SQL-2626: Add spec tests for subquery comparison expressions #38
Conversation
| @@ -4,7 +4,7 @@ use crate::{ | |||
| catalog::Catalog, | |||
| map, parser, | |||
| schema::{Atomic, Document, Schema}, | |||
| set, SchemaCheckingMode, | |||
There was a problem hiding this comment.
This was an unused import 😮
| keys, | ||
| required, |
There was a problem hiding this comment.
I mainly did this because I wanted the subquery tests to look reasonable as opposed to being something like SELECT * FROM foo WHERE arg1 = ANY(SELECT arg2 FROM foo).
| @@ -59,10 +59,10 @@ Here is an example of such a test, followed by a brief explanation: | |||
| ``` | |||
| In this example, the query contains the expression `arg1 + arg2`, an addition operation. | |||
| There is one map in the `valid_types` sequence. It indicates that `arg1` can be any | |||
| numeric type, `NULL`, or missing; it also indicates that `arg2` can be any numeric type, | |||
| `NULL`, or missing. Since these lists are included in the same map, that means any | |||
There was a problem hiding this comment.
Patrick recently realized we can't actually support MISSING in these tests using the test runner as implemented because it is invalid for a field's sole schema to be MISSING. So we removed missing from the test suite.
This change is just some lingering clean up to remove references to missing.
There was a problem hiding this comment.
Since I added the same keys/required for foo to bar, we need to disambiguate in the queries now. That's not a bad tradeoff by any means.
This PR adds type constraint spec tests for subquery comparison expressions. In the design doc, we initially said we would query test the new document and array comparison features for subquery expressions. However, as I audited the existing spec query and spec type constraint tests, I realized (1) we do not have any other type-related query tests for subquery comparison expressions and (2) we did not have any existing spec type constraint tests for subquery comparison expressions. Given that, I decided to make these new type constraint tests and they demonstrate exactly what we wanted to check.