Fix mixed line endings in CSV

What are line endings and why do they get mixed?

Every line in a text file ends with an invisible character (or pair of characters) that marks where the next line begins. Windows uses CRLF (\r\n, two bytes: carriage-return + line-feed). Unix and macOS use LF (\n, one byte: line-feed only). RFC 4180 mandates CRLF for CSV, but most real-world files use LF.

Mixed line endings appear when files are combined from different sources: you concatenate a Windows export with a Unix export, a script processes a file and rewrites some rows with the wrong style, or a copy-paste operation introduces the wrong characters. The result looks identical in a text editor (which hides line endings), but parsers that read byte-by-byte see \r characters appearing mid-row and miscount columns.

How mixed line endings break CSV parsers

A CSV parser that expects \n as the row delimiter will, on a CRLF row, consume the \r as part of the last field value — so the final column of every row gains a trailing \r. A parser that expects CRLF and encounters a bare \n inside a quoted multi-line field may treat it as a row break. The specific failure mode depends on the parser, but common symptoms are:

• Last column of every row has a trailing \r character
• Row counts are wrong (extra empty rows or missing rows)
• "Unclosed quote" errors on rows that contain quoted multi-line text
• Misaligned columns after the first occurrence of a mixed ending

How CSViser detects and fixes this

CSViser uses a quote-aware scanner that strips quoted field content before counting line-ending styles — this avoids false positives from multi-line text fields that legitimately contain bare \n inside double quotes (which is valid per RFC 4180). Only actual row delimiters are inspected. If both \r\n and bare \n are found outside of quoted spans, the file is flagged as having mixed line endings.

The fix normalises all row delimiters to \n (LF), which is accepted by virtually all modern parsers. The file content and column structure are unchanged — only the row-terminator bytes are standardised.

Quick fix in Python or shell

# Python
with open("data.csv", newline="") as f:
    content = f.read()
content = content.replace("\r\n", "\n").replace("\r", "\n")
with open("data_fixed.csv", "w", newline="") as f:
    f.write(content)

# sed (macOS / Linux)
sed 's/\r//' data.csv > data_fixed.csv

Frequently asked questions