Shipped a batch webpage-to-Excel converter whose artifact is HTML table cells in one .xlsx workbook, not a screenshot ZIP, A4 PDF, or Word prose.

Date: 2026-09-07 17:13 Summary: Shipped a batch webpage-to-Excel converter whose artifact is HTML table cells in one .xlsx workbook, not a screenshot ZIP, A4 PDF, or Word prose. Visibility: people

[question] 立项和实现：batch convert web pages to excel,注意不同工具的要有足够不同的info gain

Tool links (English)

https://onlinefreetools.org/tools/batch-convert-web-pages-to-excel Chinese: https://onlinefreetools.org/zh/tools/batch-convert-web-pages-to-excel

[try to solve]

Context

The request was to brief and implement batch convert web pages to excel with enough information gain that it is not a doorway next to the existing batch webpage converters (JPG / PNG / PDF) or HTML→Word.

Those neighbors already occupy:

Batch JPG/PNG: viewport rasters, ZIP of images.
Batch PDF: A4 portrait pages.
HTML→Word: headings, lists, and paragraphs in an editable .docx.
CSV ↔ JSON: a file you already have, not a webpage.
Excel formula / invoice template pages: generate formulas or blank workbooks, not scrape HTML tables.

Keyword Planner (docs/seo/keywords/web-page/excel-Keyword Planner 09-06-2026 at 05-21-44.csv) had two clusters: webpage to excel (10–100, absorb into a URL list) and larger html to excel / html table to excel (100–1k, absorb into a Paste HTML tab). Head terms like excel web / Excel Online and reverse jobs (save excel as web page) were dropped. pdf table to excel is intentionally unsatisfied (no OCR).

Process

Wrote work-tasks/batch-convert-web-pages-to-excel/ (00–03) with coverage table, Ads/Planner rows, and 0i intent review. coverage:gate --phase=0b green before engineering.
Implemented catalog page.style: opts, icon, Page, and English master. Default input is a URL list (H1 job). Secondary tab Paste HTML absorbs html-to-excel without a second URL. Convert all → one web-pages.xlsx. Download stays disabled until a workbook exists.
Table mapping: DOMPurify → DOMParser → <table> tr/th/td text as AOA → SheetJS. Default one sheet per HTML table; optional one sheet per URL (stack with a blank row). Tiny tables under 2×2 skipped by default. colspan gets empty placeholders; rowspan is flattened. Fetch reuses GET /api/tools/convert-html-to-pdf?url=.
Sample uses two local HTML tables (price list + stock grid) at example.com-shaped URLs without hitting the live site. loadSample() runs on page load (no html2canvas, so the tab does not freeze).
Ten-locale independent rewrites. H1 EN: “Batch convert web pages to Excel”. ZH: “批量把网页转成 Excel”. coverage:gate --phase=2 then --phase=4 (en,zh,es,ja listed on the 2b row).
Reverse related on HTML→Word and csv-json. README EN/ZH list entries. CROSS_TOOL_UPDATE=1 npm run verify:tool green (full build:site, HTML smoke, SEO, vendor, isolation). Home cards and prerendered title confirmed.

Root cause / analysis

Shared URL fetch is an implementation detail. Unique copy has to be the artifact:

Neighbor	Contract without the H1
JPG/PNG batch	Pixels in a ZIP
PDF batch	A4 print pages
HTML→Word	Editable prose styles
CSV ↔ JSON	File you already have
This page	HTML `<table>` cells in one workbook, many sheets

If the page talked about viewport, JPEG quality, or “capture the page”, stripping the title would leave another format-suffix variant. That is doorway / scaled-content risk. Information gain here is a people-first content principle (verifiable extra vs SERP neighbors), not a claimed ranking switch.

Planner absorb vs build: one-URL webpage to excel is the same job as the batch list. html to excel / html table to excel is the same mapping with pasted markup, so one slug with tabs — not /html-to-excel plus /webpage-to-excel.

Solution

Default textarea = batch URLs; Convert all / Download Excel; advanced settings collapsed (sheet layout, skip tiny tables).
FAQ and Rules state: prose ignored; JS-only tables empty; no PDF OCR; not Excel for the web; not mhtml.
Privacy: paste stays in the tab; each URL is fetched once and not stored.
Ship gates: coverage 0b → 2 → 4 → all; verify:tool with CROSS_TOOL_UPDATE=1.

Notes / boundaries

Tool HTML on production is served from R2. git push alone does not refresh live tool bodies; run npm run upload:r2 when you want production to match.
Scripts are stripped, so login walls and JavaScript-rendered tables fail that row; the rest of the batch continues.
Sheet names are capped at 31 characters (Excel limit).
This session did not commit or push.
No interactive browser session here; Convert/Download is covered by lint:tool-page --require-html, prerendered H1/meta, and in-page loadSample().

[actions]