JSON Lines and NDJSON Formatter

What JSON Lines is and why it exists

JSON Lines (also called NDJSON or LDJSON) stores one complete JSON value per line, separated by newlines instead of being wrapped in an outer array. That small change makes the file streamable: a consumer can read, process and discard one record at a time without loading everything into memory, and a producer can append a record without rewriting the file. You meet it in structured application logs, BigQuery and Snowflake exports, Elasticsearch’s _bulk API, OpenAI fine-tuning and batch files, docker logs piped through jq -c, and data pipelines that pass events between jobs. The JSON Lines vs JSON guide covers when each layout is the better choice.

Formatting a .jsonl file in the browser

Drop the file onto the editor or paste a chunk of a log. The format is recognised from its shape, so a stream of objects lands here rather than in the plain JSON formatter. The output refreshes while you type; for very large pastes press Ctrl/Cmd+Enter to run it once. Ctrl/Cmd+Shift+M forces the compact layout and Ctrl/Cmd+Shift+C copies the result.

Besides the code view there are two others. Tree shows the stream as an array, one node per record, so you can fold nested payloads. Table turns the records into rows with one column per key, which is the quickest way to scan a few thousand log events for a failing request; clicking a row jumps to its line in the input. The Tools button adds JSONPath and jq queries, schema validation and key statistics over the whole stream. Parsing is done by a lossless line-by-line JSON parser running locally, so production logs with customer emails or tokens are never uploaded.

The Records option and the toolbar

• Records has two settings. One compact record per line (valid NDJSON) strips all insignificant whitespace and writes every record on exactly one line, which is what loaders and _bulk endpoints expect. Pretty-print each record (for reading) expands each record over several indented lines; the result is easier on the eyes but is no longer valid JSON Lines, so switch back before saving.
• Indent (2 spaces, 4 spaces or tabs) only affects the pretty-printed setting.
• Sort keys orders object keys alphabetically inside every record, nested objects included. Records keep their original order, which matters for logs and for files where line order means time.
• Minify always produces the compact layout, whatever Records is set to.

Numbers are copied from the input as written, so a 19-digit order ID or a price written as 10.50 comes out unchanged, and non-ASCII text stays readable instead of being turned into \u escapes.

How messy input is handled

The parser reads the input as a sequence of JSON values rather than splitting blindly on newlines. That makes it forgiving in useful ways:

• A record that was pretty-printed across several lines is accepted and joined back onto one line, with an info note naming the line it started on.
• Two records that ended up on the same line are separated onto their own lines.
• Blank lines between records are dropped.
• A top-level JSON array counts as one record and stays on a single line. To explode an array into one line per element, use JSON to NDJSON; to go the other way, NDJSON to JSON wraps the records in an array.

When a record is broken, every bad line is reported (up to 50), each message prefixed with “Record on line N” so you can jump straight to it. A truncated last line, which is common when a log is copied while still being written, shows up as an unexpected end of input on that line.

Examples

{ "id": "ord_1001", "customer": "c_204", "total": 59.9, "status": "paid" }
{ "id": "ord_1002", "customer": "c_311", "total": 12.5, "status": "refunded" }
{ "id": "ord_1003", "customer": "c_204", "total": 230, "status": "paid" }

{"id":"ord_1001","customer":"c_204","total":59.9,"status":"paid"}
{"id":"ord_1002","customer":"c_311","total":12.5,"status":"refunded"}
{"id":"ord_1003","customer":"c_204","total":230,"status":"paid"}

{"user":"u_17","event":"signup","plan":"pro","utm":{"source":"newsletter","campaign":"sept"}}
{"user":"u_18","event":"login","device":{"os":"iOS","version":"18.1"}}

{
  "user": "u_17",
  "event": "signup",
  "plan": "pro",
  "utm": {
    "source": "newsletter",
    "campaign": "sept"
  }
}
{
  "user": "u_18",
  "event": "login",
  "device": {
    "os": "iOS",
    "version": "18.1"
  }
}

{
  "id": "evt_1",
  "type": "deploy",
  "service": "checkout"
}
{
  "id": "evt_2",
  "type": "rollback",
  "service": "checkout"
}

{"id":"evt_1","type":"deploy","service":"checkout"}
{"id":"evt_2","type":"rollback","service":"checkout"}

{"status":"paid","id":"ord_1001","total":59.9}
{"total":12.5,"id":"ord_1002","status":"refunded"}

{"id":"ord_1001","status":"paid","total":59.9}
{"id":"ord_1002","status":"refunded","total":12.5}

Common errors and how to fix them

Related tools

• ndjson validator
• JSON Lines to JSON
• JSON to JSON Lines
• json lines vs json
• JSON formatter
• Log lines (Apache, Nginx, JSON logs) formatter
• CSV formatter