Doc2X table-to-Excel verification / 表格转 Excel 核对 — 2026-10-10

Self-authored test data; no customer or personal information.
table-excel-input.jpg and table-excel-scan-input.pdf show the same page.
The PDF contains the same JPEG only, with no searchable text layer.
Each input was submitted once to the real Doc2X parsing interface.
No scan enhancement or translation was applied.

table-image-raw.xlsx: first actual image-result Table Excel download (version 0).
The original x^2 + 1 was misrecognized as $\times {72} + 1$ in 表2!B4.
table-image-reviewed.xlsx: after manually changing only that cell to x^2 + 1,
saving version 1 in the result editor, and downloading again.
table-scan-output.xlsx: actual scan-PDF Table Excel download, without editing.
The scan-PDF result already contains x^2 + 1 as plain text.
All three XLSX files retain the original download bytes.

All three workbooks have two worksheets: 表1 = 5 rows x 5 columns,
表2 = 4 rows x 3 columns. 表1 retains five merged ranges:
A1:A2, B1:B2, C1:D1, E1:E2, A3:A4.
Leading-zero IDs 0012/0013/0014 and the 18-digit identifier remain strings.
Quantities 12/18/24 and adjustments -3.5/0/8.25 are numeric cells.
Original -3.50 and 0.00 use General format, so trailing decimal zeroes
are not a fixed two-decimal display format. Set your own number format.
Percentages 12.5%/100%/0% and date 2026-10-10 are text cells.
The math expression is text, not an Excel calculation formula.
CHECKS.json lists the exact cell values, types and merge ranges.

核对要点：每表一张工作表；保留合并表头和编号文本；百分比、日期、数学表达仍为文本。
图片初次识别存在一格错误，修正后保存版本 1，再导出才带入新内容。
扫描 PDF 的同一格在本次无需修正。不要据此推断所有 PDF 都优于图片。
This is one clean page, not a cross-page, blurred-photo or embedded-image test.
No native Excel or WPS visual/editing compatibility test was completed here.
Open the files in your own spreadsheet software and verify before use.
