Doc/lib/libthread.tex

   1 \section{\module{thread} ---
   2          Multiple threads of control}
   3
   4 \declaremodule{builtin}{thread}
   5 \modulesynopsis{Create multiple threads of control within one interpreter.}
   6
   7
   8 This module provides low-level primitives for working with multiple
   9 threads (a.k.a.\ \dfn{light-weight processes} or \dfn{tasks}) --- multiple
  10 threads of control sharing their global data space.  For
  11 synchronization, simple locks (a.k.a.\ \dfn{mutexes} or \dfn{binary
  12 semaphores}) are provided.
  13 \index{light-weight processes}
  14 \index{processes, light-weight}
  15 \index{binary semaphores}
  16 \index{semaphores, binary}
  17
  18 The module is optional.  It is supported on Windows, Linux, SGI
  19 IRIX, Solaris 2.x, as well as on systems that have a \POSIX{} thread
  20 (a.k.a. ``pthread'') implementation.  For systems lacking the \module{thread}
  21 module, the \refmodule[dummythread]{dummy_thread} module is available.
  22 It duplicates this module's interface and can be
  23 used as a drop-in replacement.
  24 \index{pthreads}
  25 \indexii{threads}{\POSIX}
  26
  27 It defines the following constant and functions:
  28
  29 \begin{excdesc}{error}
  30 Raised on thread-specific errors.
  31 \end{excdesc}
  32
  33 \begin{datadesc}{LockType}
  34 This is the type of lock objects.
  35 \end{datadesc}
  36
  37 \begin{funcdesc}{start_new_thread}{function, args\optional{, kwargs}}
  38 Start a new thread and return its identifier.  The thread executes the function
  39 \var{function} with the argument list \var{args} (which must be a tuple).  The
  40 optional \var{kwargs} argument specifies a dictionary of keyword arguments.
  41 When the function returns, the thread silently exits.  When the function
  42 terminates with an unhandled exception, a stack trace is printed and
  43 then the thread exits (but other threads continue to run).
  44 \end{funcdesc}
  45
  46 \begin{funcdesc}{interrupt_main}{}
  47 Raise a KeyboardInterrupt in the main thread.  A subthread can use this
  48 function to interrupt the main thread.
  49 \versionadded{2.3}
  50 \end{funcdesc}
  51
  52 \begin{funcdesc}{exit}{}
  53 Raise the \exception{SystemExit} exception.  When not caught, this
  54 will cause the thread to exit silently.
  55 \end{funcdesc}
  56
  57 %\begin{funcdesc}{exit_prog}{status}
  58 %Exit all threads and report the value of the integer argument
  59 %\var{status} as the exit status of the entire program.
  60 %\strong{Caveat:} code in pending \keyword{finally} clauses, in this thread
  61 %or in other threads, is not executed.
  62 %\end{funcdesc}
  63
  64 \begin{funcdesc}{allocate_lock}{}
  65 Return a new lock object.  Methods of locks are described below.  The
  66 lock is initially unlocked.
  67 \end{funcdesc}
  68
  69 \begin{funcdesc}{get_ident}{}
  70 Return the `thread identifier' of the current thread.  This is a
  71 nonzero integer.  Its value has no direct meaning; it is intended as a
  72 magic cookie to be used e.g. to index a dictionary of thread-specific
  73 data.  Thread identifiers may be recycled when a thread exits and
  74 another thread is created.
  75 \end{funcdesc}
  76
  77
  78 Lock objects have the following methods:
  79
  80 \begin{methoddesc}[lock]{acquire}{\optional{waitflag}}
  81 Without the optional argument, this method acquires the lock
  82 unconditionally, if necessary waiting until it is released by another
  83 thread (only one thread at a time can acquire a lock --- that's their
  84 reason for existence).  If the integer
  85 \var{waitflag} argument is present, the action depends on its
  86 value: if it is zero, the lock is only acquired if it can be acquired
  87 immediately without waiting, while if it is nonzero, the lock is
  88 acquired unconditionally as before.  The
  89 return value is \code{True} if the lock is acquired successfully,
  90 \code{False} if not.
  91 \end{methoddesc}
  92
  93 \begin{methoddesc}[lock]{release}{}
  94 Releases the lock.  The lock must have been acquired earlier, but not
  95 necessarily by the same thread.
  96 \end{methoddesc}
  97
  98 \begin{methoddesc}[lock]{locked}{}
  99 Return the status of the lock:\ \code{True} if it has been acquired by
 100 some thread, \code{False} if not.
 101 \end{methoddesc}
 102
 103 \strong{Caveats:}
 104
 105 \begin{itemize}
 106 \item
 107 Threads interact strangely with interrupts: the
 108 \exception{KeyboardInterrupt} exception will be received by an
 109 arbitrary thread.  (When the \refmodule{signal}\refbimodindex{signal}
 110 module is available, interrupts always go to the main thread.)
 111
 112 \item
 113 Calling \function{sys.exit()} or raising the \exception{SystemExit}
 114 exception is equivalent to calling \function{exit()}.
 115
 116 \item
 117 Not all built-in functions that may block waiting for I/O allow other
 118 threads to run.  (The most popular ones (\function{time.sleep()},
 119 \method{\var{file}.read()}, \function{select.select()}) work as
 120 expected.)
 121
 122 \item
 123 It is not possible to interrupt the \method{acquire()} method on a lock
 124 --- the \exception{KeyboardInterrupt} exception will happen after the
 125 lock has been acquired.
 126
 127 \item
 128 When the main thread exits, it is system defined whether the other
 129 threads survive.  On SGI IRIX using the native thread implementation,
 130 they survive.  On most other systems, they are killed without
 131 executing \keyword{try} ... \keyword{finally} clauses or executing
 132 object destructors.
 133 \indexii{threads}{IRIX}
 134
 135 \item
 136 When the main thread exits, it does not do any of its usual cleanup
 137 (except that \keyword{try} ... \keyword{finally} clauses are honored),
 138 and the standard I/O files are not flushed.
 139
 140 \end{itemize}