JSON vs JSONL: differences and when to use each

JSON

A .json file is one JSON value, usually an array of objects or a wrapper object such as {"data": [...]}. The records sit inside outer brackets with commas between them:

[
  {"id": 1, "user": {"city": "Amsterdam"}, "price": 15},
  {"id": 2, "user": {"city": "Berlin"}, "price": 25}
]

JSONL

A .jsonl file, also called NDJSON (.ndjson) or .ldjson, drops the outer brackets and the commas. Each line is a complete JSON value, and a newline separates the records:

{"id": 1, "user": {"city": "Amsterdam"}, "price": 15}
{"id": 2, "user": {"city": "Berlin"}, "price": 25}

The last row matters for logs and exports. If a process crashes halfway through writing a .json array, the missing ] makes the whole file invalid. In a .jsonl file every line written before the crash still parses, and you lose only the last, incomplete line.

JSON

Use JSON for payloads small enough to parse in one go, or when the consumer expects one document tree: configuration files, REST API responses and nested documents that aren't a list of rows.

JSONL

Use JSONL for logs, event streams, database exports and files that keep growing. Because each line stands alone, you can split the file across workers, tail -f it while it is being written, or process it without loading it all into memory.

jq

Turn a JSON array into JSONL with -c for compact one-line output. Turn JSONL back into an array with -s (slurp), which reads everything into memory:

jq -c '.[]' data.json > data.jsonl
jq -s '.' data.jsonl > data.json

rows.page

Use rows.page when you want to filter the records before converting, or skip the command line:

1. Drop a .json, .jsonl, .ndjson or .ldjson file on rows.page, or click Paste data and paste it.
2. Optionally filter the rows or run a SQL query to keep only the records and columns you need.
3. Open Export and pick JSONL for one object per line, or JSON for one array of objects.

rows.page flattens nested objects into dot-path columns such as user.address.city when it opens a file, and its exports keep those dot-path keys instead of nesting them again; lists stay real arrays. If you need the nested tree untouched, use jq.

rows.page opens both formats with no account. DuckDB-WASM reads the file in your browser, and nothing is uploaded unless you press Share.

• Size: up to 1.5 GB per browser tab. A bigger .json file is refused with "Too big for this browser", while a bigger .jsonl file loads its first rows with the note "Loaded the first N rows: the file is X, more than the browser can hold."
• Wrapper objects: a single top-level object such as {"data": [...]} is unwrapped to its biggest list, up to 4 levels deep, with a note that the other top-level fields were skipped.
• Single records: click a row to open a side drawer with a JSON/Fields toggle and Copy as JSON. There is no whole-file tree view.
• Filter and SQL: type country:NL or price:>20, or a SQL statement against the table data with dot-path columns in double quotes, such as "user.city". SQL shows up to 1,000,000 result rows, and the grid loads 100 rows at a time as you scroll.