Add $covariancePop and $covarianceSamp window operator compatibility …

[Cassie-2025]

Summary

• Add comprehensive compatibility tests for $covariancePop and $covarianceSamp window operators
• Update TEST_COVERAGE.md §22 with overflow requirements for multi-expression operators

Test Coverage

Each operator has 7 test files covering all dimensions from TEST_COVERAGE.md §22 (Window Operator Coverage):

File	Dimension
test_window_*_argument_validation.py	Valid/invalid expression shapes, structural parse errors
test_window_*_field_paths.py	Dotted paths, missing fields, null values, array-of-objects traversal, numeric path components
test_window_*_frame_computation.py	Whole-partition, cumulative, reverse-cumulative, sliding frames
test_window_*_non_numeric_handling.py	Null, missing, string, boolean, array, object, ObjectId, Regex, Binary, Timestamp, MinKey,
MaxKey
test_window_*_numeric_precision.py	Int32/Int64/Double/Decimal128 mixing, overflow (4-case intermediate overflow per §22),
catastrophic cancellation
test_window_*_order_independence.py	Same result regardless of sort direction
test_window_*_special_floats.py	NaN/Infinity propagation in removable vs non-removable windows

Key behavioral differences tested:

• $covariancePop: divides by N (single element → 0)
• $covarianceSamp: divides by N-1 (single element → null)

Test plan

• All 241 tests pass against the reference server (8.2.4)
• Pre-commit hooks pass (black, isort, flake8, mypy, forbidden terms, DC

[documentdb-triage-tool]

🤖 Auto-triaged by documentdb-triage-tool.

Applied: compatibility test, documentation, enhancement
Project fields suggested: Component test-coverage · Priority P2 · Effort XL · Status Needs Review
Confidence: 0.88 (mixed)

Reasoning

component from path globs (test-coverage, docs); effort from diff stats (7759+2 LOC, 15 files); LLM: Adds 14 new compatibility test files (241 tests) covering $covariancePop and $covarianceSamp window operators across all §22 dimensions, plus a TEST_COVERAGE.md update — multi-file, one component, important feature coverage.

If a label is wrong, remove it manually and ping @patty-chow so the rules can be tuned. The bot will not re-label items that already have component labels.

[eerxuan]

Overall good.

Fix before merge:

• 🟠 negative_correlation_order_independent (Pop & Samp) doesn't test order-independence — it only runs
  the ascending sort and asserts -4.0, never the descending compare. Add the descending run, or rename and
  drop the claim. (Dimension is still covered elsewhere in the file, so not blocking.)

Nits:

• 🟢 argument_validation and order_independence files hand-roll pipelines where run_window_operator()
  would fit.
• 🟢 COVAR_DOCS redefined per file — could move to window/utils/.

Strengths: rigorous 4-case overflow coverage with signed ±Infinity assertions; correct single-element
split (Pop→0.0, Samp→None); expected values faithful to the reference server; the §22 doc update is
genuinely good.