From 1fff8e5987feaa8f03001eeae8ecdb74a47b3cc5 Mon Sep 17 00:00:00 2001
From: "google-labs-jules[bot]"
 <161369871+google-labs-jules[bot]@users.noreply.github.com>
Date: Thu, 23 Jul 2026 17:28:19 +0000
Subject: [PATCH] Add caching to lexicon regex compilations

Co-authored-by: zrt219 <199104500+zrt219@users.noreply.github.com>
---
 .jules/bolt.md                      |  3 +++
 openmed/openmed/clinical/context.py | 36 +++++++++++++++++++++--------
 2 files changed, 30 insertions(+), 9 deletions(-)
 create mode 100644 .jules/bolt.md

diff --git a/.jules/bolt.md b/.jules/bolt.md
new file mode 100644
index 0000000..09c307e
--- /dev/null
+++ b/.jules/bolt.md
@@ -0,0 +1,3 @@
+## 2024-07-23 - Context Lexicon Regex Compilation Bottleneck
+**Learning:** Deterministic lexicon processing (like negation and temporality cues) creates massive regular expression alternations across clinical pipelines. Re-compiling these every time `_compiled_context_lexicon` was called caused a significant CPU bottleneck.
+**Action:** Use `@functools.lru_cache` to cache compiled regular expressions and lexicons. Ensure that arguments to these functions (e.g. `cues`) are typed as `tuple` rather than unhashable `Iterable` to be compatible with caching.
diff --git a/openmed/openmed/clinical/context.py b/openmed/openmed/clinical/context.py
index 9fd11df..810bacb 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
@@ -99,10 +100,10 @@
 PSEUDO_NEGATION_CUES = _ENGLISH_CONTEXT_LEXICON.pseudo_negation
 
 
-def _cue_pattern(
-    cues: Iterable[str],
-    *,
-    token_boundaries: bool = True,
+@functools.lru_cache(maxsize=16)
+def _cached_cue_pattern(
+    cues: tuple[str, ...],
+    token_boundaries: bool,
 ) -> re.Pattern[str]:
     alternation = _cue_alternation(cues)
     if not alternation:
@@ -114,17 +115,25 @@ def _cue_pattern(
     return re.compile(pattern, re.IGNORECASE)
 
 
-def _cue_alternation(cues: Iterable[str]) -> str:
+def _cue_pattern(
+    cues: Iterable[str],
+    *,
+    token_boundaries: bool = True,
+) -> re.Pattern[str]:
+    return _cached_cue_pattern(tuple(cues), token_boundaries)
+
+
+def _cue_alternation(cues: tuple[str, ...]) -> str:
     return "|".join(
         r"\s+".join(re.escape(part) for part in cue.split())
         for cue in sorted(set(cues), key=len, reverse=True)
     )
 
 
-def _terminator_pattern(
-    cues: Iterable[str],
-    *,
-    token_boundaries: bool = True,
+@functools.lru_cache(maxsize=16)
+def _cached_terminator_pattern(
+    cues: tuple[str, ...],
+    token_boundaries: bool,
 ) -> re.Pattern[str]:
     alternation = _cue_alternation(cues)
     punctuation = r"[.!?;。！？；]"
@@ -137,6 +146,14 @@ def _terminator_pattern(
     return re.compile(rf"(?:{punctuation}|{cue_pattern})", re.IGNORECASE)
 
 
+def _terminator_pattern(
+    cues: Iterable[str],
+    *,
+    token_boundaries: bool = True,
+) -> re.Pattern[str]:
+    return _cached_terminator_pattern(tuple(cues), token_boundaries)
+
+
 @dataclass(frozen=True)
 class _CompiledContextLexicon:
     language: str
@@ -154,6 +171,7 @@ class _CompiledContextLexicon:
     backward_context_cues: frozenset[str]
 
 
+@functools.lru_cache(maxsize=16)
 def _compiled_context_lexicon(language: str | None = None) -> _CompiledContextLexicon:
     lexicon = get_clinical_cue_lexicon(language)
     token_boundaries = lexicon.token_boundaries