing code to avoid the overhead of calling through Python-level
880 callable objects, making a direct C function call instead. The essential
881 attributes of the facility have not changed; the interface allows trace
882 functions to be installed per-thread, and the basic events reported to the trace
883 function are the same as had been reported to the Python-level trace functions
884 in previous versions.
887 .. ctype:: int (*Py_tracefunc)(PyObject *obj, PyFrameObject *frame, int what, PyObject *arg)
889 The type of the trace function registered using :cfunc:`PyEval_SetProfile` and
890 :cfunc:`PyEval_SetTrace`. The first parameter is the object passed to the
891 registration function as *obj*, *frame* is the frame object to which the event
892 pertains, *what* is one of the constants :const:`PyTrace_CALL`,
893 :const:`PyTrace_EXCEPTION`, :const:`PyTrace_LINE`, :const:`PyTrace_RETURN`,
894 :const:`PyTrace_C_CALL`, :const:`PyTrace_C_EXCEPTION`, or
895 :const:`PyTrace_C_RETURN`, and *arg* depends on the value of *what*:
897 +------------------------------+--------------------------------------+
898 | Value of *what* | Meaning of *arg* |
899 +==============================+======================================+
900 | :const:`PyTrace_CALL` | Always *NULL*. |
901 +------------------------------+--------------------------------------+
902 | :const:`PyTrace_EXCEPTION` | Exception information as returned by |
903 | | :func:`sys.exc_info`. |
904 +------------------------------+--------------------------------------+
905 | :const:`PyTrace_LINE` | Always *NULL*. |
906 +------------------------------+--------------------------------------+
907 | :const:`PyTrace_RETURN` | Value being returned to the caller. |
908 +------------------------------+--------------------------------------+
909 | :const:`PyTrace_C_CALL` | Name of function being called. |
910 +------------------------------+--------------------------------------+
911 | :const:`PyTrace_C_EXCEPTION` | Always *NULL*. |
912 +------------------------------+--------------------------------------+
913 | :const:`PyTrace_C_RETURN` | Always *NULL*. |
914 +------------------------------+--------------------------------------+
917 .. cvar:: int PyTrace_CALL
919 The value of the *what* parameter to a :ctype:`Py_tracefunc` function when a new
920 call to a function or method is being reported, or a new entry into a generator.
921 Note that the creation of the iterator for a generator function is not reported
922 as there is no control transfer to the Python bytecode in the corresponding
926 .. cvar:: int PyTrace_EXCEPTION
928 The value of the *what* parameter to a :ctype:`Py_tracefunc` function when an
929 exception has been raised. The callback function is called with this value for
930 *what* when after any bytecode is processed after which the exception becomes
931 set within the frame being executed. The effect of this is that as exception
932 propagation causes the Python stack to unwind, the callback is called upon
933 return to each frame as the exception propagates. Only trace functions receives
934 these events; they are not needed by the profiler.
937 .. cvar:: int PyTrace_LINE
939 The value passed as the *what* parameter to a trace function (but not a
940 profiling function) when a line-number event is being reported.
943 .. cvar:: int PyTrace_RETURN
945 The value for the *what* parameter to :ctype:`Py_tracefunc` functions when a
946 call is returning without propagating an exception.
949 .. cvar:: int PyTrace_C_CALL
951 The value for the *what* parameter to :ctype:`Py_tracefunc` functions when a C
952 function is about to be called.
955 .. cvar:: int PyTrace_C_EXCEPTION
957 The value for the *what* parameter to :ctype:`Py_tracefunc` functions when a C
958 function has thrown an exception.
961 .. cvar:: int PyTrace_C_RETURN
963 The value for the *what* parameter to :ctype:`Py_tracefunc` functions when a C
964 function has returned.
967 .. cfunction:: void PyEval_SetProfile(Py_tracefunc func, PyObject *obj)
969 Set the profiler function to *func*. The *obj* parameter is passed to the
970 function as its first parameter, and may be any Python object, or *NULL*. If
971 the profile function needs to maintain state, using a different value for *obj*
972 for each thread provides a convenient and thread-safe place to store it. The
973 profile function is called for all monitored events except the line-number
977 .. cfunction:: void PyEval_SetTrace(Py_tracefunc func, PyObject *obj)
979 Set the tracing function to *func*. This is similar to
980 :cfunc:`PyEval_SetProfile`, except the tracing function does receive line-number
983 .. cfunction:: PyObject* PyEval_GetCallStats(PyObject *self)
985 Return a tuple of function call counts. There are constants defined for the
986 positions within the tuple:
988 +-------------------------------+-------+
990 +===============================+=======+
991 | :const:`PCALL_ALL` | 0 |
992 +-------------------------------+-------+
993 | :const:`PCALL_FUNCTION` | 1 |
994 +-------------------------------+-------+
995 | :const:`PCALL_FAST_FUNCTION` | 2 |
996 +-------------------------------+-------+
997 | :const:`PCALL_FASTER_FUNCTION`| 3 |
998 +-------------------------------+-------+
999 | :const:`PCALL_METHOD` | 4 |
1000 +-------------------------------+-------+
1001 | :const:`PCALL_BOUND_METHOD` | 5 |
1002 +-------------------------------+-------+
1003 | :const:`PCALL_CFUNCTION` | 6 |
1004 +-------------------------------+-------+
1005 | :const:`PCALL_TYPE` | 7 |
1006 +-------------------------------+-------+
1007 | :const:`PCALL_GENERATOR` | 8 |
1008 +-------------------------------+-------+
1009 | :const:`PCALL_OTHER` | 9 |
1010 +-------------------------------+-------+
1011 | :const:`PCALL_POP` | 10 |
1012 +-------------------------------+-------+
1014 :const:`PCALL_FAST_FUNCTION` means no argument tuple needs to be created.
1015 :const:`PCALL_FASTER_FUNCTION` means that the fast-path frame setup code is used.
1017 If there is a method call where the call can be optimized by changing
1018 the argument tuple and calling the function directly, it gets recorded
1021 This function is only present if Python is compiled with :const:`CALL_PROFILE`
1024 .. _advanced-debugging:
1026 Advanced Debugger Support
1027 =========================
1029 .. sectionauthor:: Fred L. Drake, Jr. <fdrake@acm.org>
1032 These functions are only intended to be used by advanced debugging tools.
1035 .. cfunction:: PyInterpreterState* PyInterpreterState_Head()
1037 Return the interpreter state object at the head of the list of all such objects.
1039 .. versionadded:: 2.2
1042 .. cfunction:: PyInterpreterState* PyInterpreterState_Next(PyInterpreterState *interp)
1044 Return the next interpreter state object after *interp* from the list of all
1047 .. versionadded:: 2.2
1050 .. cfunction:: PyThreadState * PyInterpreterState_ThreadHead(PyInterpreterState *interp)
1052 Return the a pointer to the first :ctype:`PyThreadState` object in the list of
1053 threads associated with the interpreter *interp*.
1055 .. versionadded:: 2.2
1058 .. cfunction:: PyThreadState* PyThreadState_Next(PyThreadState *tstate)
1060 Return the next thread state object after *tstate* from the list of all such
1061 objects belonging to the same :ctype:`PyInterpreterState` object.
1063 .. versionadded:: 2.2