driftpipev0.9.4docs for Sign in

Container image

NAME

container - running the immutable image directly

The image is built from the release tag with a distroless base, and its entrypoint is the binary, so arguments pass straight through with no shell in between.

DIRECT RUN

$ docker run --rm -v "$PWD:/w" -w /w \
    reg.driftpipe.sh/driftpipe:0.9.4 \
    check --schema schema.lock --input events.ndjson --sample 0
driftpipe 0.9.4   48213 records   31 paths   182ms
no drift

Mounting the working directory read-write is only needed for snapshot; a check can mount :ro and still read both the stream and the lock. The exit status of the container is the exit status of the tool, so a wrapping script reads it the same way it would a local binary.

IN A PIPELINE STEP

# any runner that can run a container
driftpipe() { docker run --rm -i -v "$PWD:/w:ro" -w /w reg.driftpipe.sh/driftpipe:0.9.4 "$@"; }
zcat dumps/latest.ndjson.gz | driftpipe check --input - --fail-on breaking

The -i flag keeps standard input attached, which is what lets the image read a piped stream through --input - exactly as the bare binary would.

GOTCHAS

  • Tags are immutable by release, so 0.9.4 is always that build; the moving latest tag is a choice you will regret on a Friday.
  • Override the entrypoint only when a job needs a shell around the tool; otherwise the pass-through is simpler and harder to get wrong.
  • The image adds nothing the binary lacks. If the runner can install a 6 MB file, the prebuilt binary is one fewer moving part.