[wenqiw777]

python-docx resolves paragraph.style and run.style through DocumentPart.get_style(), which scans the whole styles part on every call: an XPath per style id, and a full scan with enum parsing of every style for the default style. The DOCX backend reads the paragraph style several times per paragraph and once per run, so a 135-page document makes ~35k lookups for 12 distinct styles. Styles do not change during conversion, so the lookup is memoized on the loaded document part, as #4763 does for the ODF backend.

Checklist:

• Documentation has been updated, if necessary.
• Examples have been added, if necessary.
• Tests have been added, if necessary.

Review evidence:

Generated DOCX reports (headings, paragraphs with bold/italic runs, nested bullet and numbered lists, tables, header and footer), 5 warm runs each, Python 3.12, python-docx 1.2.0, Apple Silicon:

document	before (median / min)	after (median / min)	speedup
135 pages, 5.9k items	6.13 s / 5.92 s	2.61 s / 2.50 s	2.3x
270 pages, 11.8k items	11.87 s / 11.84 s	5.26 s / 5.22 s	2.3x
535 pages, 23.5k items	24.03 s / 24.01 s	10.61 s / 10.53 s	2.3x

Peak RSS is unchanged (447 / 500 / 602 MB before, 446 / 501 / 608 MB after). In cProfile of the 135-page file, CT_Styles.default_for and get_by_id took 5.2 s of 10.4 s profiled time; with the memo, the 34,793 get_style calls hit 12 cache entries.

Outputs are identical: all 97 fixtures under tests/data/{docx,html,md,xlsx,pptx}/sources were converted with main and with this branch and exported to JSON and Markdown; diff -r reports no differences. pytest tests/test_backend_msword*.py passes (292 passed, 1 skipped, including the new test). No reference data change.

The new test converts two documents whose default paragraph style differs in one process and checks that each keeps its own bold, so the memo cannot leak across documents.

[github-actions]

✅ DCO Check Passed

Thanks @wenqiw777, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[github-actions]

AI triage (advisory)

Checked commit 4d9bf8fbef. The analysis read the diff only. It did not run code from this pull request.

Warning

The AI analysis did not finish. Only the deterministic checks are shown.

Topic labels added: performance, docx.

CI safety: not assessed.

Possibly related pull requests

PR	State	Signal	Assessment
#4831	open	changes the same code	not assessed
#4584	open	changes the same code	not assessed
#4763	merged	referenced as #4763	not assessed

_{The ai:* labels are advisory. They do not approve CI and they do not add tests:full. A new push runs this check again.}

[github-actions]

AI first review (advisory): no blocker found

This review covers the full diff at 4d9bf8fbef. It read the code and did not run it.

The patch attaches functools.cache to a per-load DocumentPart instance, matching the accepted ODF pattern. The cache is scoped to each loaded document, so no cross-document leakage is possible. The regression test constructs two documents with different Normal-style bold settings and converts them in sequence; it is sufficient to catch a class-level cache mistake. No behavioral regression or contract change was found.

_{A maintainer makes the merge decision. This review never approves the pull request.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[ExtReMLapin]

I could have done the same changes for docx, but I was focused on our own usage

[PeterStaar-IBM]

lgtm!