Extend summary metrics with per-check-name breakdowns

[ghanse]

Changes

Extended DQMetricsObserver to emit a new check_metrics row alongside the existing aggregates (input_row_count, error_row_count, warning_row_count, valid_row_count). The value is a JSON array of structs — one entry per check — that fits the existing metric_name / metric_value schema without widening it:

  [{"check_name": "id_is_not_null", "error_count": 1, "warning_count": 0}, ...]

Backward compatible: existing metrics unchanged, new row is additive and opt-in.

Key design decisions:

• Backward compatible: Existing aggregate metrics are unchanged; per-check metrics are appended after them.
• Automatically enabled: no knobs, auto-enabled.
• Safe aliases: Check names with special characters are sanitized for use as Spark SQL aliases.
• No new dependencies: Uses existing Spark SQL exists() higher-order function.

Linked issues

Closes #943

Tests

• manually tested
• added unit tests
• added integration tests
• added end-to-end tests
• added performance tests

This pull request was AI-assisted by Isaac.

[ghanse]

Good implementation. The per-check metric generation is clean, backward-compatible, and well-tested. Two minor findings below — neither is blocking but both are worth addressing.

[ghanse]

Addressed duplicate alias collision — _build_per_check_metrics now detects when two check names sanitize to the same alias and raises DQXError. Added unit test for the edge case (e.g. check-1 vs check.1). Ready for cleanup pass.

[ghanse]

Cleanup pass complete.

• Renamed default_metrics to all_metrics in metrics property since it accumulates default, per-check, and custom metrics
• Fixed copy-paste docstring in test_dq_observer_custom_column_names that incorrectly said "default column names"

Tests pass (16/16). Ready for review.

[ghanse]

Post-cleanup final review: APPROVED

Reviewed the full diff across metrics_observer.py, engine.py, test_observer.py, and test_summary_metrics.py.

Correctness: The Spark SQL exists() higher-order function is the right approach for checking array-of-struct columns. Single-quote escaping, alias sanitization, and duplicate collision detection are all correctly implemented. The @cached_property to @property change on metrics is necessary and correct since check names are set dynamically.

Completeness: Fully addresses #943 with per-check-name error/warning counts. The track_extended_metrics flag defaults to False for full backward compatibility. Both apply_checks and apply_checks_by_metadata paths are covered since the latter delegates to the former.

Test coverage: 10 new unit tests and 2 parametrized integration tests (4 total runs) cover: default behavior, per-check metrics, metric ordering, custom column names, special character sanitization, check name replacement, empty alias error, and duplicate alias collision.

Code quality: Clean naming, proper docstrings, no dead code, pure helper function for sanitization. AGENTS.md compliant — type hints on all signatures, DQXError for exceptions, no new Any usage.

No changes required. Ready for human approval.

[mwojtyczka]

Review: Extend summary metrics with per-check-name breakdowns

Clean feature addition. The approach of generating per-check SQL expressions via the observer is sound. A few items below.

[mwojtyczka]

LGTM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.03%. Comparing base (0c7d688) to head (c5b9ab0).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1097      +/-   ##
==========================================
+ Coverage   91.97%   92.03%   +0.06%     
==========================================
  Files         101      101              
  Lines        9408     9418      +10     
==========================================
+ Hits         8653     8668      +15     
+ Misses        755      750       -5     

Flag	Coverage Δ
anomaly	53.97% <27.77%> (?)
anomaly-serverless	53.86% <27.77%> (?)
integration	50.14% <83.33%> (?)
integration-serverless	50.31% <83.33%> (?)
unit	55.62% <83.33%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.