PicklingError: parser phases not pickleable

[jdufresne]

Pickling an HTMLParser object fails.

The following test demonstrates the behavior I was expecting:

def test_pickle_parser():
    parser = HTMLParser()
    encoded = pickle.dumps(parser)
    decoded = pickle.loads(encoded)
    ...

This is failing because the *Phase classes are not defined at the top-level, but instead are defined inside the factory function getPhases(). (This pattern exists in the treebuilders directory as well.)

Unless there is a reason to never reuse a parser, I think the object should be pickleable.

[gsnedders]

I'd really like getPhases to go away and have some saner, ideally without the function at all. But it isn't so clear how to do so with logging still working. Would it become pickleable if we just saved a reference at the top level? (I forget the details about pickle frequently!)

It's definitely the case that some trees cannot be pickled, so after parsing you can't necessarily pickle the object, FWIW.