Use proper SQL identifier semantics for CALL arguments

[findepi]

No description provided.

[findepi]

cc @losipiuk @kasiafi

[findepi]

This feels wrong. Only one of them should pass.

[findepi]

It works like this:

• the distinction between delimited and not is first erased in

  trino/core/trino-main/src/main/java/io/trino/metadata/TableProceduresPropertyManager.java

  Line 90 in 2051e78

  .map(entry -> new Property(new Identifier(entry.getKey()), entry.getValue()))

  which may be incidental
• then the case is normalized again at

  trino/core/trino-main/src/main/java/io/trino/metadata/PropertyUtil.java

  Lines 60 to 61 in 9ee8f74

  	// property names are case-insensitive and normalized to lower case
  	String propertyName = property.getName().getValue().toLowerCase(ENGLISH);

So

• if we want ALTER TABLE EXECUTE (...) properties to be case sensitive, both these places need to be changed/diverged
• do we want all properties to be case-sensitive? especially the table properties?

cc @martint @kasiafi @dain @losipiuk

[electrum]

I would say that we only allow case-insensitive arguments for procedures. I don't see a benefit to allowing case-sensitive.

[findepi]

i don't see a direct benefit here as well, and that's why @losipiuk and I decided to go accept case-insensitivity for ALTER TABLE EXECUTE's parameters.
However, there is a concern about coherence and conformance, which i'd leave up to @martint @kasiafi to debate.

[kasiafi]

I think that we should respect the proper identifier semantics.
Although it might not seem very beneficial to allow using case-sensitive arguments for procedures, it's about consistency with the spec, and consistency of behavior across Trino. When adding new features, we generally try to be consistent with the spec wrt identifiers. In the long term, we're going to apply per-spec identifiers everywhere.

[martint]

It's not just about case sensitivity, but also about allowing characters that are invalid unless it's a delimited (quoted) identifier. Once we support quoted identifiers, we should be consistent with the specification -- if an identifier is quoted, it needs to be treated as is (case conversions for non-english characters can be problematic, too, as they depend on locale)

[findepi]

Filed #11326 to prevent scope creep in this PR.

[findepi]

Ready to review, even if we decide not to address #11163 (comment) within this PR.

[kasiafi]

Could you please separate tests for procedure name from tests for arguments (put them in different methods)? It seems that they are distinct issues. Also, could you add tests with unquoted names in different case?

[findepi]

This is just a smoke test coverage. I put the together because it's not a detailed test for ALTER TABLE EXECUTE itself. (such test doesn't exist, and I don't want to add it)

[kasiafi]

I think that we should respect the proper identifier semantics.
Although it might not seem very beneficial to allow using case-sensitive arguments for procedures, it's about consistency with the spec, and consistency of behavior across Trino. When adding new features, we generally try to be consistent with the spec wrt identifiers. In the long term, we're going to apply per-spec identifiers everywhere.

[findepi]

Please re-review.

[findepi]

(rebased)