CSV Column Editor

How to edit CSV columns

1. Paste text or import one UTF-8 CSV, TSV or TXT file. Select the input delimiter explicitly: comma, semicolon or tab. First row contains headers is on by default.
2. Press Parse columns. Review the original preview and record count. Imports are parsed before replacing a previous table or mapping.
3. Edit output labels, uncheck columns to remove them, and use the labelled up/down buttons to move them. Source column numbers remain stable. Select all, Reset mapping and Undo mapping let you restore columns or the last mapping change.
4. Choose an independent output delimiter, header presence, BOM and formula protection. Defaults preserve the input delimiter, header setting and BOM. Acknowledge blank or duplicate output names, or rename them.
5. Press Build export. Review the transformed preview and changed-cell examples, then copy or download the full result. Any input, mapping or setting change disables the previous export until a fresh build.

Worked example: names before IDs

id,name,city
001,Alice,Paris

Exclude source column 3 (city), rename source column 2 to full_name and move it above source column 1. The export is:

full_name,id
Alice,001

The identifier stays 001, not 1. Reordering moves each selected column's cells across every record; it never changes record order or count. Excluded columns remain in the original source and can be included again.

Parsing and data integrity

The bundled structured CSV parser handles quoted delimiters, doubled quotes, quoted LF/CRLF cell line breaks and one leading UTF-8 BOM. Completely empty physical records outside quotes and the terminal separator are ignored. Quoted-empty and delimiter-only records remain. Ragged field counts and malformed quotes block replacement or export with actionable parser-record errors; logical records may span multiple physical lines.

All cells remain strings, including whitespace, large integers, date-like text and leading zeros. Source headers are not trimmed or silently repaired: duplicate, empty and whitespace-only labels are valid and addressed by stable source position. Included output labels are checked for empty/whitespace-only names and exact duplicates. Export requires explicit acknowledgement if these warnings remain.

With input headers off, the mapping uses Column 1, Column 2 and so on without automatically exporting them. Editable labels stay in the interface when output headers are off; enabling output headers explicitly exports them. Header-only input can produce a header-only export, while completely empty input cannot infer columns.

Formula protection and full exports

Default-on spreadsheet protection modifies exported header/data strings: an apostrophe is prefixed to values starting with =, +, -, @, tab or carriage return, and leading whitespace before a formula prefix. Negative numeric-looking strings may change too. The result shows the affected-cell count and up to ten before/after examples. Turn protection off explicitly for raw string values. Neither CSV quoting nor protection guarantees safe interpretation in every spreadsheet program.

Downloads use standard quoting, doubled quotes and CRLF record separators while preserving line breaks inside cells. Byte-level quoting and separator formatting can differ from the input even when parsed cell values remain identical. UTF-8 BOM is optional. The output MIME type and extension follow the output delimiter: original-columns.csv, or original-columns.tsv for tab.

Original and transformed tables show up to 50 records, 20 columns and 160 characters per cell/label. The text preview shows up to 8,192 characters. Download always contains the complete accepted output; Copy CSV is available up to 1 MiB, with complete selectable output if clipboard permission is denied. Larger output is download-only.

Limits, cancellation and recovery

Limits are 5 MiB input, 50,000 data records, 200 columns and 20 MiB serialized output. Overflows fail explicitly without truncating records or returning partial exports. Explicit Parse and Build actions use local workers; the complete dataset is not copied for each label keystroke. Cancel terminates work and revision guards discard obsolete file, worker and clipboard replies. Workers and download URLs are released when invalidated or finished.

Failed imports preserve the prior draft, loaded table and mapping. If a pasted draft or input settings differ from the loaded table, export is disabled; Parse again or use Restore loaded input. Clear / Reset offers one-step Undo of input, mapping and settings, not a falsely current export. Refresh clears the draft. No joins, filters, formulas, row deletion, inferred types or Excel/OpenDocument workbooks are supported.