diff --git a/.jules/bolt.md b/.jules/bolt.md new file mode 100644 index 0000000..3bd58a5 --- /dev/null +++ b/.jules/bolt.md @@ -0,0 +1,3 @@ +## 2024-05-24 - Memoize deterministic regex compilations and lexicon generation in text-processing pipelines +**Learning:** In NLP or text-processing pipelines (e.g., within `openmed.clinical`), deterministic operations like regex compilations (`re.compile`) and lexicon generation are extremely expensive if called repeatedly per string evaluation. Without memoization, functions like `_compiled_context_lexicon` cause severe performance bottlenecks during span evaluations. +**Action:** Always memoize deterministic regex compilations and lexicon generation (e.g., using `@functools.lru_cache`) when they are repeatedly called per string or span evaluation, provided the return type is immutable (e.g., a frozen dataclass). diff --git a/openmed/openmed/clinical/context.py b/openmed/openmed/clinical/context.py index 9fd11df..e78f5ad 100644 --- a/openmed/openmed/clinical/context.py +++ b/openmed/openmed/clinical/context.py @@ -35,6 +35,7 @@ from __future__ import annotations +import functools import re from collections.abc import Iterable, Iterator, Mapping, Sequence from dataclasses import dataclass, replace @@ -154,6 +155,7 @@ class _CompiledContextLexicon: backward_context_cues: frozenset[str] +@functools.lru_cache def _compiled_context_lexicon(language: str | None = None) -> _CompiledContextLexicon: lexicon = get_clinical_cue_lexicon(language) token_boundaries = lexicon.token_boundaries