From da76296cc4a4bc49bb50f78977d10baed3b00bd0 Mon Sep 17 00:00:00 2001
From: "google-labs-jules[bot]"
 <161369871+google-labs-jules[bot]@users.noreply.github.com>
Date: Sat, 29 Aug 2026 17:33:11 +0000
Subject: [PATCH] Perf: Cache compilation of clinical context lexicons\n\nWrap
 `_compiled_context_lexicon` with `@lru_cache(maxsize=32)` to\nmemoize the
 heavy and deterministic creation and combination of multiple\nalternated
 regular expressions. Eliminates repetitive compilation overhead\nthat
 occurred across every text scan operation.

Co-authored-by: zrt219 <199104500+zrt219@users.noreply.github.com>
---
 .jules/bolt.md                      | 3 +++
 openmed/openmed/clinical/context.py | 2 ++
 2 files changed, 5 insertions(+)
 create mode 100644 .jules/bolt.md

diff --git a/.jules/bolt.md b/.jules/bolt.md
new file mode 100644
index 0000000..d8ccbb4
--- /dev/null
+++ b/.jules/bolt.md
@@ -0,0 +1,3 @@
+## 2024-08-29 - [Cache deterministic regex compilations]
+**Learning:** Deterministic regular expression compilation during text processing (e.g. NLP tasks) creates significant overhead when performed repeatedly, blocking the CPU. The `openmed.clinical.context` module builds heavy and complex alternated regular expressions out of cue lexicons over and over.
+**Action:** When working on NLP pipelines, aggressively identify regex compilation and text span transformations. Memoize expensive object constructions (like `_compiled_context_lexicon`) using `functools.lru_cache` to drastically reduce processing time.
diff --git a/openmed/openmed/clinical/context.py b/openmed/openmed/clinical/context.py
index 9fd11df..03779af 100644
--- a/openmed/openmed/clinical/context.py
+++ b/openmed/openmed/clinical/context.py
@@ -40,6 +40,7 @@
 from dataclasses import dataclass, replace
 from datetime import date
 from typing import Any, Literal
+from functools import lru_cache
 
 from openmed.clinical.lexicons import (
     ClinicalCueLexicon,
@@ -154,6 +155,7 @@ class _CompiledContextLexicon:
     backward_context_cues: frozenset[str]
 
 
+@lru_cache(maxsize=32)
 def _compiled_context_lexicon(language: str | None = None) -> _CompiledContextLexicon:
     lexicon = get_clinical_cue_lexicon(language)
     token_boundaries = lexicon.token_boundaries