Coverage

CodeStax ingests test-coverage reports and turns them into:

1. A Coverage A–E rating on every scan
2. Coverage-gap findings for low-covered files
3. The new_coverage_min quality gate input
4. Per-file coverage view on the scan dashboard

See Coverage Ingestion for the upload flow + CI recipes. This page is the scanner reference.

Supported Formats

Format detection is automatic by filename + root-element sniff. You can also force via report_format query/form param.

Two Ingestion Paths

Auto-detect

Scanner walks the cloned repo (bounded 6 directories deep) looking for files matching known patterns. Skipped: node_modules, .git, __pycache__, .venv, venv, env, dist, .next, .nuxt, .cache.

Works when coverage artifacts are committed to git.

Upload (preferred for most teams)

curl -X POST "https://codestax.co/api/quality/coverage/upload" \
  -H "Authorization: Bearer $JWT" \
  -F "scan_id=12345" \
  -F "file=@coverage/lcov.info"

Max file size: 25 MB.

Output

Per-file records (CoverageReport table)

{
  "file_path": "src/auth/handler.py",
  "lines_total": 420,
  "lines_covered": 412,
  "lines_missed": 8,
  "branches_total": 84,
  "branches_covered": 72,
  "coverage_pct": 98.1,
  "report_format": "lcov"
}

Aggregate (returned from GET /quality/coverage/:scan_id)

{
  "scan_id": 12345,
  "file_count": 847,
  "lines_total": 42100,
  "lines_covered": 38700,
  "coverage_pct": 91.92,
  "files": [ ... ]
}

Coverage-gap findings

Files below threshold emit type: "coverage_gap" findings:

• < 60% (configurable via quality_coverage_low_threshold) → medium
• < 30% (configurable via quality_coverage_critical_threshold) → high

Files < 20 lines total are skipped (noise reduction).

Parser Notes

LCOV

Line records (DA:, LF:, LH:) → total/covered/missed per file. Branch records (BRF:, BRH:) → branch coverage.

Cobertura

<class filename="..."><lines><line number="X" hits="Y"/> — counts lines + branches (condition-coverage="X% (A/B)").

JaCoCo

<report><package><sourcefile><counter type="LINE|BRANCH" missed="X" covered="Y"/> — standard JaCoCo XML.

Clover

<project><package><file><metrics statements="X" coveredstatements="Y"/> — statement-level coverage treated as line coverage.

Thresholds (Configurable)

Per-org scanner config:

{
  "quality_coverage_autodetect": true,            // enable auto-scan of repo
  "quality_coverage_low_threshold": 60,            // medium severity threshold
  "quality_coverage_critical_threshold": 30        // high severity threshold
}

Dependencies

Parser is stdlib-only — no PyYAML or lxml dependency. XML parsing uses xml.etree.ElementTree. LCOV is line-oriented text.

Limits

• 25 MB file cap (hard, backend)
• 2000 file records per scan (soft — excess rows dropped with a warning log)
• Absolute paths in reports are kept as-is (we don’t path-normalize; use relative paths in your CI config for cleanest results)

Known Gaps

• Function-level coverage — parsers capture line + branch; function-level FN:/FNDA: records in LCOV are currently ignored
• Diff-based coverage — “coverage on lines changed in this PR” not yet computed; roadmap Q2
• Coverage trends — per-repo coverage over time is stored but no trend UI yet; use the API

Related

• Coverage Ingestion — CI upload recipes
• Quality Ratings — how coverage feeds the A–E rating
• Set Up Quality Gates — new_coverage_min threshold enforcement