Miller (mlr) is a command-line tool created by John Kerl that brings the power of awk, sed, cut, join, and sort to name-indexed data formats. Unlike traditional Unix text-processing tools that operate on positional fields, Miller works with key-value-pair data, making it natural and intuitive for processing CSV, TSV, JSON, JSON Lines, and positionally-indexed data.

The tool allows you to perform on-the-fly transformations: adding new fields derived from existing ones, dropping fields, sorting, statistical aggregation, pretty-printing, and format conversion. Its natural data structure is the insertion-ordered hash map, making it a perfect fit for working with structured tabular data from the command line.

Miller is written in Go and is designed for high performance. It supports streaming processing of large files, automatic type inference, and a rich expression language for data manipulation. You can chain operations using Miller verbs (head, sort, tac, uniq, group-by, join, stats1, etc.) to build complex data pipelines directly from the command line.

Key capabilities include format conversion between CSV, TSV, JSON, and JSON Lines; statistical aggregation with mean, median, percentiles, and more; record-level filtering with boolean expressions; field transformations with arithmetic and string functions; joins between multiple data files; and sampling from large datasets. Miller complements tools like R and Pandas by handling data cleaning and preparation directly in the shell.

The tool is available via package managers on Linux (apt, yum, snap), macOS (brew, port), and Windows (choco, winget, scoop). It has comprehensive documentation at miller.readthedocs.io and is dual-licensed under BSD-2-Clause. With over 9,000 GitHub stars, Miller is one of the most widely used command-line data processing tools in the ecosystem.