In this article
What Is JSON Lines?
JSON Lines (also called JSONL or Newline-Delimited JSON, NDJSON) is a text format where each line is a valid JSON object. Unlike regular JSON which wraps everything in an array, JSONL puts one record per line with no enclosing brackets. This makes it ideal for streaming, appending, and processing large datasets line by line.
A typical JSONL file looks like this: each line contains a complete JSON object with its own keys and values. There are no commas between lines and no wrapping array. This simplicity is what makes the format powerful for data pipelines, log files, and machine learning datasets.
Why JSONL Matters
JSONL has become the preferred format for several important use cases in modern data engineering and cloud platforms.
- Streaming processing -- because each line is independent, you can process files line by line without loading the entire dataset into memory, critical for multi-gigabyte files
- BigQuery and cloud imports -- Google BigQuery, Amazon Athena, and other cloud data warehouses natively support JSONL for data loading, making it the standard format for cloud data imports
- Machine learning datasets -- many ML frameworks including Hugging Face, OpenAI fine-tuning, and BERT training expect data in JSONL format with one example per line
- Log files and event streams -- application logs often use JSONL because new events can be appended to a file without modifying existing content or maintaining array syntax
Try it free — no signup required
Convert CSV to JSONL →Converting CSV to JSONL
Converting CSV to JSONL maps each row in your CSV file to a JSON object where the column headers become the keys.
- Headers as keys -- the first row of your CSV file provides the property names for each JSON object, so column headers should be clean and consistent
- Delimiter detection -- the converter handles comma, semicolon, tab, and pipe delimiters, auto-detecting the format or letting you specify it explicitly
- Type inference -- numeric values are converted to JSON numbers instead of strings, and empty cells can be output as null or omitted entirely depending on your preference
Frequently Asked Questions
What is the difference between JSONL and regular JSON?
Regular JSON wraps all records in an array with square brackets and separates them with commas. JSONL puts one JSON object per line with no array wrapper and no commas between records. This means you can append to a JSONL file without modifying existing content, process it line by line without a full parser, and stream it without buffering the entire file.
When should I use JSONL instead of CSV?
Use JSONL when your data has nested structures, inconsistent columns, or needs to be consumed by APIs and cloud services. CSV is better for simple tabular data that will be opened in spreadsheets. JSONL preserves data types (numbers, booleans, null) while CSV treats everything as text. JSONL also handles values containing commas and newlines without escaping issues.
What file extension should I use?
The most common extensions are .jsonl and .ndjson. Some tools also accept .json with one object per line. Use .jsonl for clarity since it immediately tells anyone looking at the file that it is newline-delimited rather than standard JSON. BigQuery and most cloud tools accept all three extensions.
Can JSONL files be streamed?
Yes, that is one of the primary advantages of JSONL. Because each line is independent and self-contained, you can read and process one line at a time using simple readline logic in any programming language. This makes JSONL ideal for processing datasets that are too large to fit in memory, unlike regular JSON which requires parsing the complete array structure.