Nushell

Nushell is a functional, structured-data language that happens to be a shell. Write it that way: pipelines of transformations over records and tables, not bash with different syntax.

Why Nushell

Functional over structured data: JSON, ls, and command output stay records through the whole pipeline instead of being re-parsed as text. So no jq/sed/awk/date to install, wrap, or write twice for macOS and Linux — and Nix pins the interpreter.

Which language to pick for a given script is in the repo's AGENTS.md.

Steps

1. Read the docs for the topic at hand (see below) before writing any Nushell.
2. Write the code functionally — see Rules.
3. Verify every construct against the installed version before claiming it works. Nushell breaks and renames things between minor releases, so do not trust memory of a command's signature:

   nu --version
   nu -c '<snippet>'          # run a snippet
   nu --ide-check 10 <file>   # parse/typecheck a script
   nu -c 'help <command>'     # check a signature

4. Format before committing:

   nufmt <file>             # rewrites in place
   nufmt --dry-run <file>   # non-zero exit if it would change

Docs

Read before writing:

• https://www.nushell.sh/book/thinking_in_nu.html
• https://www.nushell.sh/book/nushell_map_functional.html
• https://www.nushell.sh/book/nushell_map_imperative.html
• https://www.nushell.sh/book/coming_to_nu.html
• https://www.nushell.sh/book/style_guide.html

Every page of the book is in the sidebar of those, as are /commands/, /cookbook/, and /lang-guide/. Follow the sidebar to whatever the task needs instead of guessing.

Rules

Functional style is the default. In review, treat each of these as a defect to fix:

• mut + for as an accumulator. Replace with reduce, each, where, group-by, zip, flatten, insert/update, generate. mut is legitimate only for genuinely sequential state that no filter expresses.
• String plumbing between steps. Pass records and tables. Serialise once at the boundary (to json --raw), parse once on the way in (from json).
• Nested if/else chains on a value's shape. Use match, including list patterns ([], [$first, ..$rest]) and guards.
• Reimplementing a builtin. Check https://www.nushell.sh/commands/categories/filters.html first.
• Shelling out for data. Prefer native commands over ^jq, ^sed, ^awk, ^date. When an external command is genuinely needed, pipe its output through from json etc. immediately.
• Bare each for effects. If the result is unused, say so with | ignore; if it produces values, keep them in the pipeline.

Also:

• Give every script a def main.
• Type custom command signatures — parameter types, --flag, and the return type. They document intent and are checked at parse time. nufmt strips a redundant : any; leave it stripped.
• Timestamps are datetime, durations are duration. Compare and subtract them directly instead of formatting to strings. Convert explicitly to UTC (| date to-timezone UTC) when the value crosses a boundary that expects it.
• Fail loudly with error make rather than returning a sentinel value.

In this repo

• CI workflow steps run Nushell via shell: nu {0} with hustcer/setup-nu. See .github/workflows/_update-flake-reusable.yaml for the established style.
• Pin the interpreter. Workflows pass an explicit version: to setup-nu; keep doing that, or a Nushell release will break a script that never changed.
• Standalone scripts live alongside their consumer, e.g. nix/packages/git-wtpr/git-wtpr.nu.
• Neither nu nor nufmt is on the interactive PATH; use the missing-tools skill to reach them.