[CaffeinatedCoder]

Closes #388

This is the 10.x port of #382 for 10.2.0 (#358). It is git cherry-pick -x of 5cf3845 from PR #392. It was stacked on PR #393 (#387), which has merged.

Problem

Over a NULL element, the translated CreateFinite gave empty and CreateUnboundedStart/CreateUnboundedEnd gave (,), instead of NULL. So DateRange.CreateUnboundedEnd(b.Visit!.From).Contains(v) selected a booking without a visit: a positive filter answering wrongly.

Change

• The three factories are wrapped by NullOverNullElement: CASE WHEN e IS NULL THEN NULL ELSE <translation> END. This is the same code as on main. EF Core folds it away for required elements, non-null parameters and constants.

• Conflicts, resolved:

  • candidates.json: kept 10.x's version. The six The range factories over a NULL element build a range: CreateFinite gives empty and CreateUnboundedStart/End give (,), where EF Core answers NULL #382 candidates exist only on main.
  • Changelogs: a ### Changed answers row under [10.2.0] in the three changelogs, linking this issue and docs/migration.md#changed-answers-in-1020.
  • Claim E04: main lists OuterJoinProbeTests, which doesn't exist here, so on 10.x E04 lists NullElementTests instead.

• Added on this line only: NullElementTests.RangeFactories_OverANullElement_AreNull. It covers Contains, its negation, == null, IsEmpty() beside an inverted-bounds row (still empty, as in memory), and a projected range.

• Added after the release review: RangeFactories_OverANullElement_AreNull_WhereAPredicateConsumesThem holds a factory over a NULL element where a predicate consumes it: Contains in a bool projection, Overlaps in a filter, its negation and a projection, and All over a captured list. Contains_OverANullElement_ProjectsFalse also projects RangeSet.Contains (10.x: Contains(element) on a range or RangeSet is NULL in SQL for a NULL element (port of #381) #387). Each new assertion was checked on its own: with the translator from before this fix, all six factory assertions fail; with the one from before Contains(element) on a range or RangeSet is NULL in SQL for a NULL element: the negation drops the row EF Core keeps, and a projection into bool throws #381's fix, the RangeSet projection throws.

Tests

• With only this commit's fix stripped, exactly two tests fail: NullOperandTranslationTests.ARangeFactoryOverANullableElement_IsNull (SQL pins) and the new live test. With the fix, both pass.
• Every suite outside the integration projects passes on the floor.
• Both integration projects pass on PostgreSQL 16, on EF Core 10.0.0 and on 10.0.13 (289 + 4 tests on each leg). After the added test, NullElementTests (5 tests) passes on both legs again.

Closes doesn't act on release/10.x, so close #388 by hand after the merge.

🤖 Generated with Claude Code