Multi-file validate data: list failing files in the summary (per-file pass/fail)

[cmungall]

The multi-file validate data F1 F2 … mode (thanks for #49) is great for batching — we run it over ~2,670 files in CI. One reporting gap shows up at that scale:

Each file gets a Validating <file> header, and failing files get a Validation Issues block under it, but the final summary only says:

Validation Summary:
  Files validated: 2670
  Issues found in 5 file(s) (7 validation issue(s))

…without naming which files. Triaging then means scanning all 2,670 Validating / Validation Issues blocks.

Request

Enumerate the failing file paths in the summary (ideally a per-file pass/fail line). The sibling linkml-term-validator already does exactly this in multi-file mode:

✅ fileA.yaml
❌ fileB.yaml - 1 issue(s):
    ❌ ERROR: …
Summary: 1/2 files failed, 1 total issue(s)

Matching that — at minimum, listing the failing files in the summary — would make large-batch runs much easier to triage.

Optional

An --errors-only / --quiet flag to suppress the Validating <file> line for passing files (2,670 of them is a lot of noise) would complement this nicely.

Running 0.2.1rc1.

[github-actions]

Summary

Feature request: improve the validate data multi-file summary to enumerate which files failed, not just how many.

Background

This is a follow-on to #49 (merged 2026-06-17), which added multi-file validate data F1 F2 … mode. The multi-file loop builds the Validator once and reuses it across all files — a ~6× speedup for bulk CI runs (10 files: ~2.1 s vs ~12 s as separate processes). The reporter gap surfaces at scale: when running across ~2,670 files, the current summary is:

Validation Summary:
  Files validated: 2670
  Issues found in 5 file(s) (7 validation issue(s))

…with no indication of which 5 files failed. The only way to find them is to scroll through all 2,670 Validating <file> / Validation Issues blocks.

Relevant code

The summary logic lives in src/linkml_reference_validator/cli/validate.py, lines 340–353. The root cause is that files_with_issues is just an int counter (line 303) — the file paths are never collected.

# Current (lines 303, 336, 348–350)
files_with_issues = 0
...
files_with_issues += 1   # only a count, path is lost
...
typer.echo(f"  Issues found in {files_with_issues} file(s) ({total_results} validation issue(s))")

Minimal fix

Change files_with_issues from int to list[Path], then enumerate the failing paths in the summary:

# Proposed
failing_files: list[Path] = []

# in the loop:
if file_results:
    ...
    failing_files.append(data_file)

# in the summary:
if failing_files:
    typer.echo(f"  Issues found in {len(failing_files)} file(s) ({total_results} validation issue(s)):")
    for f in failing_files:
        typer.echo(f"    ❌ {f}")
    raise typer.Exit(1)
else:
    typer.echo("  ✅ All validations passed!")
    raise typer.Exit(0)

This matches the pattern already used by linkml-term-validator:

✅ fileA.yaml
❌ fileB.yaml - 1 issue(s):
    ❌ ERROR: …
Summary: 1/2 files failed, 1 total issue(s)

Optional enhancement: --errors-only / --quiet

A --errors-only flag would suppress the Validating <file> header for passing files (2,670 of them is noise in CI). The suppress would only affect the per-file progress lines — the summary would always be printed. This is a purely additive CLI change; backward-compatible.

errors_only: Annotated[bool, typer.Option("--errors-only", "-q", help="Only print output for files with issues")] = False

# in the loop:
if not errors_only:
    typer.echo(f"\nValidating {data_file} against schema {schema_file}")

Implementation notes

• The fix to failing_files is a one-liner change; the summary block is ~4 lines.
• No new dependencies needed.
• Existing tests in tests/ for multi-file mode (added in feat(cli): accept multiple data files in 'validate data' #49) would need updating to assert the new summary format includes file paths.
• --errors-only can be a separate, subsequent PR if preferred.

@dragon-ai-agent please implement the failing-file enumeration in the summary (the minimal fix above) with TDD — write the test first, then update src/linkml_reference_validator/cli/validate.py. Also add --errors-only / --quiet flag as described. Make sure just test passes.