CSV to JSON Converter

CSV (comma-separated values) is a tabular format defined loosely by RFC 4180 — most parsers accept variant quoting and line endings. JSON is a hierarchical format used by every modern API. The standard mapping treats the first CSV row as field names and each subsequent row as an object whose keys are those field names.

This converter handles quoted strings with embedded commas and newlines, escapes, and BOM-prefixed UTF-8.

CSV is older than the term itself; tabular delimited text dates to the 1970s. RFC 4180, published in 2005, finally tried to standardize what was already in the wild — by then dozens of dialects existed. The spec specifies comma delimiter, double-quote string quoting, CRLF line endings, and no trailing newline. Real-world files violate every one of these constantly: tab and semicolon delimiters, single-quoted strings, mixed line endings, escaped commas inside quoted fields, BOM markers from Excel exports. A reliable parser handles all of these or detects and refuses to. JSON is the cleaner format for hierarchical data — nested objects, arrays of objects, multiple types per field — but loses the human readability of a flat CSV.

A worked example: input CSV `name,age,city\nAlice,30,"New York, NY"\nBob,25,"Tokyo"`. Output JSON: `[{"name": "Alice", "age": "30", "city": "New York, NY"}, {"name": "Bob", "age": "25", "city": "Tokyo"}]`. Note the comma inside Alice's city — preserved correctly because the field is quoted. With type inference enabled, age becomes the number 30 instead of the string "30". The first row defines field names; subsequent rows map to those names. If the CSV has no header, the converter offers a choice: assign generic keys (column_1, column_2) or treat the first row as data with auto-generated keys.

Limitations: ambiguous CSV (no spec compliance) requires guessing. The converter detects common dialects but malformed files (mixed delimiters, escaped quotes done wrong) need manual fixing. Type inference is convenient but lossy: leading-zero IDs ("00123") become numbers and lose the zero; date-like strings might or might not be parsed depending on detection. For schema-strict use, disable inference and apply types explicitly. Browser-side processing limits file size — beyond ~100 MB, performance degrades and memory pressure grows. Server-side or streaming parsers (Node CSV, Python pandas) handle multi-gigabyte files. The converter is for interactive use, not production ETL.