feat(validation): validate several files per run and look up the schema via OSSIE_SCHEMA

[kayemkim]

Summary

validate.py took exactly one file. It now accepts any number of paths, reports each one, prints a summary when there is more than one and exits non-zero if any file failed, so validate.py models/*.yaml works in a CI step without a shell loop. Output for a single file is unchanged, which keeps the converter tests that read it (NVIDIA, Cube) as they are.

The default schema path assumed this repository's layout. --schema still wins; otherwise OSSIE_SCHEMA is consulted before the repository default, and the not-found message now says how to point at a schema. I left out the ./schema/ fallback the issue floated, since the environment variable covers the vendored case without inventing a directory convention.

The check sequence itself moved out of main() into validate_document(data, schema), so a caller with a parsed document can run every check, including the ontology checks from #489, without going through the CLI.

This sits on top of #534 and #542. The schema is still read and checked once, before the loop, now in load_schema_or_report. The model side now goes through read_text_or_report, which prints the same messages #534 added but returns None instead of exiting, so an unreadable file among several fails only itself and the run continues. There is a test for that case alongside the other eleven in validation/tests. #271 changes the reporting lines that moved into validate_file unchanged, so whichever lands second is a small rebase.

Verification

• validation/tests/: 176 passed, validation/test_validate.py: 53 OK, and the full validation workflow (3.11 to 3.14) run locally on this branch merged into current main.
• Single-file output for examples/tpcds_semantic_model.yaml is byte-identical to main; the NVIDIA and Cube converter suites, which call or import validate.py, pass unchanged.
• A run over a valid file, a binary file, a missing path and a directory reports each one and ends with 1 of 4 file(s) passed, exit 1.

Related Issues

Follow-up to #266, items 2 and 3 (item 1 was #390).

Checklist

Validation

• New validation cases are covered by tests

Tests

• All existing tests pass (pytest / CI green)
• New functionality is covered by tests

Compliance

• ASF license headers are present on all new source files (no new files)
• No third-party dependencies are added without PMC/IPMC approval

🤖 Generated with Claude Code

[jbonofre]

LGTM! Thanks for the multi-file support @kayemkim!

A few minor comments:

• Only YAML parse errors and RecursionError are caught per file. Any other exception (e.g. a TypeError or KeyError on unusual input) aborts the whole run with a traceback, so remaining file are not validated and no summary prints. That conflicts a bit with the docstring's "every file is validated and reported". I would like to see a per-file catch-all that reports the failure and continues, maybe in this PR.
• The PASSED line prints only yaml_path.name, so a/model.yaml and b/model.yaml look identical, and warning/error lines carry no path. Printing the full name would make CI logs easier to attribute.
• Any starting starting with - is now reject as an unknown option, and there's no -- separator. Files like -model.yaml used to validate. Unlikely to matter in practice, but a -- separator would cover it.

[jbonofre]

@kayemkim do you want to tackle first comment in this PR or as follow-up?

[kayemkim]

Thanks for the review. I took the first and third in this PR, pushed as a second commit so the diff on top of what you approved stays small.

A check that raises now prints Error: <ExceptionType> while validating <path>: <message> on stdout, keeps the traceback on stderr and counts the file as failed, so the remaining files still run and the summary prints. A check raising on a document is a validator bug rather than a model problem, so I did not want to lose the traceback entirely; stderr keeps it out of the per-file report. There is a test that makes validate_document raise on one of two files.

-- now takes the rest of the arguments as file names, with a test for -model.json, and the usage line mentions it.

On the PASSED line: in a multi-file run every file's block starts with == <path> as given on the command line, so a/model.yaml and b/model.yaml are told apart by the header above their PASSED or error lines. I left the single-file output alone because the NVIDIA tests and validation/test_validate.py read it, and a vendored copy may have a grep on the same text. If you would rather have the full path on the PASSED line in the multi-file case anyway, it is a one-line change and I am happy to add it.

[jbonofre]

LGTM!

Two small behavior changes in valiidation/validate.py which are completely fine, but worth being aware of:

• any argument starting with - is now treated as an unknown option, so a file named -foo.yml now prints usage and exits 1. The new docstring points to -- as the workaround, which is reasonable.
• a missing schema is now reported before any input file is checked, so the order of error messages differs from before. Exit codes are unchanged.