How to use the Fix a CSV File
- 1 Drop your CSV file on the studio, or try the example.
- 2 Read the notes under Source: wrong delimiter, rows with too many or too few fields, an unclosed quote, garbled characters.
- 3 Change the delimiter or encoding under Source if the preview still looks wrong.
- 4 Keep the clean-up steps, adjust them, and download the repaired file.
What “broken” usually means
A CSV file is plain text, so it is rarely corrupted in the sense of being unreadable. It usually breaks in one of a few ways: the delimiter is not what the program expects, a quote is never closed and swallows the rest of the file, some rows have more or fewer fields than the header, the text was saved in a different encoding than it is read in, or there are blank lines and stray spaces from an export.
The studio reads the file with a forgiving parser and reports each of these instead of failing. An unclosed quote is treated as an ordinary character and its line number is shown. Rows that are too short are padded with empty fields; rows that are too long get extra columns, so no data is lost – you can also choose to join the extras into the last column or drop them.
Fixing it
The example pipeline removes completely blank rows and columns and trims spaces and invisible characters from every value. Add steps for the rest: remove duplicate rows, change the delimiter or the encoding, or restore leading zeros.
Limits
The parser follows RFC 4180, the closest thing CSV has to a standard, but real files vary, so the repairs are heuristics: check the preview. A file over 25 MB, or with more than 4 million cells, is refused rather than risking your browser.
Frequently asked questions
Why does my CSV look fine in Notepad but wrong in Excel?
Excel splits columns with the list separator of your regional settings and assumes an encoding when it opens the file. A semicolon-separated file in a region that expects commas, or UTF-8 without a byte-order mark, shows wrongly. Fix the delimiter or export UTF-8 with a BOM.
Is any data changed when it repairs rows?
Missing fields become empty and extra fields become extra columns. Nothing is discarded unless you choose to drop extra fields.
Can it repair a file with a mix of delimiters?
It picks the delimiter that splits most rows the same way and warns when that is a guess. If a file really mixes them, choose the main one under Source and fix the rest with a split step.
Last updated: