Support extensions - Githubissues

sam-writer commented 4 years ago

Part of #80 is making replaCy support extensions, so this does that.

Involves:

Some code reorg - mainly just trying to move stuff out of __init__.py and into other files
more tests
cleaned up the basic tests, using the testing code we have been using lately, that makes sure matches are for the rule being tested
add a default Scorer class that KenLMScorer extends
I think I fixed a (previously unknown) bug on line 186 of suggestion.py

Here is the thing I might hate about this:

a pipeline component signature is List[Span] -> None, that is, it takes a list of spans, and doesn't return... because it modifies that list
that's not very clean of us, should they return a List[Span]?
see, for example, adding spacy.util.filter_spans as a pipeline component, and the weird things I needed to do

sam-writer commented 4 years ago

@melisa-qordoba Here is how this would be used for #80

sam-writer commented 4 years ago

I installed this version of replaCy in a downstream project (Clarity) and the test suite still passes 💪

writer / replaCy