Initialization, Finalization, and Threads

SeePython Initialization Configuration for detailson how to configure the interpreter prior to initialization.

Before Python Initialization

In an application embedding Python, thePy_Initialize() function mustbe called before using any other Python/C API functions; with the exception ofa few functions and theglobal configuration variables.

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

Note

Despite their apparent similarity to some of the functions listed above,the following functionsshould not be called before the interpreter hasbeen initialized:Py_EncodeLocale(),Py_GetPath(),Py_GetPrefix(),Py_GetExecPrefix(),Py_GetProgramFullPath(),Py_GetPythonHome(),Py_GetProgramName(),PyEval_InitThreads(), andPy_RunMain().

Global configuration variables

Python has variables for the global configuration to control different featuresand options. By default, these flags are controlled bycommand lineoptions.

When a flag is set by an option, the value of the flag is the number of timesthat the option was set. For example,-b setsPy_BytesWarningFlagto 1 and-bb setsPy_BytesWarningFlag to 2.

intPy_BytesWarningFlag

This API is kept for backward compatibility: settingPyConfig.bytes_warning should be used instead, seePythonInitialization Configuration.

Issue a warning when comparingbytes orbytearray withstr orbytes withint. Issue an error if greateror equal to2.

Set by the-b option.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_DebugFlag

This API is kept for backward compatibility: settingPyConfig.parser_debug should be used instead, seePythonInitialization Configuration.

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

Set by the-d option and thePYTHONDEBUG environmentvariable.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_DontWriteBytecodeFlag

This API is kept for backward compatibility: settingPyConfig.write_bytecode should be used instead, seePythonInitialization Configuration.

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

Set by the-B option and thePYTHONDONTWRITEBYTECODEenvironment variable.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_FrozenFlag

This API is kept for backward compatibility: settingPyConfig.pathconfig_warnings should be used instead, seePython Initialization Configuration.

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

Private flag used by_freeze_module andfrozenmain programs.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_HashRandomizationFlag

This API is kept for backward compatibility: settingPyConfig.hash_seed andPyConfig.use_hash_seed shouldbe used instead, seePython Initialization Configuration.

Set to1 if thePYTHONHASHSEED environment variable is set toa non-empty string.

If the flag is non-zero, read thePYTHONHASHSEED environmentvariable to initialize the secret hash seed.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_IgnoreEnvironmentFlag

This API is kept for backward compatibility: settingPyConfig.use_environment should be used instead, seePython Initialization Configuration.

Ignore allPYTHON* environment variables, e.g.PYTHONPATH andPYTHONHOME, that might be set.

Set by the-E and-I options.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_InspectFlag

This API is kept for backward compatibility: settingPyConfig.inspect should be used instead, seePython Initialization Configuration.

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 whensys.stdin does not appear to be a terminal.

Set by the-i option and thePYTHONINSPECT environmentvariable.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_InteractiveFlag

This API is kept for backward compatibility: settingPyConfig.interactive should be used instead, seePython Initialization Configuration.

Set by the-i option.

Deprecated since version 3.12, will be removed in version 3.15.

intPy_IsolatedFlag

This API is kept for backward compatibility: settingPyConfig.isolated should be used instead, seePython Initialization Configuration.

Run Python in isolated mode. In isolated modesys.path containsneither the script’s directory nor the user’s site-packages directory.

Set by the-I option.

Added in version 3.4.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_LegacyWindowsFSEncodingFlag

This API is kept for backward compatibility: settingPyPreConfig.legacy_windows_fs_encoding should be used instead, seePython Initialization Configuration.

If the flag is non-zero, use thembcs encoding withreplace errorhandler, instead of the UTF-8 encoding withsurrogatepass error handler,for thefilesystem encoding and error handler.

Set to1 if thePYTHONLEGACYWINDOWSFSENCODING environmentvariable is set to a non-empty string.

SeePEP 529 for more details.

Availability: Windows.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_LegacyWindowsStdioFlag

This API is kept for backward compatibility: settingPyConfig.legacy_windows_stdio should be used instead, seePython Initialization Configuration.

If the flag is non-zero, useio.FileIO instead ofio._WindowsConsoleIO forsys standard streams.

Set to1 if thePYTHONLEGACYWINDOWSSTDIO environmentvariable is set to a non-empty string.

SeePEP 528 for more details.

Availability: Windows.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_NoSiteFlag

This API is kept for backward compatibility: settingPyConfig.site_import should be used instead, seePython Initialization Configuration.

Disable the import of the modulesite and the site-dependentmanipulations ofsys.path that it entails. Also disable thesemanipulations ifsite is explicitly imported later (callsite.main() if you want them to be triggered).

Set by the-S option.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_NoUserSiteDirectory

This API is kept for backward compatibility: settingPyConfig.user_site_directory should be used instead, seePython Initialization Configuration.

Don’t add theusersite-packagesdirectory tosys.path.

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

Deprecated since version 3.12, will be removed in version 3.14.

intPy_OptimizeFlag

This API is kept for backward compatibility: settingPyConfig.optimization_level should be used instead, seePython Initialization Configuration.

Set by the-O option and thePYTHONOPTIMIZE environmentvariable.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_QuietFlag

This API is kept for backward compatibility: settingPyConfig.quiet should be used instead, seePythonInitialization Configuration.

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

Set by the-q option.

Added in version 3.2.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_UnbufferedStdioFlag

This API is kept for backward compatibility: settingPyConfig.buffered_stdio should be used instead, seePythonInitialization Configuration.

Force the stdout and stderr streams to be unbuffered.

Set by the-u option and thePYTHONUNBUFFEREDenvironment variable.

Deprecated since version 3.12, will be removed in version 3.14.

intPy_VerboseFlag

This API is kept for backward compatibility: settingPyConfig.verbose should be used instead, seePythonInitialization Configuration.

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 equalto2, print a message for each file that is checked for whensearching for a module. Also provides information on module cleanup at exit.

Set by the-v option and thePYTHONVERBOSE environmentvariable.

Deprecated since version 3.12, will be removed in version 3.14.

Initializing and finalizing the interpreter

voidPy_Initialize()
Part of theStable ABI.

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

This initializes the table of loaded modules (sys.modules), and createsthe fundamental modulesbuiltins,__main__ andsys.It also initializes the module search path (sys.path). It does not setsys.argv; use thePython Initialization ConfigurationAPI for that. This is a no-op when called for a second time (without callingPy_FinalizeEx() first). There is no return value; it is a fatalerror if the initialization fails.

UsePy_InitializeFromConfig() to customize thePython Initialization Configuration.

Note

On Windows, changes the console mode fromO_TEXT toO_BINARY,which will also affect non-Python uses of the console using the C Runtime.

voidPy_InitializeEx(intinitsigs)
Part of theStable ABI.

This function works likePy_Initialize() ifinitsigs is1. Ifinitsigs is0, it skips initialization registration of signal handlers,which may be useful when CPython is embedded as part of a larger application.

UsePy_InitializeFromConfig() to customize thePython Initialization Configuration.

PyStatusPy_InitializeFromConfig(constPyConfig*config)

Initialize Python fromconfig configuration, as described inInitialization with PyConfig.

See thePython Initialization Configuration section for details on pre-initializing theinterpreter, populating the runtime configuration structure, and queryingthe returned status structure.

intPy_IsInitialized()
Part of theStable ABI.

Return true (nonzero) when the Python interpreter has been initialized, false(zero) if not. AfterPy_FinalizeEx() is called, this returns false untilPy_Initialize() is called again.

intPy_IsFinalizing()
Part of theStable ABI since version 3.13.

Return true (non-zero) if the main Python interpreter isshutting down. Return false (zero) otherwise.

Added in version 3.13.

intPy_FinalizeEx()
Part of theStable ABI since version 3.6.

Undo all initializations made byPy_Initialize() and subsequent use ofPython/C API functions, and destroy all sub-interpreters (seePy_NewInterpreter() below) that were created and not yet destroyed sincethe last call toPy_Initialize(). Ideally, this frees all memoryallocated by the Python interpreter. This is a no-op when called for a secondtime (without callingPy_Initialize() again first).

Since this is the reverse ofPy_Initialize(), it should be calledin the same thread with the same interpreter active. That meansthe main thread and the main interpreter.This should never be called whilePy_RunMain() is running.

Normally the return value is0.If there were errors during finalization (flushing buffered data),-1 is returned.

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

Bugs and caveats: The destruction of modules and objects in modules is donein random order; this may cause destructors (__del__() methods) to failwhen they depend on other objects (even functions) or modules. Dynamicallyloaded extension modules loaded by Python are not unloaded. Small amounts ofmemory 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 notfreed. Some memory allocated by extension modules may not be freed. Someextensions may not work properly if their initialization routine is called morethan once; this can happen if an application callsPy_Initialize() andPy_FinalizeEx() more than once.

Raises anauditing eventcpython._PySys_ClearAuditHooks with no arguments.

Added in version 3.6.

voidPy_Finalize()
Part of theStable ABI.

This is a backwards-compatible version ofPy_FinalizeEx() thatdisregards the return value.

intPy_BytesMain(intargc,char**argv)
Part of theStable ABI since version 3.8.

Similar toPy_Main() butargv is an array of bytes strings,allowing the calling application to delegate the text decoding step tothe CPython runtime.

Added in version 3.8.

intPy_Main(intargc,wchar_t**argv)
Part of theStable ABI.

The main program for the standard interpreter, encapsulating a fullinitialization/finalization cycle, as well as additionalbehaviour to implement reading configurations settings from the environmentand command line, and then executing__main__ in accordance withCommand line.

This is made available for programs which wish to support the full CPythoncommand line interface, rather than just embedding a Python runtime in alarger application.

Theargc andargv parameters are similar to those which are passed to aC program’smain() function, except that theargv entries are firstconverted towchar_t usingPy_DecodeLocale(). It is alsoimportant to note that the argument list entries may be modified to point tostrings other than those passed in (however, the contents of the stringspointed to by the argument list are not modified).

The return value will be0 if the interpreter exits normally (i.e.,without an exception),1 if the interpreter exits due to an exception,or2 if the argument list does not represent a valid Python commandline.

Note that if an otherwise unhandledSystemExit is raised, thisfunction will not return1, but exit the process, as long asPy_InspectFlag is not set. IfPy_InspectFlag is set, execution willdrop into the interactive Python prompt, at which point a second otherwiseunhandledSystemExit will still exit the process, while any othermeans of exiting will set the return value as described above.

In terms of the CPython runtime configuration APIs documented in theruntime configuration section (and without accountingfor error handling),Py_Main is approximately equivalent to:

PyConfigconfig;PyConfig_InitPythonConfig(&config);PyConfig_SetArgv(&config,argc,argv);Py_InitializeFromConfig(&config);PyConfig_Clear(&config);Py_RunMain();

In normal usage, an embedding application will call this functioninstead of callingPy_Initialize(),Py_InitializeEx() orPy_InitializeFromConfig() directly, and all settings will be appliedas described elsewhere in this documentation. If this function is insteadcalledafter a preceding runtime initialization API call, then exactlywhich environmental and command line configuration settings will be updatedis version dependent (as it depends on which settings correctly supportbeing modified after they have already been set once when the runtime wasfirst initialized).

intPy_RunMain(void)

Executes the main module in a fully configured CPython runtime.

Executes the command (PyConfig.run_command), the script(PyConfig.run_filename) or the module(PyConfig.run_module) specified on the command line or in theconfiguration. If none of these values are set, runs the interactive Pythonprompt (REPL) using the__main__ module’s global namespace.

IfPyConfig.inspect is not set (the default), the return valuewill be0 if the interpreter exits normally (that is, without raisingan exception), or1 if the interpreter exits due to an exception. If anotherwise unhandledSystemExit is raised, the function will immediatelyexit the process instead of returning1.

IfPyConfig.inspect is set (such as when the-i optionis used), rather than returning when the interpreter exits, execution willinstead resume in an interactive Python prompt (REPL) using the__main__module’s global namespace. If the interpreter exited with an exception, itis immediately raised in the REPL session. The function return value isthen determined by the way theREPL session terminates: returning0if the session terminates without raising an unhandled exception, exitingimmediately for an unhandledSystemExit, and returning1 forany other unhandled exception.

This function always finalizes the Python interpreter regardless of whetherit returns a value or immediately exits the process due to an unhandledSystemExit exception.

SeePython Configuration for an example of acustomized Python that always runs in isolated mode usingPy_RunMain().

intPyUnstable_AtExit(PyInterpreterState*interp,void(*func)(void*),void*data)
This isUnstable API. It may change without warning in minor releases.

Register anatexit callback for the target interpreterinterp.This is similar toPy_AtExit(), but takes an explicit interpreter anddata pointer for the callback.

TheGIL must be held forinterp.

Added in version 3.13.

Process-wide parameters

voidPy_SetProgramName(constwchar_t*name)
Part of theStable ABI.

This API is kept for backward compatibility: settingPyConfig.program_name should be used instead, seePythonInitialization Configuration.

This function should be called beforePy_Initialize() is called forthe first time, if it is called at all. It tells the interpreter the valueof theargv[0] argument to themain() function of the program(converted to wide characters).This is used byPy_GetPath() and some other functions below to findthe Python run-time libraries relative to the interpreter executable. Thedefault value is'python'. The argument should point to azero-terminated wide character string in static storage whose contents will notchange for the duration of the program’s execution. No code in the Pythoninterpreter will change the contents of this storage.

UsePy_DecodeLocale() to decode a bytes string to get awchar_t* string.

Deprecated since version 3.11.

wchar_t*Py_GetProgramName()
Part of theStable ABI.

Return the program name set withPyConfig.program_name, or the default.The returned string points into static storage; the caller should not modify itsvalue.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:Getsys.executable instead.

wchar_t*Py_GetPrefix()
Part of theStable ABI.

Return theprefix for installed platform-independent files. This is derivedthrough a number of complicated rules from the program name set withPyConfig.program_name and some environment variables; for example, if theprogram name is'/usr/local/bin/python', the prefix is'/usr/local'. Thereturned string points into static storage; the caller should not modify itsvalue. This corresponds to theprefix variable in the top-levelMakefile and the--prefix argument to theconfigurescript at build time. The value is available to Python code assys.base_prefix.It is only useful on Unix. See also the next function.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:Getsys.base_prefix instead, orsys.prefix ifvirtual environments need to be handled.

wchar_t*Py_GetExecPrefix()
Part of theStable ABI.

Return theexec-prefix for installed platform-dependent files. This isderived through a number of complicated rules from the program name set withPyConfig.program_name and some environment variables; for example, if theprogram name is'/usr/local/bin/python', the exec-prefix is'/usr/local'. The returned string points into static storage; the callershould not modify its value. This corresponds to theexec_prefixvariable in the top-levelMakefile and the--exec-prefixargument to theconfigure script at build time. The value isavailable to Python code assys.base_exec_prefix. It is only useful onUnix.

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

Generally speaking, a platform is a combination of hardware and softwarefamilies, e.g. Sparc machines running the Solaris 2.x operating system areconsidered the same platform, but Intel machines running Solaris 2.x are anotherplatform, and Intel machines running Linux are yet another platform. Differentmajor revisions of the same operating system generally also form differentplatforms. Non-Unix operating systems are a different story; the installationstrategies on those systems are so different that the prefix and exec-prefix aremeaningless, and set to the empty string. Note that compiled Python bytecodefiles are platform independent (but not independent from the Python version bywhich they were compiled!).

System administrators will know how to configure themount orautomount programs to share/usr/local between platformswhile having/usr/local/plat be a different filesystem for eachplatform.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:Getsys.base_exec_prefix instead, orsys.exec_prefix ifvirtual environments need to be handled.

wchar_t*Py_GetProgramFullPath()
Part of theStable ABI.

Return the full program name of the Python executable; this is computed as aside-effect of deriving the default module search path from the program name(set byPyConfig.program_name). The returned string points intostatic storage; the caller should not modify its value. The value is availableto Python code assys.executable.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:Getsys.executable instead.

wchar_t*Py_GetPath()
Part of theStable ABI.

Return the default module search path; this is computed from the program name(set byPyConfig.program_name) and some environment variables.The returned string consists of a series of directory names separated by aplatform dependent delimiter character. The delimiter character is':'on Unix and macOS,';' on Windows. The returned string points intostatic storage; the caller should not modify its value. The listsys.path is initialized with this value on interpreter startup; itcan be (and usually is) modified later to change the search path for loadingmodules.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:Getsys.path instead.

constchar*Py_GetVersion()
Part of theStable ABI.

Return the version of this Python interpreter. This is a string that lookssomething 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 characters are the major and minor version separated by aperiod. The returned string points into static storage; the caller should notmodify its value. The value is available to Python code assys.version.

See also thePy_Version constant.

constchar*Py_GetPlatform()
Part of theStable ABI.

Return the platform identifier for the current platform. On Unix, this isformed from the “official” name of the operating system, converted to lowercase, followed by the major revision number; e.g., for Solaris 2.x, which isalso known as SunOS 5.x, the value is'sunos5'. On macOS, it is'darwin'. On Windows, it is'win'. The returned string points intostatic storage; the caller should not modify its value. The value is availableto Python code assys.platform.

constchar*Py_GetCopyright()
Part of theStable ABI.

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

'Copyright1991-1995StichtingMathematischCentrum,Amsterdam'

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

constchar*Py_GetCompiler()
Part of theStable ABI.

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 itsvalue. The value is available to Python code as part of the variablesys.version.

constchar*Py_GetBuildInfo()
Part of theStable ABI.

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

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

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

voidPySys_SetArgvEx(intargc,wchar_t**argv,intupdatepath)
Part of theStable ABI.

This API is kept for backward compatibility: settingPyConfig.argv,PyConfig.parse_argv andPyConfig.safe_path should be used instead, seePythonInitialization Configuration.

Setsys.argv based onargc andargv. These parameters aresimilar to those passed to the program’smain() function with thedifference that the first entry should refer to the script file to beexecuted rather than the executable hosting the Python interpreter. If thereisn’t a script that will be run, the first entry inargv can be an emptystring. If this function fails to initializesys.argv, a fatalcondition is signalled usingPy_FatalError().

Ifupdatepath is zero, this is all the function does. Ifupdatepathis non-zero, the function also modifiessys.path according to thefollowing algorithm:

  • If the name of an existing script is passed inargv[0], the absolutepath of the directory where the script is located is prepended tosys.path.

  • Otherwise (that is, ifargc is0 orargv[0] doesn’t pointto an existing file name), an empty string is prepended tosys.path, which is the same as prepending the current workingdirectory (".").

UsePy_DecodeLocale() to decode a bytes string to get awchar_t* string.

See alsoPyConfig.orig_argv andPyConfig.argvmembers of thePython Initialization Configuration.

Note

It is recommended that applications embedding the Python interpreterfor purposes other than executing a single script pass0 asupdatepath,and updatesys.path themselves if desired.SeeCVE 2008-5983.

On versions before 3.1.3, you can achieve the same effect by manuallypopping the firstsys.path element after having calledPySys_SetArgv(), for example using:

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

Added in version 3.1.3.

Deprecated since version 3.11.

voidPySys_SetArgv(intargc,wchar_t**argv)
Part of theStable ABI.

This API is kept for backward compatibility: settingPyConfig.argv andPyConfig.parse_argv should be usedinstead, seePython Initialization Configuration.

This function works likePySys_SetArgvEx() withupdatepath setto1 unless thepython interpreter was started with the-I.

UsePy_DecodeLocale() to decode a bytes string to get awchar_t* string.

See alsoPyConfig.orig_argv andPyConfig.argvmembers of thePython Initialization Configuration.

Changed in version 3.4:Theupdatepath value depends on-I.

Deprecated since version 3.11.

voidPy_SetPythonHome(constwchar_t*home)
Part of theStable ABI.

This API is kept for backward compatibility: settingPyConfig.home should be used instead, seePythonInitialization Configuration.

Set the default “home” directory, that is, the location of the standardPython libraries. SeePYTHONHOME for the meaning of theargument string.

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

UsePy_DecodeLocale() to decode a bytes string to get awchar_t* string.

Deprecated since version 3.11.

wchar_t*Py_GetPythonHome()
Part of theStable ABI.

Return the default “home”, that is, the value set byPyConfig.home, or the value of thePYTHONHOMEenvironment variable if it is set.

This function should not be called beforePy_Initialize(), otherwiseit returnsNULL.

Changed in version 3.10:It now returnsNULL if called beforePy_Initialize().

Deprecated since version 3.13, will be removed in version 3.15:GetPyConfig.home orPYTHONHOME environmentvariable instead.

Thread State and the Global Interpreter Lock

The Python interpreter is not fully thread-safe. In order to supportmulti-threaded Python programs, there’s a global lock, called theglobalinterpreter lock orGIL, that must be held by the current thread beforeit can safely access Python objects. Without the lock, even the simplestoperations could cause problems in a multi-threaded program: for example, whentwo threads simultaneously increment the reference count of the same object, thereference count could end up being incremented only once instead of twice.

Therefore, the rule exists that only the thread that has acquired theGIL may operate on Python objects or call Python/C API functions.In order to emulate concurrency of execution, the interpreter regularlytries to switch threads (seesys.setswitchinterval()). The lock is alsoreleased around potentially blocking I/O operations like reading or writinga file, so that other Python threads can run in the meantime.

The Python interpreter keeps some thread-specific bookkeeping informationinside a data structure calledPyThreadState. There’s also oneglobal variable pointing to the currentPyThreadState: it canbe retrieved usingPyThreadState_Get().

Releasing the GIL from extension code

Most extension code manipulating theGIL has the following simplestructure:

Savethethreadstateinalocalvariable.Releasetheglobalinterpreterlock....DosomeblockingI/Ooperation...Reacquiretheglobalinterpreterlock.Restorethethreadstatefromthelocalvariable.

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

Py_BEGIN_ALLOW_THREADS...DosomeblockingI/Ooperation...Py_END_ALLOW_THREADS

ThePy_BEGIN_ALLOW_THREADS macro opens a new block and declares ahidden local variable; thePy_END_ALLOW_THREADS macro closes theblock.

The block above expands to the following code:

PyThreadState*_save;_save=PyEval_SaveThread();...DosomeblockingI/Ooperation...PyEval_RestoreThread(_save);

Here is how these functions work: the global interpreter lock is used to protect the pointer to thecurrent 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 threadstate in the global variable). Conversely, when acquiring the lock and restoringthe thread state, the lock must be acquired before storing the thread statepointer.

Note

Calling system I/O functions is the most common use case for releasingthe GIL, but it can also be useful before calling long-running computationswhich don’t need access to Python objects, such as compression orcryptographic functions operating over memory buffers. For example, thestandardzlib andhashlib modules release the GIL whencompressing or hashing data.

Non-Python created threads

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

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

ThePyGILState_Ensure() andPyGILState_Release() functions doall of the above automatically. The typical idiom for calling into Pythonfrom a C thread is:

PyGILState_STATEgstate;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 thePyGILState_* functions assume there is only one globalinterpreter (created automatically byPy_Initialize()). Pythonsupports the creation of additional interpreters (usingPy_NewInterpreter()), but mixing multiple interpreters and thePyGILState_* API is unsupported.

Cautions about fork()

Another important thing to note about threads is their behaviour in the faceof the Cfork() call. On most systems withfork(), after aprocess forks only the thread that issued the fork will exist. This has aconcrete impact both on how locks must be handled and on all stored statein CPython’s runtime.

The fact that only the “current” thread remainsmeans any locks held by other threads will never be released. Python solvesthis foros.fork() by acquiring the locks it uses internally beforethe fork, and releasing them afterwards. In addition, it resets anyLock objects in the child. When extending or embedding Python, thereis no way to inform Python of additional (non-Python) locks that need to beacquired before or reset after a fork. OS facilities such aspthread_atfork() would need to be used to accomplish the same thing.Additionally, when extending or embedding Python, callingfork()directly rather than throughos.fork() (and returning to or callinginto Python) may result in a deadlock by one of Python’s internal locksbeing held by a thread that is defunct after the fork.PyOS_AfterFork_Child() tries to reset the necessary locks, but is notalways able to.

The fact that all other threads go away also means that CPython’sruntime state there must be cleaned up properly, whichos.fork()does. This means finalizing all otherPyThreadState objectsbelonging to the current interpreter and all otherPyInterpreterState objects. Due to this and the specialnature 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 ifexec() will be called immediatelyafter.

High-level API

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

typePyInterpreterState
Part of theLimited API (as an opaque struct).

This data structure represents the state shared by a number of cooperatingthreads. Threads belonging to the same interpreter share their moduleadministration and a few other internal items. There are no public members inthis structure.

Threads belonging to different interpreters initially share nothing, exceptprocess state like available memory, open file descriptors and such. The globalinterpreter lock is also shared by all threads, regardless of to whichinterpreter they belong.

typePyThreadState
Part of theLimited API (as an opaque struct).

This data structure represents the state of a single thread. The only publicdata member is:

PyInterpreterState*interp

This thread’s interpreter state.

voidPyEval_InitThreads()
Part of theStable ABI.

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 byPy_Initialize(), so you don’thave to call it yourself anymore.

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

Deprecated since version 3.9.

PyThreadState*PyEval_SaveThread()
Part of theStable ABI.

Release the global interpreter lock (if it has been created) and reset thethread state toNULL, returning the previous thread state (which is notNULL). If the lock has been created, the current thread must haveacquired it.

voidPyEval_RestoreThread(PyThreadState*tstate)
Part of theStable ABI.

Acquire the global interpreter lock (if it has been created) and set thethread state totstate, which must not beNULL. If the lock has beencreated, the current thread must not have acquired it, otherwise deadlockensues.

Note

Calling this function from a thread when the runtime is finalizingwill terminate the thread, even if the thread was not created by Python.You can usePy_IsFinalizing() orsys.is_finalizing() tocheck if the interpreter is in process of being finalized before callingthis function to avoid unwanted termination.

PyThreadState*PyThreadState_Get()
Part of theStable ABI.

Return the current thread state. The global interpreter lock must be held.When the current thread state isNULL, this issues a fatal error (so thatthe caller needn’t check forNULL).

See alsoPyThreadState_GetUnchecked().

PyThreadState*PyThreadState_GetUnchecked()

Similar toPyThreadState_Get(), but don’t kill the process with afatal error if it is NULL. The caller is responsible to check if the resultis NULL.

Added in version 3.13:In Python 3.5 to 3.12, the function was private and known as_PyThreadState_UncheckedGet().

PyThreadState*PyThreadState_Swap(PyThreadState*tstate)
Part of theStable ABI.

Swap the current thread state with the thread state given by the argumenttstate, which may beNULL.

TheGIL does not need to be held, but will be held upon returningiftstate is non-NULL.

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

PyGILState_STATEPyGILState_Ensure()
Part of theStable ABI.

Ensure that the current thread is ready to call the Python C API regardlessof the current state of Python, or of the global interpreter lock. This maybe called as many times as desired by a thread as long as each call ismatched with a call toPyGILState_Release(). In general, otherthread-related APIs may be used betweenPyGILState_Ensure() andPyGILState_Release() calls as long as the thread state is restored toits previous state before the Release(). For example, normal usage of thePy_BEGIN_ALLOW_THREADS andPy_END_ALLOW_THREADS macros isacceptable.

The return value is an opaque “handle” to the thread state whenPyGILState_Ensure() was called, and must be passed toPyGILState_Release() to ensure Python is left in the same state. Eventhough recursive calls are allowed, these handlescannot be shared - eachunique call toPyGILState_Ensure() must save the handle for its calltoPyGILState_Release().

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

Note

Calling this function from a thread when the runtime is finalizingwill terminate the thread, even if the thread was not created by Python.You can usePy_IsFinalizing() orsys.is_finalizing() tocheck if the interpreter is in process of being finalized before callingthis function to avoid unwanted termination.

voidPyGILState_Release(PyGILState_STATE)
Part of theStable ABI.

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

Every call toPyGILState_Ensure() must be matched by a call toPyGILState_Release() on the same thread.

PyThreadState*PyGILState_GetThisThreadState()
Part of theStable ABI.

Get the current thread state for this thread. May returnNULL if noGILState API has been used on the current thread. Note that the main threadalways has such a thread-state, even if no auto-thread-state call has beenmade on the main thread. This is mainly a helper/diagnostic function.

intPyGILState_Check()

Return1 if the current thread is holding the GIL and0 otherwise.This function can be called from any thread at any time.Only if it has had its Python thread state initialized and currently isholding the GIL will it return1.This is mainly a helper/diagnostic function. It can be usefulfor example in callback contexts or memory allocation functions whenknowing that the GIL is locked can allow the caller to perform sensitiveactions or otherwise behave differently.

Added in version 3.4.

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

Py_BEGIN_ALLOW_THREADS
Part of theStable ABI.

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

Py_END_ALLOW_THREADS
Part of theStable ABI.

This macro expands toPyEval_RestoreThread(_save);}. Note that it containsa closing brace; it must be matched with an earlierPy_BEGIN_ALLOW_THREADS macro. See above for further discussion ofthis macro.

Py_BLOCK_THREADS
Part of theStable ABI.

This macro expands toPyEval_RestoreThread(_save);: it is equivalent toPy_END_ALLOW_THREADS without the closing brace.

Py_UNBLOCK_THREADS
Part of theStable ABI.

This macro expands to_save=PyEval_SaveThread();: it is equivalent toPy_BEGIN_ALLOW_THREADS without the opening brace and variabledeclaration.

Low-level API

All of the following functions must be called afterPy_Initialize().

Changed in version 3.7:Py_Initialize() now initializes theGIL.

PyInterpreterState*PyInterpreterState_New()
Part of theStable ABI.

Create a new interpreter state object. The global interpreter lock need notbe held, but may be held if it is necessary to serialize calls to thisfunction.

Raises anauditing eventcpython.PyInterpreterState_New with no arguments.

voidPyInterpreterState_Clear(PyInterpreterState*interp)
Part of theStable ABI.

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

Raises anauditing eventcpython.PyInterpreterState_Clear with no arguments.

voidPyInterpreterState_Delete(PyInterpreterState*interp)
Part of theStable ABI.

Destroy an interpreter state object. The global interpreter lock need not beheld. The interpreter state must have been reset with a previous call toPyInterpreterState_Clear().

PyThreadState*PyThreadState_New(PyInterpreterState*interp)
Part of theStable ABI.

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 isnecessary to serialize calls to this function.

voidPyThreadState_Clear(PyThreadState*tstate)
Part of theStable ABI.

Reset all information in a thread state object. The global interpreter lockmust be held.

Changed in version 3.9:This function now calls thePyThreadState.on_delete callback.Previously, that happened inPyThreadState_Delete().

Changed in version 3.13:ThePyThreadState.on_delete callback was removed.

voidPyThreadState_Delete(PyThreadState*tstate)
Part of theStable ABI.

Destroy a thread state object. The global interpreter lock need not be held.The thread state must have been reset with a previous call toPyThreadState_Clear().

voidPyThreadState_DeleteCurrent(void)

Destroy the current thread state and release the global interpreter lock.LikePyThreadState_Delete(), the global interpreter lock mustbe held. The thread state must have been reset with a previous calltoPyThreadState_Clear().

PyFrameObject*PyThreadState_GetFrame(PyThreadState*tstate)
Part of theStable ABI since version 3.10.

Get the current frame of the Python thread statetstate.

Return astrong reference. ReturnNULL if no frame is currentlyexecuting.

See alsoPyEval_GetFrame().

tstate must not beNULL.

Added in version 3.9.

uint64_tPyThreadState_GetID(PyThreadState*tstate)
Part of theStable ABI since version 3.10.

Get the unique thread state identifier of the Python thread statetstate.

tstate must not beNULL.

Added in version 3.9.

PyInterpreterState*PyThreadState_GetInterpreter(PyThreadState*tstate)
Part of theStable ABI since version 3.10.

Get the interpreter of the Python thread statetstate.

tstate must not beNULL.

Added in version 3.9.

voidPyThreadState_EnterTracing(PyThreadState*tstate)

Suspend tracing and profiling in the Python thread statetstate.

Resume them using thePyThreadState_LeaveTracing() function.

Added in version 3.11.

voidPyThreadState_LeaveTracing(PyThreadState*tstate)

Resume tracing and profiling in the Python thread statetstate suspendedby thePyThreadState_EnterTracing() function.

See alsoPyEval_SetTrace() andPyEval_SetProfile()functions.

Added in version 3.11.

PyInterpreterState*PyInterpreterState_Get(void)
Part of theStable ABI since version 3.9.

Get the current interpreter.

Issue a fatal error if there no current Python thread state or no currentinterpreter. It cannot return NULL.

The caller must hold the GIL.

Added in version 3.9.

int64_tPyInterpreterState_GetID(PyInterpreterState*interp)
Part of theStable ABI since version 3.7.

Return the interpreter’s unique ID. If there was any error in doingso then-1 is returned and an error is set.

The caller must hold the GIL.

Added in version 3.7.

PyObject*PyInterpreterState_GetDict(PyInterpreterState*interp)
Part of theStable ABI since version 3.8.

Return a dictionary in which interpreter-specific data may be stored.If this function returnsNULL then no exception has been raised andthe caller should assume no interpreter-specific dict is available.

This is not a replacement forPyModule_GetState(), whichextensions should use to store interpreter-specific state information.

Added in version 3.8.

PyObject*PyUnstable_InterpreterState_GetMainModule(PyInterpreterState*interp)
This isUnstable API. It may change without warning in minor releases.

Return astrong reference to the__main__module objectfor the given interpreter.

The caller must hold the GIL.

Added in version 3.13.

typedefPyObject*(*_PyFrameEvalFunction)(PyThreadState*tstate,_PyInterpreterFrame*frame,intthrowflag)

Type of a frame evaluation function.

Thethrowflag parameter is used by thethrow() method of generators:if non-zero, handle the current exception.

Changed in version 3.9:The function now takes atstate parameter.

Changed in version 3.11:Theframe parameter changed fromPyFrameObject* to_PyInterpreterFrame*.

_PyFrameEvalFunction_PyInterpreterState_GetEvalFrameFunc(PyInterpreterState*interp)

Get the frame evaluation function.

See thePEP 523 “Adding a frame evaluation API to CPython”.

Added in version 3.9.

void_PyInterpreterState_SetEvalFrameFunc(PyInterpreterState*interp,_PyFrameEvalFunctioneval_frame)

Set the frame evaluation function.

See thePEP 523 “Adding a frame evaluation API to CPython”.

Added in version 3.9.

PyObject*PyThreadState_GetDict()
Return value: Borrowed reference. Part of theStable ABI.

Return a dictionary in which extensions can store thread-specific stateinformation. Each extension should use a unique key to use to store state inthe dictionary. It is okay to call this function when no current thread stateis available. If this function returnsNULL, no exception has been raised andthe caller should assume no current thread state is available.

intPyThreadState_SetAsyncExc(unsignedlongid,PyObject*exc)
Part of theStable ABI.

Asynchronously raise an exception in a thread. Theid argument is the threadid of the target thread;exc is the exception object to be raised. Thisfunction does not steal any references toexc. To prevent naive misuse, youmust write your own C extension to call this. Must be called with the GIL held.Returns the number of thread states modified; this is normally one, but will bezero if the thread id isn’t found. Ifexc isNULL, the pendingexception (if any) for the thread is cleared. This raises no exceptions.

Changed in version 3.7:The type of theid parameter changed fromlong tounsignedlong.

voidPyEval_AcquireThread(PyThreadState*tstate)
Part of theStable ABI.

Acquire the global interpreter lock and set the current thread state totstate, which must not beNULL. The lock must have been created earlier.If this thread already has the lock, deadlock ensues.

Note

Calling this function from a thread when the runtime is finalizingwill terminate the thread, even if the thread was not created by Python.You can usePy_IsFinalizing() orsys.is_finalizing() tocheck if the interpreter is in process of being finalized before callingthis function to avoid unwanted termination.

Changed in version 3.8:Updated to be consistent withPyEval_RestoreThread(),Py_END_ALLOW_THREADS(), andPyGILState_Ensure(),and terminate the current thread if called while the interpreter is finalizing.

PyEval_RestoreThread() is a higher-level function which is alwaysavailable (even when threads have not been initialized).

voidPyEval_ReleaseThread(PyThreadState*tstate)
Part of theStable ABI.

Reset the current thread state toNULL and release the global interpreterlock. The lock must have been created earlier and must be held by the currentthread. Thetstate argument, which must not beNULL, is only used to checkthat it represents the current thread state — if it isn’t, a fatal error isreported.

PyEval_SaveThread() is a higher-level function which is alwaysavailable (even when threads have not been initialized).

Sub-interpreter support

While in most uses, you will only embed a single Python interpreter, thereare cases where you need to create several independent interpreters in thesame process and perhaps even in the same thread. Sub-interpreters allowyou 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 signalhandling. It is also responsible for execution during runtime initialization andis usually the active interpreter during runtime finalization. ThePyInterpreterState_Main() function returns a pointer to its state.

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

typePyInterpreterConfig

Structure containing most parameters to configure a sub-interpreter.Its values are used only inPy_NewInterpreterFromConfig() andnever modified by the runtime.

Added in version 3.12.

Structure fields:

intuse_main_obmalloc

If this is0 then the sub-interpreter will use its own“object” allocator state.Otherwise it will use (share) the main interpreter’s.

If this is0 thencheck_multi_interp_extensionsmust be1 (non-zero).If this is1 thengilmust not bePyInterpreterConfig_OWN_GIL.

intallow_fork

If this is0 then the runtime will not support forking theprocess in any thread where the sub-interpreter is currently active.Otherwise fork is unrestricted.

Note that thesubprocess module still workswhen fork is disallowed.

intallow_exec

If this is0 then the runtime will not support replacing thecurrent process via exec (e.g.os.execv()) in any threadwhere the sub-interpreter is currently active.Otherwise exec is unrestricted.

Note that thesubprocess module still workswhen exec is disallowed.

intallow_threads

If this is0 then the sub-interpreter’sthreading modulewon’t create threads.Otherwise threads are allowed.

intallow_daemon_threads

If this is0 then the sub-interpreter’sthreading modulewon’t create daemon threads.Otherwise daemon threads are allowed (as long asallow_threads is non-zero).

intcheck_multi_interp_extensions

If this is0 then all extension modules may be imported,including legacy (single-phase init) modules,in any thread where the sub-interpreter is currently active.Otherwise only multi-phase init extension modules(seePEP 489) may be imported.(Also seePy_mod_multiple_interpreters.)

This must be1 (non-zero) ifuse_main_obmalloc is0.

intgil

This determines the operation of the GIL for the sub-interpreter.It may be one of the following:

PyInterpreterConfig_DEFAULT_GIL

Use the default selection (PyInterpreterConfig_SHARED_GIL).

PyInterpreterConfig_SHARED_GIL

Use (share) the main interpreter’s GIL.

PyInterpreterConfig_OWN_GIL

Use the sub-interpreter’s own GIL.

If this isPyInterpreterConfig_OWN_GIL thenPyInterpreterConfig.use_main_obmalloc must be0.

PyStatusPy_NewInterpreterFromConfig(PyThreadState**tstate_p,constPyInterpreterConfig*config)

Create a new sub-interpreter. This is an (almost) totally separate environmentfor the execution of Python code. In particular, the new interpreter hasseparate, independent versions of all imported modules, including thefundamental modulesbuiltins,__main__ andsys. Thetable of loaded modules (sys.modules) and the module search path(sys.path) are also separate. The new environment has nosys.argvvariable. It has new standard I/O stream file objectssys.stdin,sys.stdout andsys.stderr (however these refer to the same underlyingfile descriptors).

The givenconfig controls the options with which the interpreteris initialized.

Upon success,tstate_p will be set to the first thread statecreated in the newsub-interpreter. This thread state is made in the current thread state.Note that no actual thread is created; see the discussion of thread statesbelow. If creation of the new interpreter is unsuccessful,tstate_p is set toNULL;no exception is set since the exception state is stored in thecurrent thread state and there may not be a current thread state.

Like all other Python/C API functions, the global interpreter lockmust be held before calling this function and is still held when itreturns. Likewise a current thread state must be set on entry. Onsuccess, the returned thread state will be set as current. If thesub-interpreter is created with its own GIL then the GIL of thecalling interpreter will be released. When the function returns,the new interpreter’s GIL will be held by the current thread andthe previously interpreter’s GIL will remain released here.

Added in version 3.12.

Sub-interpreters are most effective when isolated from each other,with certain functionality restricted:

PyInterpreterConfigconfig={.use_main_obmalloc=0,.allow_fork=0,.allow_exec=0,.allow_threads=1,.allow_daemon_threads=0,.check_multi_interp_extensions=1,.gil=PyInterpreterConfig_OWN_GIL,};PyThreadState*tstate=NULL;PyStatusstatus=Py_NewInterpreterFromConfig(&tstate,&config);if(PyStatus_Exception(status)){Py_ExitStatusException(status);}

Note that the config is used only briefly and does not get modified.During initialization the config’s values are converted into variousPyInterpreterState values. A read-only copy of the configmay be stored internally on thePyInterpreterState.

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

  • For modules using multi-phase initialization,e.g.PyModule_FromDefAndSpec(), a separate module object iscreated and initialized for each interpreter.Only C-level static and global variables are shared between thesemodule objects.

  • For modules using single-phase initialization,e.g.PyModule_Create(), the first time a particular extensionis imported, it is initialized normally, and a (shallow) copy of itsmodule’s dictionary is squirreled away.When the same extension is imported by another (sub-)interpreter, a newmodule is initialized and filled with the contents of this copy; theextension’sinit function is not called.Objects in the module’s dictionary thus end up shared across(sub-)interpreters, which might cause unwanted behavior (seeBugs and caveats below).

    Note that this is different from what happens when an extension isimported after the interpreter has been completely re-initialized bycallingPy_FinalizeEx() andPy_Initialize(); in thatcase, the extension’sinitmodule functionis called again.As with multi-phase initialization, this means that only C-level staticand global variables are shared between these modules.

PyThreadState*Py_NewInterpreter(void)
Part of theStable ABI.

Create a new sub-interpreter. This is essentially just a wrapperaroundPy_NewInterpreterFromConfig() with a config thatpreserves the existing behavior. The result is an unisolatedsub-interpreter that shares the main interpreter’s GIL, allowsfork/exec, allows daemon threads, and allows single-phase initmodules.

voidPy_EndInterpreter(PyThreadState*tstate)
Part of theStable ABI.

Destroy the (sub-)interpreter represented by the given thread state.The given thread state must be the current thread state. See thediscussion of thread states below. When the call returns,the current thread state isNULL. All thread states associatedwith this interpreter are destroyed. The global interpreter lockused by the target interpreter must be held before calling thisfunction. No GIL is held when it returns.

Py_FinalizeEx() will destroy all sub-interpreters thathaven’t been explicitly destroyed at that point.

A Per-Interpreter GIL

UsingPy_NewInterpreterFromConfig() you can createa sub-interpreter that is completely isolated from other interpreters,including having its own GIL. The most important benefit of thisisolation is that such an interpreter can execute Python code withoutbeing blocked by other interpreters or blocking any others. Thus asingle Python process can truly take advantage of multiple CPU coreswhen running Python code. The isolation also encourages a differentapproach to concurrency than that of just using threads.(SeePEP 554.)

Using an isolated interpreter requires vigilance in preserving thatisolation. That especially means not sharing any objects or mutablestate without guarantees about thread-safety. Even objects that areotherwise immutable (e.g.None,(1,5)) can’t normally be sharedbecause of the refcount. One simple but less-efficient approach aroundthis is to use a global lock around all use of some state (or object).Alternately, effectively immutable objects (like integers or strings)can be made safe in spite of their refcounts by making themimmortal.In fact, this has been done for the builtin singletons, small integers,and a number of other builtin objects.

If you preserve isolation then you will have access to proper multi-corecomputing without the complications that come with free-threading.Failure to preserve isolation will expose you to the full consequencesof free-threading, including races and hard-to-debug crashes.

Aside from that, one of the main challenges of using multiple isolatedinterpreters is how to communicate between them safely (not breakisolation) and efficiently. The runtime and stdlib do not provideany standard approach to this yet. A future stdlib module would helpmitigate the effort of preserving isolation and expose effective toolsfor communicating (and sharing) data between interpreters.

Added in version 3.12.

Bugs and caveats

Because sub-interpreters (and the main interpreter) are part of the sameprocess, the insulation between them isn’t perfect — for example, usinglow-level file operations likeos.close() they can(accidentally or maliciously) affect each other’s open files. Because of theway extensions are shared between (sub-)interpreters, some extensions may notwork properly; this is especially likely when using single-phase initializationor (static) global variables.It is possible to insert objects created in one sub-interpreter intoa 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 importoperations executed by such objects may affect the wrong (sub-)interpreter’sdictionary of loaded modules. It is equally important to avoid sharingobjects from which the above are reachable.

Also note that combining this functionality withPyGILState_* APIsis delicate, because these APIs assume a bijection between Python thread statesand 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 pairof matchingPyGILState_Ensure() andPyGILState_Release() calls.Furthermore, extensions (such asctypes) using these APIs to allow callingof Python code from non-Python created threads will probably be broken when usingsub-interpreters.

Asynchronous Notifications

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

intPy_AddPendingCall(int(*func)(void*),void*arg)
Part of theStable ABI.

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

When successfully queued,func will beeventually called from themain interpreter thread with the argumentarg. It will be calledasynchronously with respect to normally running Python code, but withboth these conditions met:

func must return0 on success, or-1 on failure with an exceptionset.func won’t be interrupted to perform another asynchronousnotification recursively, but it can still be interrupted to switchthreads if the global interpreter lock is released.

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

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

Warning

This is a low-level function, only useful for very special cases.There is no guarantee thatfunc will be called as quick aspossible. If the main thread is busy executing a system call,func won’t be called before the system call returns. Thisfunction is generallynot suitable for calling Python code fromarbitrary C threads. Instead, use thePyGILState API.

Added in version 3.1.

Changed in version 3.9:If this function is called in a subinterpreter, the functionfunc isnow scheduled to be called from the subinterpreter, rather than beingcalled from the main interpreter. Each subinterpreter now has its ownlist of scheduled calls.

Profiling and Tracing

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

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

typedefint(*Py_tracefunc)(PyObject*obj,PyFrameObject*frame,intwhat,PyObject*arg)

The type of the trace function registered usingPyEval_SetProfile() andPyEval_SetTrace(). The first parameter is the object passed to theregistration function asobj,frame is the frame object to which the eventpertains,what is one of the constantsPyTrace_CALL,PyTrace_EXCEPTION,PyTrace_LINE,PyTrace_RETURN,PyTrace_C_CALL,PyTrace_C_EXCEPTION,PyTrace_C_RETURN,orPyTrace_OPCODE, andarg depends on the value ofwhat:

Value ofwhat

Meaning ofarg

PyTrace_CALL

AlwaysPy_None.

PyTrace_EXCEPTION

Exception information as returned bysys.exc_info().

PyTrace_LINE

AlwaysPy_None.

PyTrace_RETURN

Value being returned to the caller,orNULL 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

AlwaysPy_None.

intPyTrace_CALL

The value of thewhat parameter to aPy_tracefunc function when a newcall 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 reportedas there is no control transfer to the Python bytecode in the correspondingframe.

intPyTrace_EXCEPTION

The value of thewhat parameter to aPy_tracefunc function when anexception has been raised. The callback function is called with this value forwhat when after any bytecode is processed after which the exception becomesset within the frame being executed. The effect of this is that as exceptionpropagation causes the Python stack to unwind, the callback is called uponreturn to each frame as the exception propagates. Only trace functions receivesthese events; they are not needed by the profiler.

intPyTrace_LINE

The value passed as thewhat parameter to aPy_tracefunc function(but not a profiling function) when a line-number event is being reported.It may be disabled for a frame by settingf_trace_lines to0 on that frame.

intPyTrace_RETURN

The value for thewhat parameter toPy_tracefunc functions when acall is about to return.

intPyTrace_C_CALL

The value for thewhat parameter toPy_tracefunc functions when a Cfunction is about to be called.

intPyTrace_C_EXCEPTION

The value for thewhat parameter toPy_tracefunc functions when a Cfunction has raised an exception.

intPyTrace_C_RETURN

The value for thewhat parameter toPy_tracefunc functions when a Cfunction has returned.

intPyTrace_OPCODE

The value for thewhat parameter toPy_tracefunc functions (but notprofiling functions) when a new opcode is about to be executed. This event isnot emitted by default: it must be explicitly requested by settingf_trace_opcodes to1 on the frame.

voidPyEval_SetProfile(Py_tracefuncfunc,PyObject*obj)

Set the profiler function tofunc. Theobj parameter is passed to thefunction as its first parameter, and may be any Python object, orNULL. Ifthe profile function needs to maintain state, using a different value forobjfor each thread provides a convenient and thread-safe place to store it. Theprofile function is called for all monitored events exceptPyTrace_LINEPyTrace_OPCODE andPyTrace_EXCEPTION.

See also thesys.setprofile() function.

The caller must hold theGIL.

voidPyEval_SetProfileAllThreads(Py_tracefuncfunc,PyObject*obj)

LikePyEval_SetProfile() but sets the profile function in all running threadsbelonging to the current interpreter instead of the setting it only on the current thread.

The caller must hold theGIL.

AsPyEval_SetProfile(), this function ignores any exceptions raised whilesetting the profile functions in all threads.

Added in version 3.12.

voidPyEval_SetTrace(Py_tracefuncfunc,PyObject*obj)

Set the tracing function tofunc. This is similar toPyEval_SetProfile(), except the tracing function does receive line-numberevents and per-opcode events, but does not receive any event related to C functionobjects being called. Any trace function registered usingPyEval_SetTrace()will not receivePyTrace_C_CALL,PyTrace_C_EXCEPTION orPyTrace_C_RETURN as a value for thewhat parameter.

See also thesys.settrace() function.

The caller must hold theGIL.

voidPyEval_SetTraceAllThreads(Py_tracefuncfunc,PyObject*obj)

LikePyEval_SetTrace() but sets the tracing function in all running threadsbelonging to the current interpreter instead of the setting it only on the current thread.

The caller must hold theGIL.

AsPyEval_SetTrace(), this function ignores any exceptions raised whilesetting the trace functions in all threads.

Added in version 3.12.

Reference tracing

Added in version 3.13.

typedefint(*PyRefTracer)(PyObject*,intevent,void*data)

The type of the trace function registered usingPyRefTracer_SetTracer().The first parameter is a Python object that has been just created (wheneventis set toPyRefTracer_CREATE) or about to be destroyed (wheneventis set toPyRefTracer_DESTROY). Thedata argument is the opaque pointerthat was provided whenPyRefTracer_SetTracer() was called.

Added in version 3.13.

intPyRefTracer_CREATE

The value for theevent parameter toPyRefTracer functions when a Pythonobject has been created.

intPyRefTracer_DESTROY

The value for theevent parameter toPyRefTracer functions when a Pythonobject has been destroyed.

intPyRefTracer_SetTracer(PyRefTracertracer,void*data)

Register a reference tracer function. The function will be called when a newPython has been created or when an object is going to be destroyed. Ifdata is provided it must be an opaque pointer that will be provided whenthe tracer function is called. Return0 on success. Set an exception andreturn-1 on error.

Not that tracer functionsmust not create Python objects inside orotherwise the call will be re-entrant. The tracer alsomust not clearany existing exception or set an exception. The GIL will be held every timethe tracer function is called.

The GIL must be held when calling this function.

Added in version 3.13.

PyRefTracerPyRefTracer_GetTracer(void**data)

Get the registered reference tracer function and the value of the opaque datapointer that was registered whenPyRefTracer_SetTracer() was called.If no tracer was registered this function will return NULL and will set thedata pointer to NULL.

The GIL must be held when calling this function.

Added in version 3.13.

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 afterinterp from the list of allsuch objects.

PyThreadState*PyInterpreterState_ThreadHead(PyInterpreterState*interp)

Return the pointer to the firstPyThreadState object in the list ofthreads associated with the interpreterinterp.

PyThreadState*PyThreadState_Next(PyThreadState*tstate)

Return the next thread state object aftertstate from the list of all suchobjects belonging to the samePyInterpreterState 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 thePython-level thread local storage API (threading.local). TheCPython C level APIs are similar to those offered by pthreads and Windows:use a thread key and functions to associate avoid* value perthread.

The GIL doesnot need to be held when calling these functions; they supplytheir own locking.

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

Note

None of these API functions handle memory management on behalf of thevoid* values. You need to allocate and deallocate them yourself.If thevoid* values happen to bePyObject*, thesefunctions 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 theCPython interpreter. This API uses a new typePy_tss_t instead ofint to represent thread keys.

Added in version 3.7.

See also

“A New C-API for Thread-Local Storage in CPython” (PEP 539)

typePy_tss_t

This data structure represents the state of a thread key, the definition ofwhich may depend on the underlying TLS implementation, and it has aninternal field representing the key’s initialization state. There are nopublic members in this structure.

WhenPy_LIMITED_API is not defined, static allocation ofthis type byPy_tss_NEEDS_INIT is allowed.

Py_tss_NEEDS_INIT

This macro expands to the initializer forPy_tss_t variables.Note that this macro won’t be defined withPy_LIMITED_API.

Dynamic Allocation

Dynamic allocation of thePy_tss_t, required in extension modulesbuilt withPy_LIMITED_API, where static allocation of this typeis not possible due to its implementation being opaque at build time.

Py_tss_t*PyThread_tss_alloc()
Part of theStable ABI since version 3.7.

Return a value which is the same state as a value initialized withPy_tss_NEEDS_INIT, orNULL in the case of dynamic allocationfailure.

voidPyThread_tss_free(Py_tss_t*key)
Part of theStable ABI since version 3.7.

Free the givenkey allocated byPyThread_tss_alloc(), afterfirst callingPyThread_tss_delete() to ensure any associatedthread locals have been unassigned. This is a no-op if thekeyargument isNULL.

Note

A freed key becomes a dangling pointer. You should reset the key toNULL.

Methods

The parameterkey of these functions must not beNULL. Moreover, thebehaviors ofPyThread_tss_set() andPyThread_tss_get() areundefined if the givenPy_tss_t has not been initialized byPyThread_tss_create().

intPyThread_tss_is_created(Py_tss_t*key)
Part of theStable ABI since version 3.7.

Return a non-zero value if the givenPy_tss_t has been initializedbyPyThread_tss_create().

intPyThread_tss_create(Py_tss_t*key)
Part of theStable ABI since version 3.7.

Return a zero value on successful initialization of a TSS key. The behavioris undefined if the value pointed to by thekey argument is notinitialized byPy_tss_NEEDS_INIT. This function can be calledrepeatedly on the same key – calling it on an already initialized key is ano-op and immediately returns success.

voidPyThread_tss_delete(Py_tss_t*key)
Part of theStable ABI since version 3.7.

Destroy a TSS key to forget the values associated with the key across allthreads, and change the key’s initialization state to uninitialized. Adestroyed key is able to be initialized again byPyThread_tss_create(). This function can be called repeatedly onthe same key – calling it on an already destroyed key is a no-op.

intPyThread_tss_set(Py_tss_t*key,void*value)
Part of theStable ABI since version 3.7.

Return a zero value to indicate successfully associating avoid*value with a TSS key in the current thread. Each thread has a distinctmapping of the key to avoid* value.

void*PyThread_tss_get(Py_tss_t*key)
Part of theStable ABI since version 3.7.

Return thevoid* value associated with a TSS key in the currentthread. This returnsNULL if no value is associated with the key in thecurrent thread.

Thread Local Storage (TLS) API

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

Note

This version of the API does not support platforms where the native TLS keyis defined in a way that cannot be safely cast toint. 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 notbe used in new code.

intPyThread_create_key()
Part of theStable ABI.
voidPyThread_delete_key(intkey)
Part of theStable ABI.
intPyThread_set_key_value(intkey,void*value)
Part of theStable ABI.
void*PyThread_get_key_value(intkey)
Part of theStable ABI.
voidPyThread_delete_key_value(intkey)
Part of theStable ABI.
voidPyThread_ReInitTLS()
Part of theStable ABI.

Synchronization Primitives

The C-API provides a basic mutual exclusion lock.

typePyMutex

A mutual exclusion lock. ThePyMutex should be initialized tozero to represent the unlocked state. For example:

PyMutexmutex={0};

Instances ofPyMutex should not be copied or moved. Both thecontents and address of aPyMutex are meaningful, and it mustremain at a fixed, writable location in memory.

Note

APyMutex currently occupies one byte, but the size should beconsidered unstable. The size may change in future Python releaseswithout a deprecation period.

Added in version 3.13.

voidPyMutex_Lock(PyMutex*m)

Lock mutexm. If another thread has already locked it, the callingthread will block until the mutex is unlocked. While blocked, the threadwill temporarily release theGIL if it is held.

Added in version 3.13.

voidPyMutex_Unlock(PyMutex*m)

Unlock mutexm. The mutex must be locked — otherwise, the function willissue a fatal error.

Added in version 3.13.

Python Critical Section API

The critical section API provides a deadlock avoidance layer on top ofper-object locks forfree-threaded CPython. They areintended to replace reliance on theglobal interpreter lock, and areno-ops in versions of Python with the global interpreter lock.

Critical sections avoid deadlocks by implicitly suspending active criticalsections and releasing the locks during calls toPyEval_SaveThread().WhenPyEval_RestoreThread() is called, the most recent critical sectionis resumed, and its locks reacquired. This means the critical section APIprovides weaker guarantees than traditional locks – they are useful becausetheir behavior is similar to theGIL.

The functions and structs used by the macros are exposed for caseswhere C macros are not available. They should only be used as in thegiven macro expansions. Note that the sizes and contents of the structures maychange in future Python versions.

Note

Operations that need to lock two objects at once must usePy_BEGIN_CRITICAL_SECTION2. Youcannot use nested criticalsections to lock more than one object at once, because the inner criticalsection may suspend the outer critical sections. This API does not providea way to lock more than two objects at once.

Example usage:

staticPyObject*set_field(MyObject*self,PyObject*value){Py_BEGIN_CRITICAL_SECTION(self);Py_SETREF(self->field,Py_XNewRef(value));Py_END_CRITICAL_SECTION();Py_RETURN_NONE;}

In the above example,Py_SETREF callsPy_DECREF, whichcan call arbitrary code through an object’s deallocation function. The criticalsection API avoids potential deadlocks due to reentrancy and lock orderingby allowing the runtime to temporarily suspend the critical section if thecode triggered by the finalizer blocks and callsPyEval_SaveThread().

Py_BEGIN_CRITICAL_SECTION(op)

Acquires the per-object lock for the objectop and begins acritical section.

In the free-threaded build, this macro expands to:

{PyCriticalSection_py_cs;PyCriticalSection_Begin(&_py_cs,(PyObject*)(op))

In the default build, this macro expands to{.

Added in version 3.13.

Py_END_CRITICAL_SECTION()

Ends the critical section and releases the per-object lock.

In the free-threaded build, this macro expands to:

PyCriticalSection_End(&_py_cs);}

In the default build, this macro expands to}.

Added in version 3.13.

Py_BEGIN_CRITICAL_SECTION2(a,b)

Acquires the per-objects locks for the objectsa andb and begins acritical section. The locks are acquired in a consistent order (lowestaddress first) to avoid lock ordering deadlocks.

In the free-threaded build, this macro expands to:

{PyCriticalSection2_py_cs2;PyCriticalSection2_Begin(&_py_cs2,(PyObject*)(a),(PyObject*)(b))

In the default build, this macro expands to{.

Added in version 3.13.

Py_END_CRITICAL_SECTION2()

Ends the critical section and releases the per-object locks.

In the free-threaded build, this macro expands to:

PyCriticalSection2_End(&_py_cs2);}

In the default build, this macro expands to}.

Added in version 3.13.