perf(xlsx): hash-indexed style interning

[eliahilse]

TL;DR:

Summary:

• style pools intern through hash indexes instead of linear scans; every memo hit re-verifies against the live pool, so public same-length mutations can never yield a stale index
• number-format allocation validates candidates against the live table and fails explicitly (NumFmtTableFull) when ids are exhausted instead of silently colliding
• range formatting stages against a cloned stylesheet and commits only on success — mid-range exhaustion leaves the workbook untouched
• note for review: Stylesheet gains private interning caches, which prevents exhaustive struct-literal construction downstream (changeset carries a minor bump)

Test plan:

• cargo test -p betteroffice-xlsx-model -p betteroffice-xlsx-ops green
• apply formatting to a large range on a style-heavy workbook — timing before/after
• round-trip styles.xml byte-fidelity on real workbooks

[openooxml-bot]

All contributors have signed the CLA — thank you! ✍️ ✅

_{Posted by the CLA bot.}

[greptile-apps]

Greptile Summary

This PR replaces linear XLSX style interning with hash-indexed, live-pool-validated caches and makes custom number-format exhaustion explicit and atomic.

• Adds private, serialization-neutral caches for fonts, fills, borders, cell formats, and number formats.
• Introduces fallible number-format allocation and propagates exhaustion through operation and collaboration paths.
• Stages range formatting against cloned style state before committing cells.
• Adds tests for cache invalidation, duplicate IDs, serialization, exhaustion, and transactional behavior.

Confidence Score: 5/5

The PR appears safe to merge; no concrete changed-code defect remains after checking cache mutation, allocation exhaustion, operation atomicity, undo behavior, and dependency changes.

Cache hits are checked against live pools, custom-format allocation rejects exhausted or colliding IDs, production callers propagate that failure, and range operations commit only after complete staging succeeds.

Important Files Changed

Filename	Overview
crates/xlsx-model/src/styles.rs	Adds validated style-pool caches, borrowed format resolution, and collision-safe fallible custom number-format allocation with extensive mutation and serialization tests.
crates/xlsx-ops/src/apply.rs	Propagates number-format exhaustion and stages complete range-format changes before committing styles, cells, and inverses.
crates/betteroffice-xlsx/src/authority.rs	Adapts collaborative style materialization to propagate the newly fallible style-interning result.
crates/xlsx-parse/src/tests.rs	Updates style-writing coverage for the fallible interning API while preserving existing package-fidelity assertions.
crates/xlsx-render/src/lib.rs	Updates rendering tests to unwrap the new fallible style-interning result.
crates/xlsx-model/Cargo.toml	Adds serde_json only as a development dependency for stylesheet serialization tests.
Cargo.lock	Records the xlsx-model serde_json development dependency without changing the flagged transitive dependency versions.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Range formatting operation] --> B[Clone stylesheet]
  B --> C[Resolve each current cell format]
  C --> D[Apply requested formatting]
  D --> E[Intern through validated hash caches]
  E --> F{Number-format ID available?}
  F -->|No| G[Return NumFmtTableFull]
  G --> H[Discard staged state]
  F -->|Yes| I[Stage cell and style indices]
  I --> J{All cells complete?}
  J -->|No| C
  J -->|Yes| K[Commit stylesheet]
  K --> L[Commit cells and build inverse]

Loading

_{Reviews (1): Last reviewed commit: "perf(xlsx): hash-indexed style interning" | Re-trigger Greptile}