Data file output

This step writes the incoming stream to a file. The file it produces can be read back with the Data file input step.

The file format is chosen automatically from the filename extension — see the table below.

Supported file types

Extension	Format
.ndjson, .json	Newline-delimited JSON. The first line is a JSON schema describing the fields; each following line is one JSON array of values, in schema order (missing values are written as null). Field names are only stored once, in the schema line.
.csv	Comma-separated values (RFC 4180). The first line is the same JSON schema; each following line is one comma-separated record. Dates are written as ISO-8601 in UTC (for example 2026-07-06T08:01:37.000Z), the decimal separator is ., and the file is UTF-8.
.ndjson.gzip, .csv.gzip	The ndjson / csv formats above, gzip-compressed. .gz works as well.
any other extension	Cube — a gzip-compressed binary format holding the Apache Hop row metadata and rows.

Note

The ndjson and csv formats are self-describing: the schema line carries the field names, types and lengths, so Data file input can rebuild the exact field layout when reading the file back.

File options

Option	Description
Transform name	Name of the transform. This name has to be unique in a single pipeline.
Filename	The filename and location of the output file. The extension decides the file format (see above).
Do not create file at start	Enable to avoid an empty file when no rows are processed; the file is created when the first row arrives.
Add filenames to result	Adds the written filename to the internal result filename set for further processing.