Data comparator

Base-case tool for the G.E.R.S.O.N. publish pipeline. One tool.yaml at the repo root. The portal clones this repo, builds the Dockerfile, and runs the container only when someone asks for a comparison.

It compares two CSV files on the first column (or --key) and reports rows added, removed, and changed.

Run locally

python compare.py --older examples/older.csv --newer examples/newer.csv --out out/report.json

Run the image

docker build -t data-comparator .
docker run --rm \
  -v "$PWD/examples/older.csv:/inputs/file_a:ro" \
  -v "$PWD/examples/newer.csv:/inputs/file_b:ro" \
  -v "$PWD/out:/outputs" \
  data-comparator

Stdout is the JSON report. The same JSON is written to /outputs/report.json.

tool.yaml

The portal builds the run screen from this file. It does not special-case the language.

Field	What the portal does
inputs[].type	file or text. One control per input.
inputs[].label	The label on that control.
inputs[].bind.path	Where the container receives that input.
outputs[].type	table, file, or text. Picks the result view.
outputs[].bind.path	Where the container writes the result.
outputs[].bind.format	json is an object with columns and rows.

tool.yaml asks for 1 vCPU, 2 GB, no GPU, and a 60 second limit. The sample files are the holdings diff from the portal: NVDA added, TLT removed, AAPL and MSFT changed.