HTML-to-Word gained an Advanced-settings Plain text checkbox: default stays structured Word; on, headings/tables/links/images flatten to editable paragraphs.

Date: 2026-09-07 22:07 Summary: HTML-to-Word gained an Advanced-settings Plain text checkbox: default stays structured Word; on, headings/tables/links/images flatten to editable paragraphs. Visibility: people

[question] 能否实现，可以根据用户选择是否纯文字

Tool links (English)

convert-html-web-pages-to-word-document

[try to solve]

Context

After explaining that this tool maps HTML structure (not CSS layout) into Word, the user asked for a choice: keep that structured mapping, or export plain text only. The job is still one .docx download, not a .txt file and not a screenshot.

Default Convert (Advanced closed) had to stay the structured path: headings, lists, tables, hyperlinks, optional images. The new mode is opt-in inside the collapsed Advanced settings, same pattern as Include images.

Process

Added checkbox #chwPlain (unchecked) under Include images in convertHtmlWebPagesToWordDocumentPage.ts. Checking it disables the images control so the two switches cannot fight.

htmlToBlocks now takes plainText. When it is on:

h1–h6 become ordinary paragraphs (no Word heading style).
ul/ol items become paragraphs without • / 1. prefixes.
table is flattened with flattenTablePlain: each row is a paragraph, cells joined by tabs — no docx.Table.
img is alt text only; Convert also forces includeImgs off.
inlineRuns(..., plain) drops bold/italic and writes links as label (https://…).

Ten-locale UI keys plain_label / plain_hint were written independently (zh 纯文字, en Plain text, es Solo texto, ja プレーンテキスト, …), and How / opts / rules mention the switch. Catalog updatedAt touched; this slug prerendered and gzipped.

Playwright against the new HTML: sample Convert produced w:tbl, Heading1, and w:hyperlink. After checking Plain text, Convert succeeded again: no table, no Heading1, no hyperlink object, and the sample link became the help article (https://example.com/help).

Root cause / analysis

“Plain text” here means no Word objects for structure, not “strip tags then dump one run.” Lawyers still get paragraphs they can copy; they do not get a Word table they have to delete, or Heading 1 they have to restyle. Putting the control in Advanced keeps the FreeConvert-style converter chrome: one Convert button, settings collapsed, sensible default.

A CSS-to-Word engine is still out of scope (see the previous CSS note). Plain text does not restore layout; it only opts out of structure mapping.

Solution

Use 高级设置 → 纯文字 (English: Advanced settings → Plain text) when you want a draft of copy. Leave it off when you want editable headings and tables. Include images is ignored while Plain text is on.

Local Cache API still keys HTML by PAGES_CACHE_VERSION (now 4.59). Restart the local Worker after that bump, or the pretty /zh/tools/… URL can keep serving the previous HTML without the checkbox.

Notes / boundaries

Output is still .docx, not .txt.
Lists lose their list style; numbering is not reconstructed.
Nested tables flatten one row at a time; inner tables are not pretty-printed.
Isolation lint flags wrangler.jsonc (cache version) plus earlier shared fetch work in src/tools/convertHtmlToPdf.ts; set CROSS_TOOL_UPDATE=1 if that gate must pass in the same working tree.

[actions]

src/pages/convertHtmlWebPagesToWordDocumentPage.ts: #chwPlain, flattenTablePlain, plainText through htmlToBlocks / inlineRuns / Convert
Ten-locale shards: plain_label, plain_hint; How / opts / rules mention the option
wrangler.jsonc: PAGES_CACHE_VERSION 4.59
Catalog touch, merge:tools, prerender + gzip this slug