NDJSON (newline-delimited JSON) is a text format where every line is one complete JSON value — usually an object — and lines are separated by \n. There is no enclosing array and no comma between records. JSON Lines (.jsonl) is the same idea under another name.

What is the difference between NDJSON and JSON?

A JSON document is one value, so a list of records is one array that a parser must read to the closing bracket before it can hand anything back. NDJSON is many documents, one per line, so a reader can process record by record, append to the file without rewriting it, and skip a corrupt line without losing the rest.

Is NDJSON the same as JSON Lines and JSONL?

For practical purposes, yes. Both put one JSON value per line, UTF-8 encoded, separated by newlines. The specifications differ in small details — JSON Lines explicitly allows \r\n line endings — but files written by one are read by the other's tools.

Can a JSON parser read an NDJSON file?

Not as a whole — JSON.parse on the file fails at the second line. Split on newlines and parse each non-empty line on its own, or wrap the lines in [ ] with commas between them to turn it into a regular JSON array.

NDJSON (JSON Lines)

NDJSON — newline-delimited JSON — is a text format with one complete JSON value on each line. No enclosing array, no commas between records. The same convention ships as JSON Lines, with the extension .jsonl.

{"id":1,"status":"passed","ms":412}
{"id":2,"status":"failed","ms":1290}
{"id":3,"status":"passed","ms":388}

The equivalent JSON is one array that must be read to the final ] before any of it is valid.

Why it exists

Need	JSON array	NDJSON
Process a 10 GB file	Load it all, or a streaming parser	Read a line, parse a line
Append a record	Rewrite the closing `]`	Write one line
Survive a truncated write	Whole file invalid	Lose the last line
Split across workers	Hard	`split -l`
`grep` / `wc -l` / `tail -f`	Meaningless	One record per match

That is why it is the default shape for structured logs, Elasticsearch's _bulk API, BigQuery and Snowflake loads, OpenAI-style batch and fine-tuning files, and most streaming HTTP responses.

The rules

One JSON value per line. Usually an object, but any value is legal.
No raw newlines inside a value — they must be escaped as \n, which every JSON serialiser already does when not pretty-printing.
UTF-8. A byte order mark at the start breaks the first line in strict parsers.
Blank lines are commonly skipped, but not every reader tolerates them.

Converting

NDJSON → JSON array: jq -s . file.ndjson, or in JavaScript text.split("\n").filter(Boolean).map(JSON.parse).
JSON array → NDJSON: jq -c '.[]' file.json.
CSV → NDJSON: the CSV to JSON converter has an NDJSON output shape.
NDJSON → table: JSON to HTML table and JSON to CSV both accept NDJSON input when the text isn't a single JSON document.

Making it readable

NDJSON is built for programs. When a person needs to look at the records — a failed batch, a slice of logs — a rendered table they can sort and comment on beats a 4,000-column-wide line in a terminal. Paste the lines into JSON to HTML table and share the result as a link.

Try it

Comma is free — unlimited reports, unlimited commenters, unlimited revision history.

Turn NDJSON into a shareable table →

Byte order mark · CSV injection · Mojibake
CSV to JSON · JSON formatter
Glossary