rush

R Scripting at the Command Line

rush brings R to the command line. Run an R expression, wrangle a CSV, query a Parquet file with SQL, or draw a plot right in your terminal, all without opening an R session or writing a script.

rush

R has a wonderful ecosystem for working with data, but reaching for it usually means launching R, loading packages, and reading files by hand. rush collapses that into a single command you can pipe, redirect, and drop into any shell pipeline. R becomes just another Unix tool, at home next to grep, awk, and jq.

Highlights

• Quick one-liners. Evaluate any R expression straight from the shell.
• Scaffold scripts. Use --dry-run to generate self-contained R scripts you can save, edit, and share.
• Pipeline-native. Reads from standard input and writes to standard output, so it composes with every other command-line tool.
• Reads what you have. CSV, Parquet, JSON/JSONL, Excel, Arrow IPC, DuckDB, SPSS/Stata/SAS, SQLite, YAML, TOML, XML, and more, chosen automatically by file extension.
• Converts between formats. The convert command converts files directly between any supported format.
• Query with SQL. The sql command runs DuckDB queries directly against your files, no import step required.
• Plots in your terminal. The plot command renders ggplot2 graphics as ANSI/ASCII art, or saves them to PNG, PDF, and more.
• Zero dependency management. Packages are resolved on demand by ir.

How it works

Under the hood, each invocation assembles a small, self-contained R script and runs it with ir, the R interpreter launcher from r-lib. The script declares its own package dependencies in #| frontmatter, and ir installs whatever it needs on the fly. That means rush itself stays lightweight, and you never have to manage a library of packages just to run a one-liner.

TipReady to try it?

Head over to Get started for installation instructions and a quick tour.

Related tools

rush is, to our knowledge, the only CLI tool that combines R expression evaluation, automatic package resolution, multi-format I/O, SQL queries, and plotting in a single command. It also doubles as a script scaffolder: every invocation can emit a self-contained R script via --dry-run, giving you a starting point you can save, tweak, and version-control. That said, several excellent tools cover parts of this space, often faster or with fewer dependencies:

• DuckDB CLI: An embedded SQL engine that queries CSV, Parquet, and JSON files directly. Faster for pure SQL workloads; no R expressions or plotting.
• Miller (mlr): A streaming processor for CSV, TSV, and JSON with its own verb language. Handles files larger than RAM; no Parquet, SQL, or plotting.
• qsv: A fast Rust-based CSV toolkit with 90+ subcommands, Polars SQL, and Luau scripting. CSV-focused; no R ecosystem access.
• csvkit: A Python suite for converting, slicing, and querying CSV files. Mature and widely used; slower and CSV-only for output.
• jq: The standard CLI JSON processor. Powerful for reshaping JSON in pipelines; JSON-only.
• yq: Like jq but for YAML, TOML, XML, and JSON. Aimed at config-file editing rather than data analysis.
• Nushell: A shell where every command outputs structured tables. Requires adopting a new shell; limited analytical depth.
• VisiData: An interactive terminal spreadsheet for exploring tabular data. Designed for interactive use, not scripted pipelines.
• littler: A lightweight front-end for running R one-liners from the shell. No format I/O helpers, no dependency resolution, no SQL or plotting, but a useful building block for R scripting.