`node-lookup -jF ipaddress,enc_role` returned `{}` because it queried a single
fact literally named "ipaddress,enc_role". Requesting several facts per host is
a natural need (e.g. pairing an address with its role).
- Split -F on commas (splitFactNames) and match any of them via an "or" over
["=","name",<n>] clauses (nameFilter); a single name keeps the plain "="
form.
- Key JSON output by each result's real fact name so all requested facts appear
under the host (previously keyed by the raw -F string).
- Update the -F flag help and add unit tests (split, single vs multi query
shape, multi-fact JSON).
pflag does not attach a space-separated value to a string flag that is grouped
with a bool flag, so `node-lookup -R -pm k8s` parsed -p and left `k8s` as a
stray positional, failing with "unknown command k8s". Only `-pm=k8s` or the
un-grouped `-p -m k8s` worked, which is surprising.
- Allow one positional argument (cobra.MaximumNArgs(1)) and fall back to it for
the match value when -m is empty (matchValue()), so -pm/-im/-ipm <value> and
a bare `-p <value>` all work. -m still wins when both are given.
- Add matchValue unit tests.
- Document the positional value and the pflag grouping quirk in AGENTS.md.
## Why
The unit tests stopped compiling after the `--pm` → `-p`/`-i` match-modifier refactor was left uncommitted, there was no RPM/completions distribution story, and invoking the tool without a TTY against an empty pipe silently returned nothing. This makes the project releasable and safe to run from agents/CI.
## Changes
- Make stdin handling robust: replace the fragile `!isTerminal` check with `stdinReader()`, which only reads node names when stdin is a real pipe/redirect carrying data. Terminals, `/dev/null`, and empty/closed pipes now fall through to a normal query, so running without a TTY behaves like an interactive run.
- Repair and expand `main_test.go` to match the current `buildQuery`/`run` signatures; add coverage for the match modifiers, all output modes, config precedence, and the new `stdinReader` logic. `httptest` stubs PuppetDB (no live deps).
- Add nfpm packaging (`packaging/nfpm.yaml`, `scripts/build-rpm.sh`): installs the binary to `/usr/bin/node-lookup` and bundles generated bash/zsh/fish completions under the standard system paths.
- Rework the Makefile to build into `dist/` and add `completions`/`rpm` targets.
- Split PR CI into `build`, `test`, and `pre-commit` workflows and extend `release` to build the RPM and `PUT` it to the artifactapi `rpm-internal` repo. Every step sets a `serviceAccount` and k8s resources.
The project directory has also been relocated under `prodenv`.
Reviewed-on: #13
Co-authored-by: Ben Vincent <ben@unkin.net>
Co-committed-by: Ben Vincent <ben@unkin.net>
-a requires -n and prints all PuppetDB facts for the specified node
as 'fact_name value' sorted alphabetically by fact name, useful for
discovering available facts to query against.
Query PuppetDB for node facts via CLI. Replaces the original Python script.
- XDG config (~/.config/node-lookup/config.yaml) with env var overrides
- All flags from original tool preserved (-n, -F, -R, -m, --pm, -1, -2, -C, -A, -j)
- config init / config show subcommands
- Unit tests (23), Makefile, GoReleaser config, pre-commit hooks
💘 Generated with Crush
Assisted-by: Claude Sonnet 4.6 via Crush <crush@charm.land>
unkinben