Pattern matcher regression

[emste]

Hey!
I'm working with the Matcher and experienced changed behaviour between v2.0.0a17 and v2.0.0a19. The pattern operator '*' seems to be less greedy than it was in a17:

import spacy

from spacy.matcher import Matcher

nlp = spacy.load('en_core_web_sm')

def on_match(matcher, doc, i, matches):
    match_id, start, end = matches[i]
    match = doc[start : end].text
    print(f"on_match: {match}")

matcher = Matcher(nlp.vocab)
matcher.add('MyMatcher', on_match,  [ {'ORTH': 'Hello'}, {'ORTH': 'World', 'OP': '*'} ])

doc1 = nlp('Hello World')
doc2 = nlp('Hello World World World')
matcher(doc1)
matcher(doc2)
# Output for a17
# on_match: Hello World
# on_match: Hello World World World
# Output for a19
# on_match: Hello World
# on_match: Hello World

Is this an intended change?

Env a17

• spaCy version: 2.0.0a17
• Platform: Linux-4.9.49-moby-x86_64-with-debian-8.9
• Python version: 3.6.3
• Models: en_core_web_sm, xx_ent_wiki_sm

Env a19

• spaCy version: 2.0.0a19
• Platform: Linux-4.9.49-moby-x86_64-with-debian-8.9
• Python version: 3.6.3
• Models: en_core_web_sm, xx_ent_wiki_sm, de_core_news_sm

[honnibal]

Thanks!

[wpm]

Looks like the same thing is happening with the + qualifier.

[ines]

Should be fixed in #1876!

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.