Convert CSV to JSON

Choose JSON types and output structure deliberately

CSV fields start as text. Converting them to JSON requires decisions about identifiers, numbers and missing values that CSV itself cannot express. Confirm the first-row setting before exporting: object keys come from the detected or selected headers. UseUse as header when you know the first record contains column names.

The exact type inference rules

With type inference enabled, each cell is inspected separately. There is no whole-column schema, so one column can contain both numbers and strings. Leading and trailing spaces are ignored for recognition; an ordinary text value that does not match a rule retains its original spaces.

Decimal conversion uses JavaScript numbers, not exact decimal arithmetic. Disable inference when monetary precision or the original notation must survive unchanged. Disabling it also keeps true and null as literal strings. A quoted CSV field is still eligible for inference: CSV quotation protects separators, not a JSON type.

Empty string, null and absent field are different

There are two controls, not three independent missing-value modes. With inference off, empty fields stay "" and whitespace remains unchanged; the empty-as-null checkbox has no effect. With inference on and empty-as-null off, empty or whitespace-only fields become "". With both enabled, they become null. There is no option to omit empty properties from object output.

A short row is a structural issue, not proof of an intentional null. Objects and JSON Lines supply an empty value for a missing trailing column and then apply those settings. Array output preserves the row's actual length. Entirely empty records are skipped during CSV parsing. Check malformed rows with the CSV validatorbefore treating an export as application-ready data.

A reproducible conversion example

id,active,amount,note
000184,true,12.50,
000219,false,0,null

With inference on and empty-as-null off, thenulls and booleans fixtureproduces this object array. Automated tests exercise the real worker export and parse its downloaded bytes to verify these values:

[
  { "id": "000184", "active": true, "amount": 12.5, "note": "" },
  { "id": "000219", "active": false, "amount": 0, "note": null }
]

Notice that 000184 stays an identifier, 12.50 loses its formatting zero, and the empty field differs from the word null. Theleading zeros fixtureprovides longer identifiers for a second check.

Object arrays, row arrays and JSON Lines

Choose objects when an API expects named properties. Choose arrays when a consumer expects positions: the first exported array contains headers, followed by data arrays. Array shape also preserves separate columns with identical labels. In objects or JSON Lines, exactly repeated header keys overwrite earlier values with the last column's value. Resolve duplicate headers before choosing a keyed format.

JSON Lines is useful for consumers that process one JSON object per line. It has no surrounding array or commas between records. A newline inside a CSV field becomes an escaped \n, so one record still occupies one output line:

{"id":"001","note":"first\nsecond"}
{"id":"002","note":"done"}

This output choice does not remove the workbench's 250,000-row retention limit or make its conversion streaming. Export is blocked for preview-only sessions. For the reverse direction, use JSON to CSV; for nested objects, see thenested JSON conversion guide.