CSV ↔ JSON Converter

How to convert between CSV and JSON

1. 1

   Pick a direction

   Choose CSV to JSON or JSON to CSV. Each tab keeps its own input, settings and result.

2. 2

   Add your data

   Type or paste it, import a local UTF-8 file, or load the example. Check the delimiter and options.

3. 3

   Convert and use the result

   Press Convert, read the summary, then copy the output, download it, or use it as the input of the other tab.

CSV to JSON

• Every value stays a string. The result is an array of objects, and every value is a JSON string exactly as written: zip codes such as 01234, numbers such as 1.50, spaces at either end, and empty cells (which become ""). Numbers, dates, true/false and null are never guessed. Convert them in the program that reads the JSON.
• Delimiters. Comma, semicolon and tab. Auto-detect counts each delimiter in the first row, outside quotes, and picks the most frequent one. The detected delimiter is shown before and after converting, and you can choose another. If only one column results but the first row holds another delimiter, the summary points that out.
• Column names. By default the first row holds the names and becomes the keys. Empty names and repeated names are errors, because an object cannot hold the same key twice; nothing is renamed or overwritten. Without a header row the keys column1, column2 and so on are generated. A name such as __proto__ is an ordinary key.
• What the CSV reader accepts. Values may be wrapped in double quotes and then contain delimiters, doubled quotes and line breaks. A UTF-8 byte order mark is ignored, and CRLF, LF and CR line breaks all work. Quotes in the wrong place, an unclosed quote, and rows with a different number of values than the first row are errors with the record number and line.
• Empty lines. One line break at the very end is ignored. An empty line anywhere else is a record: in a CSV with several columns it is an error (remove the line), and in a one-column CSV it is an empty value. Empty lines are never dropped quietly.
• Header only. A CSV with only a header row converts to an empty array, and the summary says 0 rows.

JSON to CSV

• Flat objects only. The JSON must be an array of flat objects, one object per row: strings, numbers, true/false and null. Columns are all keys in the order they first appear, and rows keep their order. A nested object or array is refused with its location and a reason; it is not flattened or turned into text for you. So are repeated keys in one object, items that are not objects, and a single object outside an array.
• An empty array. An empty array has no keys, so no columns can be inferred. You get an explanation and no file, not an empty or header-less CSV.
• CSV cannot keep every distinction. A missing key, null and an empty string all become an empty cell, and the summary counts the missing and null ones. The number 1 and the text "1" look the same, and so do true and the text "true". Strings keep their leading zeros and spaces, and numbers are written exactly as in the JSON (12345678901234567890, 1.50 and 1e30 do not change). A spreadsheet program may still change a cell such as 01234 when it opens the file.
• Quoting and delimiter. Choose comma, semicolon or tab, and whether to write the header row. A value is wrapped in double quotes when it contains the delimiter, a quote, a line break, or spaces at either end, and quotes inside are doubled. Line breaks in the output are LF.
• Protect against spreadsheet formulas. On by default, under Advanced settings. Text values and headers that start with =, +, -, @, a tab or a carriage return get an apostrophe in front, the approach OWASP recommends against CSV injection. This changes the exported text, so =1+1 becomes '=1+1. Numbers in the JSON, and text that is itself a plain number such as -5, are not changed. It lowers the risk but does not promise safety in every spreadsheet program; turn it off if you need the text unchanged.
• Byte order mark. Optional, off by default. It writes the three bytes that mark a file as UTF-8, which helps some spreadsheet programs show accents and non-English text correctly. The mark is invisible in the output box but is included when you copy or download.

Results, errors and the other tab

• Errors are never a quiet empty result. If the input cannot be converted you get an error and the output is cleared. The message says what is wrong, for example “Record 3 has 2 values, but record 1 has 3”, with the line and a Go to error button. Nothing is repaired or guessed. The first problem is shown; fix it and convert again.
• Out of date output. Nothing converts while you type. If you change the input or a setting afterwards, the output and summary are marked out of date until you press Convert again. Copy and Download always give the output you see, made with the settings shown in the summary.
• Use the output in the other tab. Moves a successful result into the input of the other direction, so you can check a round trip. Undo brings the previous input of that tab back, also after Clear, Import or Load Example.
• Files. Import a UTF-8 .csv, .tsv or .txt file for CSV, or a .json or .txt file for JSON. The file is read in your browser and not uploaded. Downloads are UTF-8 files named converted.json or converted.csv.

Limits

maximum input size

5 MB

maximum rows

100,000

maximum columns

500

Larger input, more rows or columns, or a result of more than 16,000,000 characters is refused with a message rather than cut. JSON can be nested at most 128 levels deep before it is refused, though only flat objects convert. Conversion runs in a background thread (a Web Worker), so the page stays responsive, and a newer conversion replaces an older one that is still running.

Frequently asked questions