Output · Files
Write Parquet
Saves the table as a compressed Parquet file that reloads fast here, in DuckDB, pandas or Spark. Good for large in-between results.
Also known as: Columnar export
How it works
Type the Parquet file name (output.parquet by default) and pick a Location; blank means the project folder. Compression is Snappy by default, with Gzip, Zstd or None as alternatives. The write fails if the file already exists unless you tick Overwrite if the file already exists. Column types are kept as they are, so the file reloads with the same schema in a Parquet Reader, DuckDB, pandas or Spark. The written file is passed on downstream as a File.
Inputs and outputs
Inputs
data— Table
Outputs
file— File
Settings
The Attributes panel of Write Parquet, as in the app: the controls work, but nothing runs — there is no data behind it.
The same settings, as the Assistant writes them
{
fileName: string
directory?: string
compression: "snappy"|"gzip"|"zstd"|"uncompressed"
overwrite?: boolean
}fileName is a basename, e.g. "sales.parquet"; directory empty = the project folder; overwrite off (default) fails rather than replacing an existing file
Example
A long cleaning stage feeds several reports. Write its result to clean-orders.parquet with Overwrite ticked, and start each report pipeline from a Parquet Reader on that file instead of repeating the cleaning. Choose Zstd when the file goes over a slow link and size matters more than write speed.