Glosarium
Contents

Data Engineering › Storage, Formats & Lakehouse

JSON Lines

One JSON object per line, easy to stream and append.

Data Engineer · Junior

Also known as: JSONL, NDJSON, newline-delimited JSON

JSON Lines (also called JSONL or NDJSON, for newline-delimited JSON) is a file format with one complete JSON object per line.

{"event": "signup", "user_id": 1, "ts": "2026-03-01T09:00:00Z"}
{"event": "login", "user_id": 1, "ts": "2026-03-01T09:05:12Z"}
{"event": "purchase", "user_id": 1, "amount": 19.99}

Compare with a normal JSON file holding an array: [ {...}, {...}, {...} ]. That file must be read as a whole before you know it is valid, and appending means editing the end of the file.

Why data teams like it

  • Streams well. Read it one line at a time without loading it all into memory.
  • Easy to append. New records are just new lines.
  • Fault tolerant. A bad line affects only that record, not the whole file.
  • Splittable. A big file can be cut at line boundaries and processed in parallel.
  • Flexible. Each line can have different fields, which suits logs and events.
import json

with open("events.jsonl") as f:
    for line in f:
        record = json.loads(line)
        print(record["event"])

Things to watch

  • Newlines inside values must be escaped (\n); each record has to stay on one physical line. Standard JSON serializers do this for you.
  • No schema. Types can drift between lines ("amount": "19.99" vs 19.99), and fields come and go. Validate on ingestion. See schema drift.
  • Bigger than binary formats. Field names repeat on each line and it compresses less efficiently than columnar formats. For analytics at scale, convert to Parquet. See row vs columnar formats.
  • Compressing the files (e.g. gzip) saves space, at the cost of splitting.

It is a good “landing” format for raw event data.