tods-validate 0.4.0

Validator for Transit Operational Data Standard (TODS) feeds, formerly the Operational Data Standard (ODS)

tods-validate

A validator for Transit Operational Data Standard (TODS) feeds, with a CLI and a GitHub Action.

TODS is an open standard for describing scheduled transit operations: crew runs, deadheads, vehicle assignments, and other non-public service that GTFS does not cover. It works as an overlay on an agency's GTFS feed. The standard was originally published by Cal-ITP as the Operational Data Standard (ODS) and is now maintained with MobilityData under its current name. This validator checks feeds against the current spec, TODS v2.1.0.

tods-validate reads a TODS package, checks it against the spec, and reports findings in language a scheduler can act on. Each finding says what is wrong, where, and what good looks like, and cites the spec section it comes from.

Install

Requires Python 3.11 or newer.

pipx install tods-validate

or pip install tods-validate into an environment of your choice. For CI environments without Python, a container image is published on releases:

docker run --rm -v "$PWD/feed:/feed:ro" ghcr.io/chelseakr/tods-validate /feed/tods --gtfs /feed/gtfs

There is also a pre-commit hook; see .pre-commit-hooks.yaml for usage.

Usage

Point it at the directory or .zip file containing your TODS files. If your GTFS feed lives in a separate file, pass it with --gtfs so trip, stop, service, and block references can be checked:

tods-validate exports/tods/ --gtfs exports/gtfs.zip

When the TODS files sit next to the GTFS files in one package, the GTFS files are picked up automatically:

$ tods-validate /tmp/demo-feed
tods-validate: /tmp/demo-feed (TODS v2.1.0)

2 errors:
  ERROR TODS-E203 [run_events.txt, row 4, field 'end_time']
    run_events.txt row 4: end_time is '9:45', which is not a valid time. Use HH:MM:SS, e.g. '09:45:00' or '25:10:00' for 1:10 AM the next service day.
  ERROR TODS-E307 [run_events.txt, row 4, field 'trip_id']
    run_events.txt row 4: trip_id 'WKDY-1002' does not exist in the companion GTFS trips.txt (after applying trips_supplement.txt). Run events that represent work on a trip must reference a scheduled trip.
    Fix: Correct the trip_id, or add the trip via trips_supplement.txt if it is non-revenue service.

Summary: 2 error(s), 0 warning(s), 0 info.
$ echo $?
1

The exit code is 0 when no errors are found, 1 when there are errors, and 2 when the package cannot be read at all. Warnings do not fail the run unless you pass --fail-on warning.

Other output formats:

• --format json prints a stable JSON document for tooling.
• --format markdown prints a report suitable for pasting into an issue (--stamp adds a provenance footer for a citable compliance artifact).
• --format github prints GitHub Actions workflow annotations.
• --format sarif prints SARIF for GitHub code-scanning and security dashboards.
• --format html prints a standalone, shareable report.

On large feeds, --max-findings N caps how many findings are listed (the summary is unaffected) and --quiet prints only the summary. Text and Markdown reports group findings by rule and add a root-cause hint when one rule clusters.

New developers can also call the validator in-process; see docs/api.md. Not a programmer? Start with docs/getting-started.md.

To suppress findings your agency has decided to accept, pass --ignore TODS-W206 (repeatable), or put the policy in a tods-validate.toml next to where you run the validator:

ignore = ["TODS-W206", "TODS-I108"]
fail-on = "warning"

Command-line flags win over the file. A config file in another location can be passed with --config path/to/file.toml. A config may also extends = "../base.toml" to inherit a shared house policy, and profile = "strict" (or lenient) applies a named preset that other settings can still override.

Some checks are off by default because they surface judgement calls rather than spec violations. Turn them on with --enable coverage (which GTFS trips have no run event; which blocks have no vehicle) or --enable advisory (e.g. long runs with no break), or by rule ID. See docs/rules.md.

References into GTFS are resolved after applying the supplement files, so a trip added by trips_supplement.txt is a valid target for run_events.trip_id, and a stop deleted by stops_supplement.txt is not.

Merging supplements into GTFS

The spec says that GTFS plus the supplement files should form a valid GTFS dataset (the "TODS-Supplemented GTFS"). The merge subcommand materializes that dataset so you can test the claim, or hand the operational feed to a tool that only speaks GTFS:

tods-validate merge exports/tods/ --gtfs exports/gtfs.zip -o supplemented.zip

GTFS files without a supplement are copied through unchanged; supplemented files get their rows deleted, updated, and added per the spec's evaluation rules, and the command reports what changed per file. Validate the TODS package first so the merge rests on clean inputs.

A CI job that checks the merged feed with MobilityData's gtfs-validator:

- uses: ChelseaKR/tods-validate@v0.4.0
  with:
    path: feed/tods
    gtfs: feed/gtfs
- run: |
    pipx install tods-validate
    tods-validate merge feed/tods --gtfs feed/gtfs -o supplemented.zip
- run: |
    curl -sSL -o gtfs-validator.jar https://github.com/MobilityData/gtfs-validator/releases/latest/download/gtfs-validator-cli.jar
    java -jar gtfs-validator.jar -i supplemented.zip -o validator-report

Other subcommands

• tods-validate stats feed/ --gtfs gtfs/ prints descriptive metrics (run events, distinct runs, revenue vs non-revenue minutes, employees, vehicles, and GTFS coverage) — facts about a feed, not a quality score.
• tods-validate diff old/ new/ validates two versions of a feed and reports which findings were fixed, newly introduced, or still present; it exits non-zero only on newly introduced errors, which is useful in review.
• tods-validate batch a/ b/ c/ validates several feeds and prints a roll-up table (--format json for tooling).
• tods-validate anonymize feed/ -o feed-anon/ writes a copy with person-identifying fields (employee IDs, license plates, vehicle IDs) pseudonymized before sharing. This is pseudonymization, not guaranteed anonymity; see SECURITY.md.

To fail CI only on findings introduced since a known-good run, capture a baseline (--format json > baseline.json) and pass --baseline baseline.json.

GitHub Action

If your TODS export lives in a repository, this workflow validates it on every pull request and annotates findings inline:

name: Validate TODS feed
on: [pull_request]
jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: ChelseaKR/tods-validate@v0.4.0
        with:
          path: feed/tods
          gtfs: feed/gtfs        # omit if GTFS files sit next to the TODS files

Rules

The full catalog of checks, with IDs, severities, and spec citations, is in docs/rules.md, or from the tool itself with tods-validate rules (--format json for tooling). Rule IDs are stable: a CI pipeline can safely filter or suppress specific IDs. The JSON report format is described by docs/report.schema.json.

Ambiguities in the spec discovered while building the validator are tracked in docs/spec-questions.md.

What this does not check

tods-validate validates the TODS files and their references into the companion GTFS feed. It does not re-validate the GTFS feed itself, and it does not check that the merged ("TODS-Supplemented") GTFS dataset is valid GTFS. For those, run MobilityData's gtfs-validator, optionally on the merged feed.

Development

git clone https://github.com/ChelseaKR/tods-validate
cd tods-validate
python -m venv .venv && . .venv/bin/activate
pip install -e ".[dev]"
pytest

Lint and type-check with ruff check src tests scripts and mypy. The rule catalog is generated: after adding or changing a rule, run python scripts/generate_rules_doc.py and commit the result; CI fails if it drifts.

License

Apache-2.0, matching the TODS specification repository.