bug: preserve explicit false run properties across read/write

[kiki830621]

Problem

DocxReader.parseRunProperties treats <w:b>, <w:i>, <w:strike>, and <w:noProof> as true whenever the element exists, ignoring explicit off values such as w:val="0", false, or off. RunProperties.toXML() can then emit only the on form. A read → unrelated mutation → write round-trip therefore turns an untouched <w:b w:val="0"/> into <w:b/>.

Type

bug

Expected

• Typed on/off run properties preserve three states: absent, explicitly on, explicitly off.
• An untouched explicit-off property remains semantically off after save.
• Existing source-compatible reads such as properties.bold still return false for absent and explicit-off states.
• Merge/formatting APIs can intentionally apply false, rather than treating it as “unspecified”.

Actual

• RunProperties.bold, italic, strikethrough, and noProof are plain Bool values.
• The reader sets them to true on element presence.
• The emitter omits false and emits naked elements for true.
• merge(with:) applies only true, so callers cannot remove an existing boolean property.

Scope / Acceptance

• Add RED→GREEN read/write tests for <w:b w:val="0"/>, including an unrelated paragraph mutation before save.
• Cover accepted ST_OnOff spellings (0/false/off and omitted/1/true/on).
• Preserve source compatibility for callers reading/writing the existing Bool properties.
• Make merge distinguish an unspecified patch from an explicit false patch.
• Cover italic, strikethrough, and noProof through the same typed mechanism; untyped raw run-property children remain outside this issue.
• Run focused reader/writer/model tests and the full package suite.

Cross-repo tracking

• Parent: Boolean run property w:val="0" dropped on save — bold spreads to untouched runs macdoc#173
• Consumer behavior: format_text bold:false is a no-op — cannot remove existing boolean formatting macdoc#174 and a forthcoming che-word-mcp child issue.

---

Current Status

Phase: verified
Last updated: 2026-08-31 by idd-verify

Key Decisions

• Preserve public Bool access while storing absent/on/off assignment presence internally.
• Merge and copy by presence, not truthiness; explicit false and underline none survive parser, writer, RunPayload, reducer, reverse extractor, and revision paths.
• rPrChange delegates to the single canonical RunProperties.toXML() emitter.

Blocking

• (none)

Follow-up

• bug: OMath splice rPr modes are bypassed by Run.rawXML #117 tracks the pre-existing OMath splice rawXML/rPr carrier defect.

Commits

• 563039c — final program behavior.
• a3fab8b — frozen consumer proof.

Verification

• PR #116 full report
• Focused 146/146; full 1,451 tests with 31 skips and 0 failures.
• Consumer draft PR refactor: honor false text formatting che-word-mcp#198: 347 tests, 12 skips, 0 failures.
• Spectra analyze 0 findings; strict/diff gate/Analyze (swift)/CodeQL PASS.
• PR refactor: preserve explicit false run properties #116 MERGEABLE/CLEAN; verified tag idd-115-verified; still awaiting human merge.

Pull Request

• refactor: preserve explicit false run properties #116

[kiki830621]

Implementation Complete

Changes

• 8cc8e48 — source-compatible RunProperties presence state、ST_OnOff parse/emit、explicit false/nil merge、DOCX regressions。
• PR: refactor: preserve explicit false run properties #116
• Spectra change: explicit-false-run-properties（analyze 0 findings；strict valid）。

Verification

• RED: 5 tests／12 assertion failures against pre-change main。
• Focused Run/Paragraph/table/EditAlgebra/OMath matrix: 142/142 PASS。
• Full package: 1,447 tests，31 skips，0 failures。
• Consumer integration: bug: format_text false values are reported as success but ignored che-word-mcp#197 draft PR #198，2/2 focused + 347 full tests PASS against this exact revision。

Next

PR #116 進入獨立 IDD verify；PASS 前不合併、不發版。

[kiki830621]

Verify (via PR #116)

Result: PASS — frozen a3fab8b 通過 5 個獨立 reviewer、Codex CLI、1,451-test full suite、consumer integration、Spectra、CI 與 freshness gate；0 blocking、1 follow-up（#117）。

Full report: #116 (comment)

This issue's findings: see "#115" section in the linked report.