Remove unnecessary and brittle tests

[martint]

Column names coming out of the query are not necessarily related to the column names in the table function. These tests are testing behavior that is not necessarily expected or guaranteed, so they are brittle and can break at any time.

Release notes

(x) This is not user-visible or docs only and no release notes are required.

[kasiafi]

It seems that the method hasColumnNames() is only used in tests for table functions. Are you going to remove it?

[hashhar]

Column names coming out of the query are not necessarily related to the column names in the table function

It might not be true for all table functions but if it doesn't remain true for query passthrough functions then the usability of those will become questionable since the expectation is that if you select some columns within the PTF the output includes columns of the same name so that you can access them in the outer query.

Or do I misunderstand your statement?

[kasiafi]

Column names coming out of the query are not necessarily related to the column names in the table function

@hashhar it is not about table functions but about usefulness of testing table function's output names by inspecting the query output. There are two reasons why we shouldn't do it I can think of:

1. we don't follow ansi identifier semantics. The column names might change between the output of the TF, and the query output
2. at the query output all columns have names. Within the query they might not. A TF can produce an anonymous column, but the test will see "_col0".

[hashhar]

at the query output all columns have names. Within the query they might not. A TF can produce an anonymous column, but the test will see "_col0".

Good point. I hadn't considered this case.

Also the first point you mention already happens for example for H2 where column names differ in case.

Thanks for explaining the issues.

[hashhar]

@martint do you think it'd be good to include the two possible cases as examples in commit message. I'm sure others would wonder as well when looking at commit in future.