大师兄的Python源码学习笔记(四十五）: Python的多线

大师兄的Python源码学习笔记(四十四）: Python的多线程机制(六）
大师兄的Python源码学习笔记(四十六）: Python的内存管理机制(一）

八、threading库

• _thread库和Lock对象是Python提供的低级线程控制工具，而threading库是在_thread的基础上构建的高级线程控制库。

demo.py

>>>import threading
>>>import time

>>>class MyThread(threading.Thread):
>>>    def run(self):
>>>        while True:
>>>            print(f"thread id: {threading.get_ident()}")
>>>            time.sleep(1)

>>>if __name__ == '__main__':
>>>    mythread = MyThread()
>>>    mythread.start()
>>>    while True:
>>>        print(f"thread id:{threading.get_ident()}")
>>>        time.sleep(1)
thread id: 9524
thread id:13284
thread id: 9524
thread id:13284
... ...

1. threading概述

• 由于threading库是在_thread库的基础上构建的，所以暴露了很多_thread中的属性：

threading.py

>>>"""Thread module emulating a subset of Java's threading model."""

>>>import os as _os
>>>import sys as _sys
>>>import _thread

... ...
>>># Rename some stuff so "from threading import *" is safe
>>>_start_new_thread = _thread.start_new_thread
>>>_allocate_lock = _thread.allocate_lock
>>>_set_sentinel = _thread._set_sentinel
>>>get_ident = _thread.get_ident
>>>try:
>>>    get_native_id = _thread.get_native_id
>>>    _HAVE_THREAD_NATIVE_ID = True
>>>    __all__.append('get_native_id')
>>>except AttributeError:
>>>    _HAVE_THREAD_NATIVE_ID = False
>>>ThreadError = _thread.error
>>>try:
>>>    _CRLock = _thread.RLock
>>>except AttributeError:
>>>    _CRLock = None
>>>TIMEOUT_MAX = _thread.TIMEOUT_MAX
>>>del _thread
... ...

• 在threading中，有一套记录当前所有通过继承threading.Thread而创建的Python线程机制，是用一个lock和两个dict完成：

threading.py

... ...
>>># Active thread administration
>>>_active_limbo_lock = _allocate_lock()
>>>_active = {}    # maps thread id to Thread object
>>>_limbo = {}
>>>_dangling = WeakSet()
... ...

• threading.Thread创建多线程分两个阶段:
• 第一阶段：调用threading.Thread.start

• 这时还没有调用_thread.start_new_thread创建原生子线程，线程记录在_limbo中。
• 由于没有创建子线程，所以也没有线程id，记录方式为：_limbo[thread] = thread。

• 第二阶段：调用threading.Thread.run

• 这时已调用_thread.start_new_thread创建原生子线程，并从_limbo中删除子线程，而记录到_active中。
• 记录的方式为_active[thread_id] = thread。

• 所以，_limbo和_active分别维护了等待创建和已创建的子线程集合，且对他们的访问都在_active_limbo_lock的保护下进行。
• 在threading中，提供了列举当前所有子线程的操作：threading.enumerate:

threading.py

>>>def enumerate():
>>>    """Return a list of all Thread objects currently alive.

>>>    The list includes daemonic threads, dummy thread objects created by
>>>    current_thread(), and the main thread. It excludes terminated threads and
>>>    threads that have not yet been started.

>>>    """
>>>    with _active_limbo_lock:
>>>        return list(_active.values()) + list(_limbo.values())

• enumerate实际就是将_limbo和_active维护的线程集合返回。

2. threading的线程同步工具

threading.py

... ...
>>>import _thread

... ...
>>>_allocate_lock = _thread.allocate_lock
... ...
>>>Lock = _allocate_lock
... ...

• 从源码中可以看出，threading对_thread中的Lock对象没做任何包装，直接将其展示出来，通过调用threading.Lock就可以创建一个Lock对象。
• 在这个Lock对象的基础上，就可以进行acquire、release等操作。
• threading的线程同步工具实际就是建立在这个Lock对象的基础上的。

2.1 RLock

threading.py

class _RLock:
    """This class implements reentrant lock objects.

    A reentrant lock must be released by the thread that acquired it. Once a
    thread has acquired a reentrant lock, the same thread may acquire it
    again without blocking; the thread must release it once for each time it
    has acquired it.

    """

    def __init__(self):
        self._block = _allocate_lock()
        self._owner = None
        self._count = 0

    def __repr__(self):
        owner = self._owner
        try:
            owner = _active[owner].name
        except KeyError:
            pass
        return "<%s %s.%s object owner=%r count=%d at %s>" % (
            "locked" if self._block.locked() else "unlocked",
            self.__class__.__module__,
            self.__class__.__qualname__,
            owner,
            self._count,
            hex(id(self))
        )

    def acquire(self, blocking=True, timeout=-1):
        """Acquire a lock, blocking or non-blocking.

        When invoked without arguments: if this thread already owns the lock,
        increment the recursion level by one, and return immediately. Otherwise,
        if another thread owns the lock, block until the lock is unlocked. Once
        the lock is unlocked (not owned by any thread), then grab ownership, set
        the recursion level to one, and return. If more than one thread is
        blocked waiting until the lock is unlocked, only one at a time will be
        able to grab ownership of the lock. There is no return value in this
        case.

        When invoked with the blocking argument set to true, do the same thing
        as when called without arguments, and return true.

        When invoked with the blocking argument set to false, do not block. If a
        call without an argument would block, return false immediately;
        otherwise, do the same thing as when called without arguments, and
        return true.

        When invoked with the floating-point timeout argument set to a positive
        value, block for at most the number of seconds specified by timeout
        and as long as the lock cannot be acquired.  Return true if the lock has
        been acquired, false if the timeout has elapsed.

        """
        me = get_ident()
        if self._owner == me:
            self._count += 1
            return 1
        rc = self._block.acquire(blocking, timeout)
        if rc:
            self._owner = me
            self._count = 1
        return rc

    __enter__ = acquire

    def release(self):
        """Release a lock, decrementing the recursion level.

        If after the decrement it is zero, reset the lock to unlocked (not owned
        by any thread), and if any other threads are blocked waiting for the
        lock to become unlocked, allow exactly one of them to proceed. If after
        the decrement the recursion level is still nonzero, the lock remains
        locked and owned by the calling thread.

        Only call this method when the calling thread owns the lock. A
        RuntimeError is raised if this method is called when the lock is
        unlocked.

        There is no return value.

        """
        if self._owner != get_ident():
            raise RuntimeError("cannot release un-acquired lock")
        self._count = count = self._count - 1
        if not count:
            self._owner = None
            self._block.release()

    def __exit__(self, t, v, tb):
        self.release()

    # Internal methods used by condition variables

    def _acquire_restore(self, state):
        self._block.acquire()
        self._count, self._owner = state

    def _release_save(self):
        if self._count == 0:
            raise RuntimeError("cannot release un-acquired lock")
        count = self._count
        self._count = 0
        owner = self._owner
        self._owner = None
        self._block.release()
        return (count, owner)

    def _is_owned(self):
        return self._owner == get_ident()

• RLock对象是Lock对象的变种，其内部维护着一个Lock对象，但他是一种可以重入的锁。
• 对于Lock对象而言，如果一个线程连续两次进行acquire操作，那么由于第一次acquire之后没有release，第二次acquire时将被挂起进程，导致Lock对象永远不会release，形成死锁。
• RLock对象允许同一个线程多次对其进行acquire，因为其内部通过一个counter变量维护线程的acquire次数。
• 所以每一次acquire必须对应一次release，在所有release操作完成后，别的线程才能申请该RLock对象。

2.2 Condition

threading.py

class Condition:
    """Class that implements a condition variable.

    A condition variable allows one or more threads to wait until they are
    notified by another thread.

    If the lock argument is given and not None, it must be a Lock or RLock
    object, and it is used as the underlying lock. Otherwise, a new RLock object
    is created and used as the underlying lock.

    """

    def __init__(self, lock=None):
        if lock is None:
            lock = RLock()
        self._lock = lock
        # Export the lock's acquire() and release() methods
        self.acquire = lock.acquire
        self.release = lock.release
        # If the lock defines _release_save() and/or _acquire_restore(),
        # these override the default implementations (which just call
        # release() and acquire() on the lock).  Ditto for _is_owned().
        try:
            self._release_save = lock._release_save
        except AttributeError:
            pass
        try:
            self._acquire_restore = lock._acquire_restore
        except AttributeError:
            pass
        try:
            self._is_owned = lock._is_owned
        except AttributeError:
            pass
        self._waiters = _deque()

    def __enter__(self):
        return self._lock.__enter__()

    def __exit__(self, *args):
        return self._lock.__exit__(*args)

    def __repr__(self):
        return "<Condition(%s, %d)>" % (self._lock, len(self._waiters))

    def _release_save(self):
        self._lock.release()           # No state to save

    def _acquire_restore(self, x):
        self._lock.acquire()           # Ignore saved state

    def _is_owned(self):
        # Return True if lock is owned by current_thread.
        # This method is called only if _lock doesn't have _is_owned().
        if self._lock.acquire(0):
            self._lock.release()
            return False
        else:
            return True

    def wait(self, timeout=None):
        """Wait until notified or until a timeout occurs.

        If the calling thread has not acquired the lock when this method is
        called, a RuntimeError is raised.

        This method releases the underlying lock, and then blocks until it is
        awakened by a notify() or notify_all() call for the same condition
        variable in another thread, or until the optional timeout occurs. Once
        awakened or timed out, it re-acquires the lock and returns.

        When the timeout argument is present and not None, it should be a
        floating point number specifying a timeout for the operation in seconds
        (or fractions thereof).

        When the underlying lock is an RLock, it is not released using its
        release() method, since this may not actually unlock the lock when it
        was acquired multiple times recursively. Instead, an internal interface
        of the RLock class is used, which really unlocks it even when it has
        been recursively acquired several times. Another internal interface is
        then used to restore the recursion level when the lock is reacquired.

        """
        if not self._is_owned():
            raise RuntimeError("cannot wait on un-acquired lock")
        waiter = _allocate_lock()
        waiter.acquire()
        self._waiters.append(waiter)
        saved_state = self._release_save()
        gotit = False
        try:    # restore state no matter what (e.g., KeyboardInterrupt)
            if timeout is None:
                waiter.acquire()
                gotit = True
            else:
                if timeout > 0:
                    gotit = waiter.acquire(True, timeout)
                else:
                    gotit = waiter.acquire(False)
            return gotit
        finally:
            self._acquire_restore(saved_state)
            if not gotit:
                try:
                    self._waiters.remove(waiter)
                except ValueError:
                    pass

    def wait_for(self, predicate, timeout=None):
        """Wait until a condition evaluates to True.

        predicate should be a callable which result will be interpreted as a
        boolean value.  A timeout may be provided giving the maximum time to
        wait.

        """
        endtime = None
        waittime = timeout
        result = predicate()
        while not result:
            if waittime is not None:
                if endtime is None:
                    endtime = _time() + waittime
                else:
                    waittime = endtime - _time()
                    if waittime <= 0:
                        break
            self.wait(waittime)
            result = predicate()
        return result

    def notify(self, n=1):
        """Wake up one or more threads waiting on this condition, if any.

        If the calling thread has not acquired the lock when this method is
        called, a RuntimeError is raised.

        This method wakes up at most n of the threads waiting for the condition
        variable; it is a no-op if no threads are waiting.

        """
        if not self._is_owned():
            raise RuntimeError("cannot notify on un-acquired lock")
        all_waiters = self._waiters
        waiters_to_notify = _deque(_islice(all_waiters, n))
        if not waiters_to_notify:
            return
        for waiter in waiters_to_notify:
            waiter.release()
            try:
                all_waiters.remove(waiter)
            except ValueError:
                pass

    def notify_all(self):
        """Wake up all threads waiting on this condition.

        If the calling thread has not acquired the lock when this method
        is called, a RuntimeError is raised.

        """
        self.notify(len(self._waiters))

    notifyAll = notify_all

• Condition对象是对Lock对象的包装，在创建Condition对象时，其构造函数需要一个Lock对象作为参数，如果没有传入Lock对象，Condition对象将在其内部自行创建一个RLock对象。
• Condition对象的价值在于其提供了wait和notify的语义，这两条语义有利于线程通信和解决复杂线程的同步问题。
• 假设有Condition对象c，当线程A调用c.wait()时，线程A将释放c中的Lock对象，并进入阻塞状态，直到有别的线程调用c.notify()，A才会重新通过acquire申请c中的Lock对象，并退出wait。

2.3 Semaphore

threading.py

class Semaphore:
    """This class implements semaphore objects.

    Semaphores manage a counter representing the number of release() calls minus
    the number of acquire() calls, plus an initial value. The acquire() method
    blocks if necessary until it can return without making the counter
    negative. If not given, value defaults to 1.

    """

    # After Tim Peters' semaphore class, but not quite the same (no maximum)

    def __init__(self, value=1):
        if value < 0:
            raise ValueError("semaphore initial value must be >= 0")
        self._cond = Condition(Lock())
        self._value = value

    def acquire(self, blocking=True, timeout=None):
        """Acquire a semaphore, decrementing the internal counter by one.

        When invoked without arguments: if the internal counter is larger than
        zero on entry, decrement it by one and return immediately. If it is zero
        on entry, block, waiting until some other thread has called release() to
        make it larger than zero. This is done with proper interlocking so that
        if multiple acquire() calls are blocked, release() will wake exactly one
        of them up. The implementation may pick one at random, so the order in
        which blocked threads are awakened should not be relied on. There is no
        return value in this case.

        When invoked with blocking set to true, do the same thing as when called
        without arguments, and return true.

        When invoked with blocking set to false, do not block. If a call without
        an argument would block, return false immediately; otherwise, do the
        same thing as when called without arguments, and return true.

        When invoked with a timeout other than None, it will block for at
        most timeout seconds.  If acquire does not complete successfully in
        that interval, return false.  Return true otherwise.

        """
        if not blocking and timeout is not None:
            raise ValueError("can't specify timeout for non-blocking acquire")
        rc = False
        endtime = None
        with self._cond:
            while self._value == 0:
                if not blocking:
                    break
                if timeout is not None:
                    if endtime is None:
                        endtime = _time() + timeout
                    else:
                        timeout = endtime - _time()
                        if timeout <= 0:
                            break
                self._cond.wait(timeout)
            else:
                self._value -= 1
                rc = True
        return rc

    __enter__ = acquire

    def release(self):
        """Release a semaphore, incrementing the internal counter by one.

        When the counter is zero on entry and another thread is waiting for it
        to become larger than zero again, wake up that thread.

        """
        with self._cond:
            self._value += 1
            self._cond.notify()

    def __exit__(self, t, v, tb):
        self.release()

• Semaphore对象内部维护着一个Condition对象，并在Condition对象的基础上实现了对共享资源池进行保护的线程同步机制。
• 可以理解Semaphore对象是一把锁多把钥匙。

2.4 BoundedSemaphore

threading.py

class BoundedSemaphore(Semaphore):
    """Implements a bounded semaphore.

    A bounded semaphore checks to make sure its current value doesn't exceed its
    initial value. If it does, ValueError is raised. In most situations
    semaphores are used to guard resources with limited capacity.

    If the semaphore is released too many times it's a sign of a bug. If not
    given, value defaults to 1.

    Like regular semaphores, bounded semaphores manage a counter representing
    the number of release() calls minus the number of acquire() calls, plus an
    initial value. The acquire() method blocks if necessary until it can return
    without making the counter negative. If not given, value defaults to 1.

    """

    def __init__(self, value=1):
        Semaphore.__init__(self, value)
        self._initial_value = value

    def release(self):
        """Release a semaphore, incrementing the internal counter by one.

        When the counter is zero on entry and another thread is waiting for it
        to become larger than zero again, wake up that thread.

        If the number of releases exceeds the number of acquires,
        raise a ValueError.

        """
        with self._cond:
            if self._value >= self._initial_value:
                raise ValueError("Semaphore released too many times")
            self._value += 1
            self._cond.notify()

• BoundedSemaphore对象与BoundedSemaphore对象类似，但是他限制了钥匙的数量。

2.5 Event

threading.py

class Event:
    """Class implementing event objects.

    Events manage a flag that can be set to true with the set() method and reset
    to false with the clear() method. The wait() method blocks until the flag is
    true.  The flag is initially false.

    """

    # After Tim Peters' event class (without is_posted())

    def __init__(self):
        self._cond = Condition(Lock())
        self._flag = False

    def _reset_internal_locks(self):
        # private!  called by Thread._reset_internal_locks by _after_fork()
        self._cond.__init__(Lock())

    def is_set(self):
        """Return true if and only if the internal flag is true."""
        return self._flag

    isSet = is_set

    def set(self):
        """Set the internal flag to true.

        All threads waiting for it to become true are awakened. Threads
        that call wait() once the flag is true will not block at all.

        """
        with self._cond:
            self._flag = True
            self._cond.notify_all()

    def clear(self):
        """Reset the internal flag to false.

        Subsequently, threads calling wait() will block until set() is called to
        set the internal flag to true again.

        """
        with self._cond:
            self._flag = False

    def wait(self, timeout=None):
        """Block until the internal flag is true.

        If the internal flag is true on entry, return immediately. Otherwise,
        block until another thread calls set() to set the flag to true, or until
        the optional timeout occurs.

        When the timeout argument is present and not None, it should be a
        floating point number specifying a timeout for the operation in seconds
        (or fractions thereof).

        This method returns the internal flag on exit, so it will always return
        True except if a timeout is given and the operation times out.

        """
        with self._cond:
            signaled = self._flag
            if not signaled:
                signaled = self._cond.wait(timeout)
            return signaled

• Event对象与Semaphore对象类似，提供了set和wait语义，由主一个线程发送信号，其他的线程等待信号，用于主线程控制其他线程。

2.6 Barrier

threading.py

class Barrier:
    """Implements a Barrier.

    Useful for synchronizing a fixed number of threads at known synchronization
    points.  Threads block on 'wait()' and are simultaneously awoken once they
    have all made that call.

    """

    def __init__(self, parties, action=None, timeout=None):
        """Create a barrier, initialised to 'parties' threads.

        'action' is a callable which, when supplied, will be called by one of
        the threads after they have all entered the barrier and just prior to
        releasing them all. If a 'timeout' is provided, it is used as the
        default for all subsequent 'wait()' calls.

        """
        self._cond = Condition(Lock())
        self._action = action
        self._timeout = timeout
        self._parties = parties
        self._state = 0 #0 filling, 1, draining, -1 resetting, -2 broken
        self._count = 0

    def wait(self, timeout=None):
        """Wait for the barrier.

        When the specified number of threads have started waiting, they are all
        simultaneously awoken. If an 'action' was provided for the barrier, one
        of the threads will have executed that callback prior to returning.
        Returns an individual index number from 0 to 'parties-1'.

        """
        if timeout is None:
            timeout = self._timeout
        with self._cond:
            self._enter() # Block while the barrier drains.
            index = self._count
            self._count += 1
            try:
                if index + 1 == self._parties:
                    # We release the barrier
                    self._release()
                else:
                    # We wait until someone releases us
                    self._wait(timeout)
                return index
            finally:
                self._count -= 1
                # Wake up any threads waiting for barrier to drain.
                self._exit()

    # Block until the barrier is ready for us, or raise an exception
    # if it is broken.
    def _enter(self):
        while self._state in (-1, 1):
            # It is draining or resetting, wait until done
            self._cond.wait()
        #see if the barrier is in a broken state
        if self._state < 0:
            raise BrokenBarrierError
        assert self._state == 0

    # Optionally run the 'action' and release the threads waiting
    # in the barrier.
    def _release(self):
        try:
            if self._action:
                self._action()
            # enter draining state
            self._state = 1
            self._cond.notify_all()
        except:
            #an exception during the _action handler.  Break and reraise
            self._break()
            raise

    # Wait in the barrier until we are released.  Raise an exception
    # if the barrier is reset or broken.
    def _wait(self, timeout):
        if not self._cond.wait_for(lambda : self._state != 0, timeout):
            #timed out.  Break the barrier
            self._break()
            raise BrokenBarrierError
        if self._state < 0:
            raise BrokenBarrierError
        assert self._state == 1

    # If we are the last thread to exit the barrier, signal any threads
    # waiting for the barrier to drain.
    def _exit(self):
        if self._count == 0:
            if self._state in (-1, 1):
                #resetting or draining
                self._state = 0
                self._cond.notify_all()

    def reset(self):
        """Reset the barrier to the initial state.

        Any threads currently waiting will get the BrokenBarrier exception
        raised.

        """
        with self._cond:
            if self._count > 0:
                if self._state == 0:
                    #reset the barrier, waking up threads
                    self._state = -1
                elif self._state == -2:
                    #was broken, set it to reset state
                    #which clears when the last thread exits
                    self._state = -1
            else:
                self._state = 0
            self._cond.notify_all()

    def abort(self):
        """Place the barrier into a 'broken' state.

        Useful in case of error.  Any currently waiting threads and threads
        attempting to 'wait()' will have BrokenBarrierError raised.

        """
        with self._cond:
            self._break()

    def _break(self):
        # An internal error was detected.  The barrier is set to
        # a broken state all parties awakened.
        self._state = -2
        self._cond.notify_all()

    @property
    def parties(self):
        """Return the number of threads required to trip the barrier."""
        return self._parties

    @property
    def n_waiting(self):
        """Return the number of threads currently waiting at the barrier."""
        # We don't need synchronization here since this is an ephemeral result
        # anyway.  It returns the correct value in the steady state.
        if self._state == 0:
            return self._count
        return 0

    @property
    def broken(self):
        """Return True if the barrier is in a broken state."""
        return self._state == -2

• Barrier对象内部维护了一个Condition对象，他会设置一个线程障碍数量parties，当等待的线程到达了parties时唤醒所有的等待线程，在这之前会一直处于阻塞状态。
• 这对于需要等待多个线程结束后，再同时执行的任务非常有用。

3. threading中的Thread

• Thread是threading中的关键组件：

threading.py

class Thread:
    """A class that represents a thread of control.

    This class can be safely subclassed in a limited fashion. There are two ways
    to specify the activity: by passing a callable object to the constructor, or
    by overriding the run() method in a subclass.

    """

    _initialized = False

    def __init__(self, group=None, target=None, name=None,
                 args=(), kwargs=None, *, daemon=None):
        """This constructor should always be called with keyword arguments. Arguments are:

        *group* should be None; reserved for future extension when a ThreadGroup
        class is implemented.

        *target* is the callable object to be invoked by the run()
        method. Defaults to None, meaning nothing is called.

        *name* is the thread name. By default, a unique name is constructed of
        the form "Thread-N" where N is a small decimal number.

        *args* is the argument tuple for the target invocation. Defaults to ().

        *kwargs* is a dictionary of keyword arguments for the target
        invocation. Defaults to {}.

        If a subclass overrides the constructor, it must make sure to invoke
        the base class constructor (Thread.__init__()) before doing anything
        else to the thread.

        """
        assert group is None, "group argument must be None for now"
        if kwargs is None:
            kwargs = {}
        self._target = target
        self._name = str(name or _newname())
        self._args = args
        self._kwargs = kwargs
        if daemon is not None:
            self._daemonic = daemon
        else:
            self._daemonic = current_thread().daemon
        self._ident = None
        if _HAVE_THREAD_NATIVE_ID:
            self._native_id = None
        self._tstate_lock = None
        self._started = Event()
        self._is_stopped = False
        self._initialized = True
        # Copy of sys.stderr used by self._invoke_excepthook()
        self._stderr = _sys.stderr
        self._invoke_excepthook = _make_invoke_excepthook()
        # For debugging and _after_fork()
        _dangling.add(self)

    def _reset_internal_locks(self, is_alive):
        # private!  Called by _after_fork() to reset our internal locks as
        # they may be in an invalid state leading to a deadlock or crash.
        self._started._reset_internal_locks()
        if is_alive:
            self._set_tstate_lock()
        else:
            # The thread isn't alive after fork: it doesn't have a tstate
            # anymore.
            self._is_stopped = True
            self._tstate_lock = None

    def __repr__(self):
        assert self._initialized, "Thread.__init__() was not called"
        status = "initial"
        if self._started.is_set():
            status = "started"
        self.is_alive() # easy way to get ._is_stopped set when appropriate
        if self._is_stopped:
            status = "stopped"
        if self._daemonic:
            status += " daemon"
        if self._ident is not None:
            status += " %s" % self._ident
        return "<%s(%s, %s)>" % (self.__class__.__name__, self._name, status)

    def start(self):
        """Start the thread's activity.

        It must be called at most once per thread object. It arranges for the
        object's run() method to be invoked in a separate thread of control.

        This method will raise a RuntimeError if called more than once on the
        same thread object.

        """
        if not self._initialized:
            raise RuntimeError("thread.__init__() not called")

        if self._started.is_set():
            raise RuntimeError("threads can only be started once")
        with _active_limbo_lock:
            _limbo[self] = self
        try:
            _start_new_thread(self._bootstrap, ())
        except Exception:
            with _active_limbo_lock:
                del _limbo[self]
            raise
        self._started.wait()

    def run(self):
        """Method representing the thread's activity.

        You may override this method in a subclass. The standard run() method
        invokes the callable object passed to the object's constructor as the
        target argument, if any, with sequential and keyword arguments taken
        from the args and kwargs arguments, respectively.

        """
        try:
            if self._target:
                self._target(*self._args, **self._kwargs)
        finally:
            # Avoid a refcycle if the thread is running a function with
            # an argument that has a member that points to the thread.
            del self._target, self._args, self._kwargs

    def _bootstrap(self):
        # Wrapper around the real bootstrap code that ignores
        # exceptions during interpreter cleanup.  Those typically
        # happen when a daemon thread wakes up at an unfortunate
        # moment, finds the world around it destroyed, and raises some
        # random exception *** while trying to report the exception in
        # _bootstrap_inner() below ***.  Those random exceptions
        # don't help anybody, and they confuse users, so we suppress
        # them.  We suppress them only when it appears that the world
        # indeed has already been destroyed, so that exceptions in
        # _bootstrap_inner() during normal business hours are properly
        # reported.  Also, we only suppress them for daemonic threads;
        # if a non-daemonic encounters this, something else is wrong.
        try:
            self._bootstrap_inner()
        except:
            if self._daemonic and _sys is None:
                return
            raise

    def _set_ident(self):
        self._ident = get_ident()

    if _HAVE_THREAD_NATIVE_ID:
        def _set_native_id(self):
            self._native_id = get_native_id()

    def _set_tstate_lock(self):
        """
        Set a lock object which will be released by the interpreter when
        the underlying thread state (see pystate.h) gets deleted.
        """
        self._tstate_lock = _set_sentinel()
        self._tstate_lock.acquire()

        if not self.daemon:
            with _shutdown_locks_lock:
                _shutdown_locks.add(self._tstate_lock)

    def _bootstrap_inner(self):
        try:
            self._set_ident()
            self._set_tstate_lock()
            if _HAVE_THREAD_NATIVE_ID:
                self._set_native_id()
            self._started.set()
            with _active_limbo_lock:
                _active[self._ident] = self
                del _limbo[self]

            if _trace_hook:
                _sys.settrace(_trace_hook)
            if _profile_hook:
                _sys.setprofile(_profile_hook)

            try:
                self.run()
            except:
                self._invoke_excepthook(self)
        finally:
            with _active_limbo_lock:
                try:
                    # We don't call self._delete() because it also
                    # grabs _active_limbo_lock.
                    del _active[get_ident()]
                except:
                    pass

    def _stop(self):
        # After calling ._stop(), .is_alive() returns False and .join() returns
        # immediately.  ._tstate_lock must be released before calling ._stop().
        #
        # Normal case:  C code at the end of the thread's life
        # (release_sentinel in _threadmodule.c) releases ._tstate_lock, and
        # that's detected by our ._wait_for_tstate_lock(), called by .join()
        # and .is_alive().  Any number of threads _may_ call ._stop()
        # simultaneously (for example, if multiple threads are blocked in
        # .join() calls), and they're not serialized.  That's harmless -
        # they'll just make redundant rebindings of ._is_stopped and
        # ._tstate_lock.  Obscure:  we rebind ._tstate_lock last so that the
        # "assert self._is_stopped" in ._wait_for_tstate_lock() always works
        # (the assert is executed only if ._tstate_lock is None).
        #
        # Special case:  _main_thread releases ._tstate_lock via this
        # module's _shutdown() function.
        lock = self._tstate_lock
        if lock is not None:
            assert not lock.locked()
        self._is_stopped = True
        self._tstate_lock = None
        if not self.daemon:
            with _shutdown_locks_lock:
                _shutdown_locks.discard(lock)

    def _delete(self):
        "Remove current thread from the dict of currently running threads."
        with _active_limbo_lock:
            del _active[get_ident()]
            # There must not be any python code between the previous line
            # and after the lock is released.  Otherwise a tracing function
            # could try to acquire the lock again in the same thread, (in
            # current_thread()), and would block.

    def join(self, timeout=None):
        """Wait until the thread terminates.

        This blocks the calling thread until the thread whose join() method is
        called terminates -- either normally or through an unhandled exception
        or until the optional timeout occurs.

        When the timeout argument is present and not None, it should be a
        floating point number specifying a timeout for the operation in seconds
        (or fractions thereof). As join() always returns None, you must call
        is_alive() after join() to decide whether a timeout happened -- if the
        thread is still alive, the join() call timed out.

        When the timeout argument is not present or None, the operation will
        block until the thread terminates.

        A thread can be join()ed many times.

        join() raises a RuntimeError if an attempt is made to join the current
        thread as that would cause a deadlock. It is also an error to join() a
        thread before it has been started and attempts to do so raises the same
        exception.

        """
        if not self._initialized:
            raise RuntimeError("Thread.__init__() not called")
        if not self._started.is_set():
            raise RuntimeError("cannot join thread before it is started")
        if self is current_thread():
            raise RuntimeError("cannot join current thread")

        if timeout is None:
            self._wait_for_tstate_lock()
        else:
            # the behavior of a negative timeout isn't documented, but
            # historically .join(timeout=x) for x<0 has acted as if timeout=0
            self._wait_for_tstate_lock(timeout=max(timeout, 0))

    def _wait_for_tstate_lock(self, block=True, timeout=-1):
        # Issue #18808: wait for the thread state to be gone.
        # At the end of the thread's life, after all knowledge of the thread
        # is removed from C data structures, C code releases our _tstate_lock.
        # This method passes its arguments to _tstate_lock.acquire().
        # If the lock is acquired, the C code is done, and self._stop() is
        # called.  That sets ._is_stopped to True, and ._tstate_lock to None.
        lock = self._tstate_lock
        if lock is None:  # already determined that the C code is done
            assert self._is_stopped
        elif lock.acquire(block, timeout):
            lock.release()
            self._stop()

    @property
    def name(self):
        """A string used for identification purposes only.

        It has no semantics. Multiple threads may be given the same name. The
        initial name is set by the constructor.

        """
        assert self._initialized, "Thread.__init__() not called"
        return self._name

    @name.setter
    def name(self, name):
        assert self._initialized, "Thread.__init__() not called"
        self._name = str(name)

    @property
    def ident(self):
        """Thread identifier of this thread or None if it has not been started.

        This is a nonzero integer. See the get_ident() function. Thread
        identifiers may be recycled when a thread exits and another thread is
        created. The identifier is available even after the thread has exited.

        """
        assert self._initialized, "Thread.__init__() not called"
        return self._ident

    if _HAVE_THREAD_NATIVE_ID:
        @property
        def native_id(self):
            """Native integral thread ID of this thread, or None if it has not been started.

            This is a non-negative integer. See the get_native_id() function.
            This represents the Thread ID as reported by the kernel.

            """
            assert self._initialized, "Thread.__init__() not called"
            return self._native_id

    def is_alive(self):
        """Return whether the thread is alive.

        This method returns True just before the run() method starts until just
        after the run() method terminates. The module function enumerate()
        returns a list of all alive threads.

        """
        assert self._initialized, "Thread.__init__() not called"
        if self._is_stopped or not self._started.is_set():
            return False
        self._wait_for_tstate_lock(False)
        return not self._is_stopped

    def isAlive(self):
        """Return whether the thread is alive.

        This method is deprecated, use is_alive() instead.
        """
        import warnings
        warnings.warn('isAlive() is deprecated, use is_alive() instead',
                      DeprecationWarning, stacklevel=2)
        return self.is_alive()

    @property
    def daemon(self):
        """A boolean value indicating whether this thread is a daemon thread.

        This must be set before start() is called, otherwise RuntimeError is
        raised. Its initial value is inherited from the creating thread; the
        main thread is not a daemon thread and therefore all threads created in
        the main thread default to daemon = False.

        The entire Python program exits when only daemon threads are left.

        """
        assert self._initialized, "Thread.__init__() not called"
        return self._daemonic

    @daemon.setter
    def daemon(self, daemonic):
        if not self._initialized:
            raise RuntimeError("Thread.__init__() not called")
        if self._started.is_set():
            raise RuntimeError("cannot set daemon status of active thread")
        self._daemonic = daemonic

    def isDaemon(self):
        return self.daemon

    def setDaemon(self, daemonic):
        self.daemon = daemonic

    def getName(self):
        return self.name

    def setName(self, name):
        self.name = name

• 在threading.Thread中，维护着一个Event对象：_started:

threading.py 

    def __init__(self, group=None, target=None, name=None,
                 args=(), kwargs=None, *, daemon=None):
        ... ...
        self._tstate_lock = None
        self._started = Event()
        self._is_stopped = False
        ... ...

• 在源码中可以看到，在调用threading.Thread.start时，会在_limbo中记录线程，然后通过_thread.start_new_thread创建原生线程，并将Event对象阻塞。线程过程为_bootstrap。

threading.py

    def start(self):
       ... ...
        with _active_limbo_lock:
            _limbo[self] = self
        try:
            _start_new_thread(self._bootstrap, ())
        except Exception:
            with _active_limbo_lock:
                del _limbo[self]
            raise
        self._started.wait()

• 在_bootstrap中，会获取_tstate_lock锁，并从_limbo中删除线程记录，转而存储到_active中。

threading.py

    def _bootstrap(self):
        ... ...
        try:
            self._bootstrap_inner()
        except:
            if self._daemonic and _sys is None:
                return
            raise

threading.py 

    def _bootstrap_inner(self):
        try:
            self._set_ident()
            self._set_tstate_lock()
            if _HAVE_THREAD_NATIVE_ID:
                self._set_native_id()
            self._started.set()
            with _active_limbo_lock:
                _active[self._ident] = self
                del _limbo[self]

            if _trace_hook:
                _sys.settrace(_trace_hook)
            if _profile_hook:
                _sys.setprofile(_profile_hook)

            try:
                self.run()
            except:
                self._invoke_excepthook(self)
        finally:
            with _active_limbo_lock:
                try:
                    # We don't call self._delete() because it also
                    # grabs _active_limbo_lock.
                    del _active[get_ident()]
                except:
                    pass

• 然后调用run，实现用户自定义线程过程。

threading.py 

    def run(self):
       ... ...
        try:
            if self._target:
                self._target(*self._args, **self._kwargs)
        finally:
            # Avoid a refcycle if the thread is running a function with
            # an argument that has a member that points to the thread.
            del self._target, self._args, self._kwargs

• 在run结束后，C源码会释放._tstate_lock锁，并被_wait_for_tstate_lock发现，接着线程同步调用_stop操作，结束线程。

    def _wait_for_tstate_lock(self, block=True, timeout=-1):
        ... ...
        lock = self._tstate_lock
        if lock is None:  # already determined that the C code is done
            assert self._is_stopped
        elif lock.acquire(block, timeout):
            lock.release()
            self._stop()

threading.py 

    def _stop(self):
        ... ...
        lock = self._tstate_lock
        if lock is not None:
            assert not lock.locked()
        self._is_stopped = True
        self._tstate_lock = None
        if not self.daemon:
            with _shutdown_locks_lock:
                _shutdown_locks.discard(lock)