CSV ↔ JSON Converter

Delimiters, quoting and escaping (the RFC 4180 rules)

CSV looks trivial until a value contains a comma. The de-facto rules are set out in RFC 4180: fields are separated by commas, and any field containing a comma, a double-quote or a line break must be wrapped in double-quotes. A literal double-quote inside such a field is escaped by doubling it. Get this wrong and a single address field with a comma in it silently shifts every later column by one, corrupting the whole row without any error.

The 'C' in CSV is also a polite fiction. European locales that use a comma as the decimal separator often delimit with semicolons instead, and tab-separated (TSV) files are common exports from spreadsheets and databases. A good converter lets you pick the delimiter; assuming a comma when the file is semicolon-delimited produces one giant column. Always confirm the actual separator before trusting the parse.

Header rows and type-inference traps

• The first row is usually headers that become JSON keys — but not always. If your data legitimately starts with a row that looks like data, a converter that assumes a header will eat your first record. Check the toggle.
• CSV has no types — every cell is text. Inferring that "42" is a number and "true" is a boolean is a guess that's usually helpful and occasionally destructive.
• The classic data-loss bug: a ZIP code like 02118, a phone number, or an ID like 007 read as a number loses its leading zero and becomes 2118. Long numeric IDs can also exceed JavaScript's safe integer range and lose their last digits. Keep these as strings.
• Ambiguous values bite too: "NA" might be a country code (Namibia) or 'not available'; a bare 2026-06-19 might be parsed as a date or left as text depending on the tool. Know which your data means.

Flattening nested JSON — and what doesn't survive

JSON is a tree; CSV is a flat grid. Converting JSON to CSV only works cleanly when each record is a flat object with the same keys. Nested objects have to be flattened — typically by joining the path with a dot, so a nested user name becomes a column user.name. Arrays are the real problem: a record with a list of three tags has no natural single-row representation, so converters either join the array into one cell, explode it into multiple rows, or spill it across numbered columns. None of these is lossless in both directions.

The honest summary: CSV-to-JSON is a perfect round-trip only for flat, uniform, string-typed tabular data. The moment you have nesting, ragged records or types that matter, something has to give. For working with the JSON side directly — validating, formatting, querying — use the JSON Formatter; for a spreadsheet view of the same rows, open the Spreadsheet.