Regression in PostgreSQL Contains on array

[roji]

Since syncing EFCore.PG from commit 725adf7 to ac46335, Contains on arrays is broken. AllAnyContainsRewritingExpressionVisitor seems rewrites the Contains to Any, which then produces a translation failure. This rewriting seems problematic to me (shouldn't we leave alone calls on arrays or other stuff?), but let me know if this is intentional and I should change the provider to recognize Any on arrays instead of Contains.

[smitpatel]

Suspected PR which adding offending code #19744

Related issue to improve #20155 (not sure if that would fix this too. Blocked on #20164

Can you also post expression tree which fails?

[roji]

Sure, here you go:

[Microsoft.EntityFrameworkCore.Query.QueryRootExpression].Single(e => e.SomeArray.Contains(3))

After NormalizeQueryableMethodCall:

[Microsoft.EntityFrameworkCore.Query.QueryRootExpression].Single(e => e.SomeArray.AsQueryable().Contains(3))

After AllAnyContainsRewritingExpressionVisitor:

[Microsoft.EntityFrameworkCore.Query.QueryRootExpression].Single(e => e.SomeArray.AsQueryable().Any(p => (p == 3)))

[ajcvickers]

I'm assuming, because we have tests for it, that Contains on byte arrays is still working. Do we special case byte arrays somewhere?

[roji]

I thought about this some more... The Npgsql provider should actually be able to translate a simple Any on arrays - as long as the tree above doesn't produce any exceptions in preprocessing and makes it to translation, I should just handle this structure. So this is probably not an issue (though let me confirm).

[roji]

FYI just unskipped the tests and they're green, thanks!