Support customizing how built-in types are pickled for cloudpickle

[AdrS]

This is to enable customizing how sets are serialized to increase the pickling determinism. I'm modifying the vendored cloudpickle as a stop-gap measure until the cloudpickle maintainers review cloudpipe/cloudpickle#563.

Note: It's easiest to review this change with the setting to hide whitespace.

Issue: #34410

[github-actions]

Assigning reviewers. If you would like to opt out of this review, comment assign to next reviewer:

R: @jrmccluskey for label python.

Available commands:

• stop reviewer notifications - opt out of the automated review tooling
• remind me after tests pass - tag the comment author after tests pass
• waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

[jrmccluskey]

Passing to @claudevdm who has been working with dill and cloudpickle as of late, should have better context here

[claudevdm]

As discussed offline we probably want to have this default to the fast pickler, and add an option to use slow pickler.

[claudevdm]

Should we vendor the cloudpickle tests as well if we are modifying the implementation this much?

[claudevdm]

There are a bunch of failing tests e.g.

self = <apache_beam.internal.cloudpickle.cloudpickle.PurePythonPickler object at 0x7a8974470940>
obj = b''

    def memoize(self, obj):
        """Store an object in the memo."""
    
        # The Pickler memo is a dictionary mapping object ids to 2-tuples
        # that contain the Unpickler memo key and the object being memoized.
        # The memo key is written to the pickle and will become
        # the key in the Unpickler's memo.  The object is stored in the
        # Pickler memo so that transient objects are kept alive during
        # pickling.
    
        # The use of the Unpickler memo length as the memo key is just a
        # convention.  The only requirement is that the memo values be unique.
        # But there appears no advantage to any other scheme, and this
        # scheme allows the Unpickler memo to be implemented as a plain (but
        # growable) array, indexed by memo key.
        if self.fast:
            return
>       assert id(obj) not in self.memo
E       AssertionError

[github-actions]

Reminder, please take a look at this pr: @jrmccluskey

[github-actions]

Assigning new set of reviewers because Pr has gone too long without review. If you would like to opt out of this review, comment assign to next reviewer:

R: @tvalentyn for label python.

Available commands:

• stop reviewer notifications - opt out of the automated review tooling
• remind me after tests pass - tag the comment author after tests pass
• waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

[tvalentyn]

Should we vendor the cloudpickle tests as well if we are modifying the implementation this much?

+1.

Given cloudpickle authors are not responsive, we might have to maintain this fork until our changes can be upstreamed.

[github-actions]

Reminder, please take a look at this pr: @tvalentyn

[tvalentyn]

waiting on author

[github-actions]

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 1 week if no further activity occurs. If you think that’s incorrect or this pull request requires a review, please simply write any comment. If closed, you can revive the PR at any time and @mention a reviewer or discuss it on the dev@beam.apache.org list. Thank you for your contributions.

[github-actions]

This pull request has been closed due to lack of activity. If you think that is incorrect, or the pull request requires review, you can revive the PR at any time.