CSV to JSON

CSV is harder than it looks, and every value here stays a string

Many converters guess types; this one deliberately does not. A converter that "helpfully" turns 007 into 7 and 3/4 into a date has destroyed your data while appearing to succeed. This one converts nothing, so you get exactly the characters that were in the cell — and if your code wants numbers, it casts them itself, where you can see it happening.

The eleven things that break CSV, and which of them this handles

Written by reading the parser, not by guessing.

Two more worth knowing. If a row has fewer cells than there are headers, the missing keys appear with an empty string rather than being left out — so every object in the array has the same shape. If a row has more cells than headers, the extras are dropped, because there is no name to file them under.

A deliberately awkward CSV

Everything below is invented. It packs four of the traps into three rows: a comma inside a quoted name, a doubled quote, a line break inside a field, and a postcode that begins with a zero.

The result is three objects. ref is "A-1", customer is "Okonkwo, Ada" with the comma intact, note on the second row is wants a 6" bracket, and postcode on the first row is the string "01810" — not the number 1810. Paste the result into the JSON formatter if you want to check the nesting, or run it back through JSON to CSV to confirm nothing was lost on the way in.

The file that opens as a single column

This tool detects comma, semicolon, tab and pipe, so a semicolon file converts as it is. The background still matters when a file moves between other programs: Excel writes .csv files using the list separator from the Windows regional settings, and in places where the comma is the decimal mark that separator is often a semicolon. The file is still called .csv. It is not comma-separated.

Three ways out, best first:

• Re-export properly. In Google Sheets, Download → Comma-separated values. In Excel the separator follows the Windows list separator even for the CSV formats, so open the saved file in a text editor and check before sending it on.
• Change the separator at the source. The separator Excel writes follows the list separator in the operating system's regional settings, so changing that setting and re-exporting produces a genuinely comma-separated file. Worth doing once if you export regularly.
• Swap the characters by hand with find and replace — but only after checking that no field contains a comma of its own. If any value does, replacing every ; with , will corrupt the row count in a way that is very hard to spot afterwards. It is a last resort, not a shortcut.

And the leading zeros were probably already gone

If your postcodes or product codes arrive here as 1810 instead of 01810, this tool did not eat the zero — it cannot, because it never parses numbers. Excel ate it when the file was opened, long before you copied anything. Reopening a CSV in a spreadsheet and saving it again is a lossy operation. Take the export straight from the source system to here, and skip the round trip through a spreadsheet entirely.

What this is not for

It is a flat converter: one CSV row becomes one flat JSON object, always. Nesting cannot be expressed — a column called address.city produces a key literally named address.city. There is no type inference, no date parsing, no schema and no validation, so a malformed row raises no warning; it just produces an object with empty strings in it.

Nor is it built for very large exports. The whole text sits in the textarea and the parser walks it character by character on every keystroke, so tens of thousands of rows make typing feel sticky and a file of hundreds of megabytes belongs in a script. To see how big the thing you are holding actually is, the character counter beats guessing.

Why it matters that this one runs locally

CSV is the format business data travels in, which means the file in your clipboard right now is unusually likely to be the sensitive kind: a customer export with names, email addresses and order totals; a payroll extract; a patient list; a leavers' report. That is the whole point of the format.

Pasting that into a converter hosted by someone you cannot name is a transfer of personal data to a third party you have no agreement with, no record of and no way to audit — under the GDPR, a processor relationship created by accident. It is the sort of thing that surfaces in a security questionnaire a year later, when nobody remembers which website they used.

This page has nothing in the path to leak to. The parser is about twenty lines of JavaScript with no fetch, no XMLHttpRequest and no analytics call in it. The check takes ten seconds: load the page, disconnect from the network, paste your CSV. The JSON still appears. Everything on the developer tools page is built the same way, so the output can go straight into JSON to YAML without touching a server.