ANALYSIS DIAGNOSTIC DATA

CSV Repair / Validator

Drop in a CSV file. Real quote-aware parsing (not a naive comma-split) — detects the delimiter, finds unterminated quoted fields, rows with the wrong number of columns, duplicate header names, and mixed line endings. Reports exactly which row is wrong and why.

Input

Common error messages this diagnoses

If you're here because something threw one of these, this tool shows you exactly what's wrong and how much is recoverable:

  • "Error tokenizing data. C error: Expected X fields in line Y, saw Z" — (pandas) — a row with the wrong column count; this tool finds every such row, not just the first.
  • "unexpected number of columns" — same symptom, different tool's wording.
  • "EOF within quoted string" — a quoted field was opened but never closed before the file ended.

What this tool doesn't do

This validates CSV syntax — it doesn't know what your columns are supposed to contain, so it can't catch a wrong data type or a swapped column (a "date" column full of numbers passes just fine, since that's not a CSV structure error). Column-count mismatches are reported as warnings, not errors, because trailing-column omission is extremely common and often intentional. Delimiter detection is a best-effort sniff of the first 20 lines — pick a different one manually if it's wrong. No fixed CSV standard actually exists (RFC 4180 is a description of common practice, not something every tool follows), so a file can pass every check here and still be rejected by a stricter downstream parser (Excel, a specific database's bulk loader). Only the first 2,000 rows are shown in the preview table, though every row is checked. Very large files (hundreds of MB+) are read fully into memory like every other tool here, not streamed.