`repack`

Recompress a local file at maximum compression by default, without changing the data format. Use convert when you need a different format or a cloud URI.

Two paths:

Container codecs (.gz, .zst, .bz2, .xz, .lz4, …): stream-decompress and recompress. Same codec unless the output extension or --compression says otherwise.
Built-in compression (Parquet / ORC / AVRO): rewrite the file with format-native compression. Parquet and ORC default to zstd; AVRO defaults to zstandard.

Omitting OUTPUT writes to a sibling temp file and atomically replaces the input on success.

# In-place max recompress of a gzip file
undatum repack data.csv.gz

# Explicit output and a faster level
undatum repack data.jsonl.zst out.jsonl.zst --level 3

# Parquet → zstd (format-native compression)
undatum repack data.parquet out.parquet

# Wrap an uncompressed file into a codec container
undatum repack data.csv data.csv.zst

# Change container codec (gzip → zstd)
undatum repack data.csv.gz data.csv.zst

Key options:

--level / -l — compression level (overrides the default max profile). Applies to container codecs and Parquet; ORC and AVRO ignore --level.
--compression — override codec. For containers: gz, zst, bz2, xz, lz4, … For Parquet/ORC/AVRO: a format-native codec (for example zstd, snappy).
--progress / --no-progress — progress bar (default: on)
--verbose — verbose logging

Limitations:

Local files only. Remote URIs (s3://, gs://, az://) are rejected. Download first, or use convert.
Uncompressed input needs a codec. undatum repack data.csv fails. Point OUTPUT at a compressed path (data.csv.zst) or pass --compression.
Format is preserved. repack does not convert CSV to Parquet. Use convert for that, then repack if you want max native compression.
ZIP / 7z are available as codecs but are not the primary path. Prefer single-stream containers (gz, zst, bz2, xz, lz4).
Parquet repack requires pyarrow.

See the format support matrix for the codec list and extras (undatum[compression]).