Rules Based Matcher with OP "!" and "?"

[galtay-tempus]

How to reproduce the behaviour

I think this is a bug, but would be interested in hearing feedback.

nlp = spacy.load("en_core_web_sm")

matcher = Matcher(nlp.vocab)
patterns = [[{"TEXT": "token1", "OP": "!"}, {"TEXT": "token2", "OP": "?"}, {"TEXT": "token3"}]]
matcher.add("TEST", patterns)

doc = nlp("token1 token2 token3")
matches = matcher(doc, with_alignments=True)
print(matches)
[(1046765068364475028, 1, 3, [0, 2])]

doc = nlp("token1 token3")
matches = matcher(doc, with_alignments=True)
print(matches)
[]

My expectations are that the "match 0 times" token with OP "!" would cause the first example to produce no matches.

I also think maybe this is a bug b/c indexing into the doc with start = match[1] = 1 and end=match[2] = 3 raises an IndexError.

EDIT:
sorry, my confusion, doc[1:3] works fine in the first example.

Thanks for the great library!

Your Environment

• spaCy version: 3.2.0
• Platform: Linux-4.14.248-189.473.amzn2.x86_64-x86_64-with-glibc2.17
• Python version: 3.8.12
• Pipelines: en_core_web_sm (3.2.0)

[adrianeboyd]

Yes, it's the same as array slice notation, so doc[1:3] is correct, where the end index is exclusive.

[polm]

Also, to clarify:

My expectations are that the "match 0 times" token with OP "!" would cause the first example to produce no matches.

This is not correct, the output is as expected. You can remove token1 from the text and it will still work, because token2 is not token1 (matches ! rule), token2 is present zero times after that (matches ? rule), and token3 is next (matches final rule).

[galtay-tempus]

thanks! yes, this makes it clear, i was not interpreting the pattern i wrote correctly.