CsvLens · CSV/TSV viewer · zero upload

The CSV opens in this tab.
Quoted commas, line breaks, EUC-KR.
Nothing uploaded.

A comma inside a quoted field, a line break inside a cell, a doubled "" — the parser follows RFC 4180 instead of splitting on the delimiter and hoping. The delimiter is detected, the bytes are decoded as UTF-8 or EUC-KR in the browser, and the table underneath searches, sorts and counts every column. Then it leaves as JSON, a Markdown table, or a UTF-8 BOM CSV that Excel opens without breaking Korean.

0 bytes uploaded 0 external requests RFC 4180 parser UTF-8 · EUC-KR

1 · Bring the table in

read in memory · never sent

Drop a .csv or .tsv file

It is read into this tab's memory and parsed here.

No file yet — nothing has been read.

No table yet

2 · Search, sort, read

click a column name to sort
0 rows

Nothing loaded yet — drop a file or paste text above.

Table summary

  • Data rows—
  • Columns—
  • Delimiter—
  • Encoding—
  • Size—

Counts and exports cover every row, not only the rows drawn in the table.

Columns

Load a table to see its columns.

Pick a column for its statistics.

3 · Take it out

JSON · Markdown · BOM CSV

A JSON array of objects keyed by the header row. Duplicate header names get a numeric suffix so no value is lost.

—
Waiting for a table

The CSV download starts with EF BB BF (the UTF-8 BOM) and uses CRLF line endings, which is what makes Excel read Korean instead of code-page mojibake. Fields are quoted only when they contain a delimiter, a quote or a line break, and doubled quotes are written back as "".

How it reads a file

three details that decide correctness

01Quotes win over delimiters

The parser walks the text character by character with a single piece of state: whether it is inside a quoted field. Inside, a comma is text, a line break is text, and "" becomes one quote. Outside, a quote only opens a field when the field is still empty, which is why an apostrophe in O'Brien does not swallow the rest of the row.

02The delimiter is measured, not assumed

Four candidates — comma, tab, semicolon, pipe — are tried against the first rows and scored by how many fields they produce and how consistently they produce that count. A tie goes to the comma. The summary always shows which one won, and the picker beside it overrides it when a file is unusual.

03Bytes before characters

A CSV has no encoding label, so the bytes are kept and decoded here with TextDecoder: UTF-8 by default, EUC-KR when an older Korean system wrote it. Switching encoding re-decodes the original bytes, so a wrong guess costs one click rather than a round trip through another converter.

Questions worth asking

privacy · Excel · size
Is my CSV uploaded anywhere?
No. The file is read with the File API as bytes, decoded with TextDecoder and parsed in JavaScript in this tab. There is no server call, no analytics, no storage and no cookie, and the footer counts cross-origin requests with a PerformanceObserver and stays at 0 — the one honest way to show that instead of just asserting it. A CSV often carries customer names, salaries or prices, so a free viewer that uploads it is a real disclosure.
Why does Excel break the Korean?
Without a byte order mark Excel guesses the encoding, and on a Korean Windows machine it tends to guess the local code page, which turns UTF-8 Korean into mojibake. The CSV here starts with EF BB BF and uses CRLF, so Excel recognises UTF-8. The reverse case matters too: a file written in EUC-KR arrives as broken text in a reader that assumes UTF-8, which is what the encoding selector is for.
Why is everything in one column?
Either the delimiter is not one of the four detected, or the detection guessed from the first rows and the rest of the file disagrees. The delimiter chip in the summary shows the guess and the picker forces another; the table redraws immediately. A fixed-width file with no delimiter is not CSV and needs a fixed-width split, which this page does not pretend to do.
How big a file can it take?
The table draws the first 500 matching rows to keep scrolling smooth, while the statistics and all three exports use every row — a 100,000-row file is workable. Past roughly 20 MB the tab itself becomes the limit rather than any quota, and the status line says so instead of freezing silently. Nothing is truncated in what you download.