Streaming documents

Transform(string) assumes its input is a single document. For many documents at once, JoltTransformer (which the JoltJsonTransformer types inherit from) offers streaming methods.

One document per line

For NDJSON / JSON Lines, where each line holds one complete object:

Method	Description
TransformLines(TextReader reader, TextWriter writer, StreamingOptions? options)	Reads one line at a time, transforms each, and writes the output.
TransformLines(Stream input, Stream output, StreamingOptions? options)	Same, over streams.
TransformLinesAsync(TextReader reader, TextWriter writer, CancellationToken? cancellationToken, StreamingOptions? options)	Asynchronous reading and writing.
TransformLinesAsync(Stream input, Stream output, CancellationToken? cancellationToken, StreamingOptions? options)	Asynchronous, over streams.

Other sequence formats

Method	Description
TransformSequence(TextReader reader, TextWriter writer, StreamingOptions? options)	Reads one JSON object at a time, transforms each, and writes the output.
TransformSequence(Stream input, Stream output, StreamingOptions? options)	Same, over streams.
TransformSequenceAsync(TextReader reader, TextWriter writer, CancellationToken? cancellationToken, StreamingOptions? options)	Asynchronous reading and writing.
TransformSequenceAsync(Stream input, Stream output, CancellationToken? cancellationToken, StreamingOptions? options)	Asynchronous, over streams.

TransformSequence detects the input format automatically, trying these in order:

1. RFC 7464 JSON text sequences: objects separated by the 0x1E record separator byte
2. Whitespace-separated objects: prettified or not, separated by newlines or other whitespace
3. A JSON array of objects

Arrays load fully into memory

The JSON array format reads the entire array into memory before processing. Prefer one of the first two formats for large inputs.

Output format

By default, output uses the same format and delimiter as the input. To control it, pass a StreamingOptions, which lets you choose how each output document is written (single line, multi-line, and others) and which delimiter to use (RFC 7464, newline, and so on).