fix: stop HyponymDetector mutating the global BASE_PATTERNS

[Chessing234]

Bug

HyponymDetector.__init__ does:

self.patterns = BASE_PATTERNS
if extended:
    self.patterns.extend(EXTENDED_PATTERNS)

self.patterns = BASE_PATTERNS aliases the module-global list, so .extend() mutates it in place. Every extended=True detector permanently appends EXTENDED_PATTERNS to the global BASE_PATTERNS (and duplicates them on each later instance); a subsequent extended=False detector then wrongly inherits the extended patterns.

Repro

base = len(BASE_PATTERNS)
HyponymDetector(nlp, extended=True); HyponymDetector(nlp, extended=True)
len(BASE_PATTERNS)  # before: base + 2*len(EXTENDED_PATTERNS); after: base

Fix

Copy the list before extending: self.patterns = list(BASE_PATTERNS).

Made with Cursor