Last updated 12 October 2026
File formats
| Format | Extensions | Support |
|---|---|---|
| Delimited text | .csv .tsv .psv | Read and written |
| JSON | .json .jsonl | Read and written |
| Stata, SPSS and SAS | .dta .sav .zsav .por .sas7bdat .xpt | Stored and restored byte for byte |
| Parquet | .parquet | Through a separate converter |
Delimited text
The delimiter is detected from the header row. Comma, tab and pipe are recognised, as well as other single-character delimiters.
JSON
Polypress reads and writes an array of records, an object of column arrays, or JSON Lines (one record per line).
Stata, SPSS and SAS
The original file is stored and restored byte for byte, including value labels. Polypress reads these files with the ReadStat library, so an archive of a statistical package file can also be restored to delimited text or JSON. In that case Polypress reports any information that the target format cannot hold, such as value labels or display formats. Polypress does not create Stata, SPSS or SAS files from other formats.
Parquet
Parquet files are converted to and from CSV with the pyarrow library. With a pip installation,
the converter is the polypress-parquet command:
# compress
polypress-parquet to-csv data.parquet \
| polypress compress - -o data.ppz
# restore
polypress restore data.ppz -o - \
| polypress-parquet from-csv - data.parquet
In a source build, the same converter is python3 work/py/parquet.py.
A restored Parquet file is not identical to the original
Its columns are stored as text, and the original column types, compression settings and metadata are not kept.