From 1b0e5472ce5802a08a244b3669bf59465a7074bd Mon Sep 17 00:00:00 2001
From: "google-labs-jules[bot]"
 <161369871+google-labs-jules[bot]@users.noreply.github.com>
Date: Tue, 1 Sep 2026 17:31:50 +0000
Subject: [PATCH] =?UTF-8?q?=E2=9A=A1=20Bolt:=20Memoize=20expensive=20regex?=
 =?UTF-8?q?=20compilations=20in=20ConText=20lexicon?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Optimizes the `_compiled_context_lexicon` function in `openmed.clinical.context` by wrapping it with `@functools.lru_cache(maxsize=None)` and ensuring immutability with `types.MappingProxyType`. This prevents compiling identical and complex regex patterns multiple times across different span resolutions, significantly improving performance for heavy clinical document evaluation pipelines.

Co-authored-by: zrt219 <199104500+zrt219@users.noreply.github.com>
---
 .jules/bolt.md                      | 3 +++
 openmed/openmed/clinical/context.py | 5 ++++-
 2 files changed, 7 insertions(+), 1 deletion(-)
 create mode 100644 .jules/bolt.md

diff --git a/.jules/bolt.md b/.jules/bolt.md
new file mode 100644
index 0000000..3d9bf00
--- /dev/null
+++ b/.jules/bolt.md
@@ -0,0 +1,3 @@
+## 2024-05-18 - Missing Memoization for Regex Compilations
+**Learning:** `_compiled_context_lexicon` function inside `openmed/openmed/clinical/context.py` compiles multiple heavy regular expressions dynamically without memoization. The function is called for every span context resolution (`resolve_temporality`, `resolve_uncertainty`, etc), leading to severe performance bottlenecks.
+**Action:** Always memoize deterministic regex compilations and lexicon generation in NLP pipelines (e.g. `openmed.clinical`) using `@functools.lru_cache` to prevent bottlenecks during repeated string/span evaluations. Make sure to only cache hashable parameters.
diff --git a/openmed/openmed/clinical/context.py b/openmed/openmed/clinical/context.py
index 9fd11df..ceacbcc 100644
--- a/openmed/openmed/clinical/context.py
+++ b/openmed/openmed/clinical/context.py
@@ -35,7 +35,9 @@
 
 from __future__ import annotations
 
+import functools
 import re
+import types
 from collections.abc import Iterable, Iterator, Mapping, Sequence
 from dataclasses import dataclass, replace
 from datetime import date
@@ -154,6 +156,7 @@ class _CompiledContextLexicon:
     backward_context_cues: frozenset[str]
 
 
+@functools.lru_cache(maxsize=None)
 def _compiled_context_lexicon(language: str | None = None) -> _CompiledContextLexicon:
     lexicon = get_clinical_cue_lexicon(language)
     token_boundaries = lexicon.token_boundaries
@@ -195,7 +198,7 @@ def _compiled_context_lexicon(language: str | None = None) -> _CompiledContextLe
             ),
             token_boundaries=token_boundaries,
         ),
-        category_by_text=_cue_category_lookup(lexicon),
+        category_by_text=types.MappingProxyType(_cue_category_lookup(lexicon)),
         backward_context_cues=frozenset(
             _normalize_cue_text(cue) for cue in lexicon.backward
         ),