19 releases (7 breaking)

0.8.1 Mar 1, 2023
0.7.2 Feb 2, 2023
0.6.0 Oct 2, 2022
0.3.0 May 19, 2022
0.1.4 Mar 29, 2021

#248 in Encoding

Download history 8/week @ 2022-12-02 8/week @ 2022-12-09 19/week @ 2022-12-16 9/week @ 2022-12-23 9/week @ 2022-12-30 30/week @ 2023-01-06 25/week @ 2023-01-13 12/week @ 2023-01-20 46/week @ 2023-01-27 47/week @ 2023-02-03 56/week @ 2023-02-10 73/week @ 2023-02-17 54/week @ 2023-02-24 22/week @ 2023-03-03 11/week @ 2023-03-10 46/week @ 2023-03-17

137 downloads per month


195 lines

CSV to Parquet


Convert CSV files to Apache Parquet. This package is part of Arrow CLI tools.


Download prebuilt binaries

You can get the latest releases from https://github.com/domoritz/arrow-tools/releases.

With Cargo

cargo install csv2parquet

With Cargo B(inary)Install

To avoid re-compilation and speed up installation, you can install this tool with cargo binstall:

cargo binstall csv2parquet


Usage: csv2parquet [OPTIONS] <CSV> <PARQUET>
  <CSV>      Input CSV file
  <PARQUET>  Output file
  -s, --schema-file <SCHEMA_FILE>
          File with Arrow schema in JSON format
      --max-read-records <MAX_READ_RECORDS>
          The number of records to infer the schema from. All rows if not present. Setting max-read-records to zero will stop schema inference and all columns will be string typed
      --header <HEADER>
          Set whether the CSV file has headers [possible values: true, false]
  -d, --delimiter <DELIMITER>
          Set the CSV file's column delimiter as a byte character [default: ,]
  -c, --compression <COMPRESSION>
          Set the compression [possible values: uncompressed, snappy, gzip, lzo, brotli, lz4, zstd]
  -e, --encoding <ENCODING>
          Sets encoding for any column [possible values: plain, rle, bit-packed, delta-binary-packed, delta-length-byte-array, delta-byte-array, rle-dictionary]
      --data-pagesize-limit <DATA_PAGESIZE_LIMIT>
          Sets data page size limit
      --dictionary-pagesize-limit <DICTIONARY_PAGESIZE_LIMIT>
          Sets dictionary page size limit
      --write-batch-size <WRITE_BATCH_SIZE>
          Sets write batch size
      --max-row-group-size <MAX_ROW_GROUP_SIZE>
          Sets max size for a row group
      --created-by <CREATED_BY>
          Sets "created by" property
          Sets flag to enable/disable dictionary encoding for any column
      --statistics <STATISTICS>
          Sets flag to enable/disable statistics for any column [possible values: none, chunk, page]
      --max-statistics-size <MAX_STATISTICS_SIZE>
          Sets max statistics size for any column. Applicable only if statistics are enabled
  -p, --print-schema
          Print the schema to stderr
  -n, --dry
          Only print the schema
  -h, --help
          Print help information
  -V, --version
          Print version information

The --schema-file option uses the same file format as --dry and --print-schema.


Convert a CSV to Parquet

csv2parquet data.csv data.parquet

Convert a CSV with no header to Parquet

csv2parquet --header false <CSV> <PARQUET>

Get the schema from a CSV with header

csv2parquet --header true --dry <CSV> <PARQUET>

Convert a CSV using schema-file to Parquet

Below is an example of the schema-file content:

  "fields": [
      "name": "col1",
      "data_type": "Utf8",
      "nullable": false,
      "dict_id": 0,
      "dict_is_ordered": false,
      "metadata": {}
      "name": " col2",
      "data_type": "Utf8",
      "nullable": false,
      "dict_id": 0,
      "dict_is_ordered": false,
      "metadata": {}
  " metadata": {}

Then add the schema-file schema.json in the command:

csv2parquet --header false --schema-file schema.json <CSV> <PARQUET>


~1M SLoC