Parquet CLI via Docker and nanoparquet

Warning

This project is experimental.

Features

• Supports all Parquet files that https://github.com/r-lib/nanoparquet does.
• Small size: the Docker image is 34MB compressed.

Requirements

• Docker, on Windows, macOS or Linux.

Installation

Download the nanoparquet.cmd file and save it into a place that is on your PATH.

Linux and macOS

You could put the script in /usr/local/bin and name it nanoparquet. Make sure that it is executable.

curl -L -o /usr/local/bin/nanoparquet \
  https://raw.githubusercontent.com/gaborcsardi/nanoparquet-cli/refs/heads/main/nanoparquet.cmd
chmod +x /usr/local/bin/nanoparquet

Windows

There is no standard place for user-installed scripts on Windows, so it is up to you where to save nanoparquet.cmd. The example below saves it into tools within the user's home directory. Make sure you add this directory to the PATH.

Invoke-WebRequest https://raw.githubusercontent.com/gaborcsardi/nanoparquet-cli/refs/heads/main/nanoparquet.cmd -OutFile ~/tools/nanoparquet.cmd

Usage

Call the script without arguments to see the help screen:

$ nanoparquet
Error:
Usage:
  nanoparquet ( --help | --version )
  nanoparquet info <parquet-file>
  nanoparquet schema <parquet-file>
  nanoparquet export <parquet-file> [ <csv-file> ]

Commands:
  info    Show information about a Parquet file.
  schema  Show the schema of tha Parquet file.
  export  Export a Parquet file to a CSV file.

Arguments:
  <parquet-file>  Parquet file.
  <csv-file>      CSV file. Uses standard output if missing.

Options:
  --help     Print this help message.
  --version  Print nanoparquet version.
Execution halted

Input and output files

The nanoparquet script bind mounts the current directory into the nanoparquet container. So you can call nanoparquet on any file that is anywhere inside the current working directory.

Output files work the same way: they must reside somewhere within the working directory. For example:

nanoparquet export data/2024-01.parquet data/2024-01.csv

You can use backward or forward slashes as path separators on Windows. Paths will always use forward slashed in nanoparquet output.

📘 Licenses

GNU R

GPL-2 or GPL-3 © The R Foundation

The tools in this repo

MIT © Posit Software, PBC