Read-write lock with only one underlying lock?

问题

I've written a read-write lock using Python's concurrency primitives (I think!). Every implementation I've read on SO or elsewhere seems to use 2 locks -- one for reads, and another for writes. My implementation contains only one monitor for reads, but I may be missing something crucial -- can anyone confirm that this will work? If so, what is the benefit to using an additional write lock?

This is the classic read-write lock with preference for readers (may starve writers). I use a dummy cache to demonstrate the reading and writing.

    import threading as t

    class ReadWriteCache(object):
        def __init__(self):
            self.cache = {}
            self.reads = 0
            self.read_cond = t.Condition(t.Lock())

        def read(self, key):
            with self.read_cond:      # Register the read, so writes will wait()
                self.reads += 1

            result = self.cache[key]

            with self.read_cond:
                self.reads -= 1
                if not self.reads:
                    self.read_cond.notify_all()
                return result

        def update(self, key, value):
            with self.read_cond:
                while self.reads:
                    self.read_cond.wait()   # Wait for reads to clear
                self.cache[key] = value     # With read lock, update value

回答1:

You are not using a single lock.
You are using a lock and a condition variable

self.read_lock = t.Condition(t.Lock())

A condition variable is a concurrency primitive too. A more complex one than a lock.

note : please do not call a condition variable object read_lock

edit: Your code seems correct to me, as it solves the First readers-writers problem. As you said it may starve writer. This is not a small issue. The logic behind reader writer is that there may be a lot more reads than writes
An additional lock allow to solve the Second readers-writers problem, where a writer doesn't starve. Indeed, readers have to wait when there is a writer waiting for the resource.

回答2:

One more solution using a lock and a condition. Takes care of the starvation issue and also supports promotion of a read lock to write lock when requested from the same thread.

# From O'Reilly Python Cookbook by David Ascher, Alex Martelli
# With changes to cover the starvation situation where a continuous
#   stream of readers may starve a writer, Lock Promotion and Context Managers

class ReadWriteLock:
  """ A lock object that allows many simultaneous "read locks", but
  only one "write lock." """

  def __init__(self, withPromotion=False):
    self._read_ready = threading.Condition(threading.RLock(  ))
    self._readers = 0
    self._writers = 0
    self._promote = withPromotion
    self._readerList = []  # List of Reader thread IDs
    self._writerList = []  # List of Writer thread IDs

  def acquire_read(self):
    logging.debug("RWL : acquire_read()")
    """ Acquire a read lock. Blocks only if a thread has
    acquired the write lock. """
    self._read_ready.acquire(  )
    try:
      while self._writers > 0:
        self._read_ready.wait()
      self._readers += 1
    finally:
      self._readerList.append(threading.get_ident())
      self._read_ready.release(  )

  def release_read(self):
    logging.debug("RWL : release_read()")
    """ Release a read lock. """
    self._read_ready.acquire(  )
    try:
      self._readers -= 1
      if not self._readers:
        self._read_ready.notifyAll(  )
    finally:
      self._readerList.remove(threading.get_ident())
      self._read_ready.release(  )

  def acquire_write(self):
    logging.debug("RWL : acquire_write()")
    """ Acquire a write lock. Blocks until there are no
    acquired read or write locks. """
    self._read_ready.acquire(  )   # A re-entrant lock lets a thread re-acquire the lock
    self._writers += 1
    self._writerList.append(threading.get_ident())
    while self._readers > 0:
      # promote to write lock, only if all the readers are trying to promote to writer
      # If there are other reader threads, then wait till they complete reading
      if self._promote and threading.get_ident() in self._readerList and set(self._readerList).issubset(set(self._writerList)):
        break
      else:
        self._read_ready.wait(  )

  def release_write(self):
    logging.debug("RWL : release_write()")
    """ Release a write lock. """
    self._writers -= 1
    self._writerList.remove(threading.get_ident())
    self._read_ready.notifyAll(  )
    self._read_ready.release(  )

#----------------------------------------------------------------------------------------------------------

class ReadRWLock:
  # Context Manager class for ReadWriteLock
  def __init__(self, rwLock):
    self.rwLock = rwLock

  def __enter__(self):
    self.rwLock.acquire_read()
    return self         # Not mandatory, but returning to be safe

  def __exit__(self, exc_type, exc_value, traceback):
    self.rwLock.release_read()
    return False        # Raise the exception, if exited due to an exception

#----------------------------------------------------------------------------------------------------------

class WriteRWLock:
  # Context Manager class for ReadWriteLock
  def __init__(self, rwLock):
    self.rwLock = rwLock

  def __enter__(self):
    self.rwLock.acquire_write()
    return self         # Not mandatory, but returning to be safe

  def __exit__(self, exc_type, exc_value, traceback):
    self.rwLock.release_write()
    return False        # Raise the exception, if exited due to an exception

#----------------------------------------------------------------------------------------------------------

来源：https://stackoverflow.com/questions/38862104/read-write-lock-with-only-one-underlying-lock