CSV to PDF Table Converter

CSV to PDF turns a comma-separated table into a real, printable PDF table. Pick a CSV file, choose A4 or Letter and portrait or landscape, and the tool reads the first row as the header, lays every data row out underneath it in columns, and gives you a PDF you can download, print or email. It is the everyday job of getting a spreadsheet export — a price list, a mailing list, a stock report, a set of results — out of a cell-and-comma file and into something that looks like a document. The whole layout runs inside your browser tab: the CSV is read with the same file-picker tools the rest of the site uses, and nothing you open is uploaded anywhere.

The table is honest about its geometry. The first non-blank row is the header, drawn in Helvetica-Bold on a light grey band; every row below it is body text in regular Helvetica at 9 pt. Each column is measured in the font it will be drawn in — the header in bold, the data in regular — and sized to the widest text it holds, so a column of long customer names gets more room than a column of single digits. If the columns together are narrower than the page they are widened to fill the width; if they are wider, they are shrunk in proportion to how far each is above a floor, so the widest column gives up the most and no column collapses. The header is drawn once per page, so a table that runs onto a second page still begins that page with a bold header row — a reader never has to scroll back to find out what a column is.

No data is dropped. A value too long for its column is wrapped onto extra lines and shown in full, and a column that spare width would otherwise stretch is capped at half the usable width so a long value wraps onto several readable lines rather than running the full page as one very wide line. The single bounded case is a cell that would need more lines than a page can hold: it is cut at the page's row limit and the last visible line ends with an ellipsis (…), so the reader can see that something was left out rather than being silently misled. A row is never split across a page break. A CSV that is empty, has no data rows, or has a row whose cell count does not match the header is refused by name, because silently padding or shifting a column would misalign the whole table.

The one limit to know before you start is character coverage. Only the two standard PDF fonts are used — Helvetica and Helvetica-Bold — which encode the WinAnsi character set (PDF 32000-1 Annex D): Latin letters including Western European accents (é, ü, ñ), digits, and common punctuation such as curly quotes, the em dash and the euro sign. Characters outside that set — Chinese, Japanese, Korean, Arabic, Hebrew, Cyrillic, Greek, Devanagari, mathematical symbols such as ≤ or ∞, and all emoji — cannot be drawn. Every character in the CSV is checked against the embedded font before anything is drawn, and the offenders are named in a specific message rather than being replaced with an empty box or dropped.

This is a one-way converter: it builds a new PDF from a CSV, and it does not read an existing PDF. If you need the reverse — a table pulled out of a PDF — that is a different tool. The CSV itself is never modified, so you can change the page size and create another version from the same file. Everything runs on your own device, and your browser's Network tab will show that no request carries your file.

1. The CSV text is parsed into a grid, honouring RFC 4180 quoting: a double quote at the start of a field opens a quoted field, a pair of double quotes inside one is a literal quote, and a comma or line break inside a quoted field does not split the field. A double quote inside an already non-empty unquoted field is kept as a literal character, so a value such as Café — "quoted" • prints exactly as written. CRLF line endings are tolerated. 2. Blank rows (every cell empty) are ignored, so a trailing newline is not a ragged row. The file must not be empty, must have at least a header and one data row, must not exceed 1,000,000 characters, and every row must have exactly as many cells as the header — otherwise it is refused by row number. 3. The page box is resolved from the size (A4 = 595.28 × 841.89 pt, Letter = 612 × 792 pt) and the orientation (landscape swaps the axes). A 36 pt (half-inch) margin is taken on all four sides, and a table so wide that a column would have less than a glyph's room is refused with the page named in the message. 4. Each column's natural width is the widest of its header cell (measured in Helvetica-Bold) and its data cells (measured in Helvetica), plus 10 pt of inner padding. Columns are then laid out to the content width: if the natural widths fit, the spare width is shared equally but each column is capped at half the usable width; if they overflow, every column keeps a 24 pt minimum and the remaining width is shared in proportion to how far each column sits above that minimum. 5. Every character in the header and the data is probed against the embedded font, and any character outside the WinAnsi set is refused by name before anything is drawn. 6. Each cell is word-wrapped to its column's usable width, with a word wider than the column broken character by character. A row is as tall as its tallest cell. A cell that would need more lines than fit between the header and the bottom margin is cut at that limit and its last line ends with a visible ellipsis (…). 7. The header is drawn at the top of the page on a grey band, then each row below it, dropping by the row height. When the next row would cross the bottom margin, a new page begins and the header is drawn again at the top. The two faces are registered under the fixed resource names /F1 (body) and /F2 (header). The document is saved with AllUtil as producer and returned to the browser.