Help with matcher patterns

[Dan-Essin]

Hello,
Newcomer question. I am working a tutorial based on spacy v2 but my development platform has version 3.1.3 installed.
The tutorial presents this code:

# Create a pattern matching two tokens: "iPhone" and "X"
pattern = [{"TEXT": "iPhone"}, {"TEXT": "X"}]
# Add the pattern to the matcher
matcher.add("IPHONE_X_PATTERN", None, pattern)

In version 3 that generates errors but this code works:

pattern = [[{"TEXT": "iOS"}, {"IS_DIGIT": True}]]
matcher.add("IPHONE_X_PATTERN", pattern)

My question is:
Since [{"TEXT": "iPhone"}, {"TEXT": "X"}] is a list of dicts
Why does that fail in version 3 but [[{"TEXT": "iOS"}, {"IS_DIGIT": True}]], which is a list of a list of dicts, works?

p.s. obviously version 3 has reduced the number of args in matcher.add from 3 to 2.

Thank you

[kinghuang]

As you noted, the arguments to Matcher.add changed from spaCy 2 to 3. In spaCy 2, the method signature was:

matcher.add(match_id, on_match, *patterns)

Notice the * on *patterns. This is a variable-length argument. So, after the match_id and on_match argument, any remaining arguments to matcher.add(…) would get collected together as a list for patterns.

pattern1 = [{"TEXT": "iPhone", {"TEXT": "X"}]
pattern2 = [{"TEXT": "iOS"}, {"IS_DIGIT": True}]

matcher.add("IPHONE_X_PATTERN", None, pattern1, pattern2)

# The method receives the equivalent of patterns=[pattern1, pattern2],
# from gathering the variable arguments.

In spaCy 3, the method signature changed to:

matcher.add(match_id, patterns, on_match=None, greedy=None)

Putting aside the differences in the order of the arguments, the lack of a * on patterns means that it will no longer collect arbitrary arguments passed to the method. This is why you need to put together all the patterns into a list, and pass that to the method.

pattern1 = [{"TEXT": "iPhone", {"TEXT": "X"}]
pattern2 = [{"TEXT": "iOS"}, {"IS_DIGIT": True}]

matcher.add("IPHONE_X_PATTERN", [pattern1, pattern2])

# The method receives patterns=[pattern1, pattern2] like before,
# but because you've passed the list instead of a variable number of arguments being collected.