Test: PDF → Excel (1 Nov 2026, round 4)
Now passing 21 of 21 tests
Test PDFs with typical problems of real documents. Round 2 (tests 09–13): images, charts, two-level headers, mixed fonts, two-column text. Round 3 (14–19): sideways table, a row broken by a page break, watermark, merged cells, tricky numbers, two tables side by side. Round 4 (20): portrait and landscape pages in one PDF and choosing the Excel print orientation. “Before” is version 4 of the converter, “Now” is version 5. Every cell is compared row by row, and the column names (header) separately. Click a preview to see the page.
| # | Preview | What is tested | Before (v4) | Now (v5) | Files |
|---|---|---|---|---|---|
| 00 | ![]() | 14-page price list, ruled table, header repeated on every page | 540 of 540 rows, header OK | 540 of 540 rows, header OK | PDF · Excel · expected |
| 01 | ![]() | Simple ruled table on 3 pages + total below the table | 90 of 90 rows, header OK | 90 of 90 rows, header OK, print: portrait | PDF · Excel · expected |
| 02 | ![]() | Bank statement WITHOUT lines, empty Debit/Credit cells, 2 pages | 45 of 45 rows, header OK | 45 of 45 rows, header OK | PDF · Excel · expected |
| 03 | ![]() | Long descriptions wrap to 2–3 lines, horizontal lines only | 30 of 30 rows, header OK | 30 of 30 rows, header OK | PDF · Excel · expected |
| 04 | ![]() | Two different tables on one page: summary and details | 28 of 28 rows, header OK | 28 of 28 rows, header OK | PDF · Excel · expected1 · expected2 |
| 05 | ![]() | Group subtotals and a grand total inside the table | 18 of 18 rows, header OK | 18 of 18 rows, header OK | PDF · Excel · expected |
| 06 | ![]() | Landscape page, numbers like “€ 1.234,56”, negatives | 12 of 12 rows, header OK | 12 of 12 rows, header OK, print: landscape | PDF · Excel · expected |
| 07 | ![]() | Invoice: company details on top, line items, Subtotal/VAT/Total | 8 of 8 rows, header OK | 8 of 8 rows, header OK | PDF · Excel · expected |
| 08 | ![]() | Scan (image without a text layer) | “needs OCR” message — correct | “needs OCR” message — correct | PDF · expected |
| 09 | ![]() | Round 2. Logo, vector chart, a product photo in every table row | 16 of 16 rows, header OK | 16 of 16 rows, header OK | PDF · Excel · expected |
| 10 | ![]() | Round 2. Two-level header: “Size” over 3 columns, “Price by quantity” over 5 sub-columns, 9 columns | 24 of 24 rows, header OK | 24 of 24 rows, header OK, print: landscape | PDF · Excel · expected |
| 11 | ![]() | Round 2. Cyrillic, 7 different fonts and sizes (7–11 pt), bold, italic, coloured rows, no lines, “Итого” total | 28 of 28 rows, header OK | 28 of 28 rows, header OK | PDF · Excel · expected |
| 12 | ![]() | Round 2. Article: text in 2 columns, picture, table, paragraph right after the table | 12 of 12 rows, header OK | 12 of 12 rows, header OK | PDF · Excel · expected |
| 13 | ![]() | Round 2. Report without a grid: “2026” over 5 columns, “2025” over 2, italics | 14 of 14 rows, header OK | 14 of 14 rows, header OK | PDF · Excel · expected |
| 14 | ![]() | Round 3. Table printed sideways: portrait page, text and lines rotated 90° | 22 of 22 rows, header OK | 22 of 22 rows, header OK, print: landscape | PDF · Excel · expected |
| 15 | ![]() | Round 3. No lines, 3 pages: a row broken by a page break (twice), no header on page 3, page headers/footers | 34 of 34 rows, header OK | 34 of 34 rows, header OK | PDF · Excel · expected |
| 16 | ![]() | Round 3. CONFIDENTIAL watermark over the table, striped rows without lines, headers/footers, 2 pages | 48 of 48 rows, header OK | 48 of 48 rows, header OK | PDF · Excel · expected |
| 17 | ![]() | Round 3. Vertically merged cells: the category is written once for 2–5 rows (repeated in Excel) | 18 of 18 rows, header OK | 18 of 18 rows, header OK | PDF · Excel · expected |
| 18 | ![]() | Round 3. Numbers: (1,234.56), 1,234.56-, −12.50, 150.00 CR / 75.20 DR, USD 99.00, 1'234.50, 1 234, +5.0 % | 28 of 28 rows, header OK | 28 of 28 rows, header OK | PDF · Excel · expected |
| 19 | ![]() | Round 3. Two different tables without lines side by side on one page (left and right), different lengths | 23 of 23 rows, header OK | 23 of 23 rows, header OK | PDF · Excel · expected1 · expected2 |
| 20 | ![]() | NEW. One PDF: page 1 portrait (4 columns), page 2 landscape (10 columns). Sheet 1 prints portrait, sheet 2 landscape; choosing “portrait” / “landscape” manually changes all sheets | no such test | 39 of 39 rows, header OK, print: portrait / landscape | PDF · Excel · expected1 · expected2 |
Portrait or landscape
- Auto (default): each Excel sheet prints the same way as the PDF page its table came from. A wide table that does not fit a portrait page prints landscape.
- Portrait or Landscape — for all sheets at once, when one format is needed.
- In every mode the full table width fits on one page and the header repeats on every printed page. Paper: A4 or US Letter.
Good to know
- The test PDFs are generated (not real client documents). Real PDFs can be harder — real samples are needed for the next round.
- Scans (images) are not processed yet: that needs OCR, which has not been tested. The converter says so clearly.
- Totals are checked automatically: if a column sum matches a total in the PDF, the Check sheet shows OK.
- A column with photos is not carried over to Excel (it has no text) — it is removed.
- A vertically merged cell is repeated on every row in Excel — easy to filter.
- Percentages become numbers as written (+5.0 % → 5.0); DR and brackets become negative.
- Header without a grid: a group label (“2026”) is linked to columns by the line under it. Without that line the columns keep simple names.




















