Test a cell on each page
A supplier sends two pages of item descriptions and amounts. Try selecting an individual word on each page. Check where the table starts, how the amounts line up and whether page two repeats the column headings. The tool applies the same relative grid to the selected pages, so a table that moves halfway down the second page needs a separate extraction.
Place the grid on the sample
- Open PDF table to CSV and choose the source PDF below. Set the page range to 1-2. On the preview enclose the table, including its headings.
- For this sample the left, top, right and bottom edges are approximately 6.72%, 19.48%, 93.28% and 39.43% of the page. Within the rectangle set column boundaries to 0,33.01,67.96,100 and row boundaries to 0,25,50,75,100.
- Choose Extract table. In the first page’s rows, change the note for Apple, red to Edited, "quote" and set the last item label to =SUM(1,2). Leave page two unchanged as a comparison. Download the CSV.
Read the eight records back
The 192-byte sample starts with a UTF-8 byte order mark. A CSV reader returns eight rows of three fields, including both header rows. Apple, red occupies one field. The edited note contains a comma and a pair of quotation marks. The note next to 中文项目 stays empty. Check these particular cells first because a shifted delimiter can leave a sheet looking plausible while putting amounts in the wrong column.
Check how the spreadsheet treats amounts
The exported values beginning with a formula marker carry a leading apostrophe. In the sample this also affects -7. Inspect the imported cell type before summing the Amount column. Account for the second header when combining the rows. A scan, a merged heading crossing columns or wrapped text inside a cell needs a different extraction or a manual transcription checked against the source.