Consider updating Dedup definition to apply de-duplication to error values

[kasei]

As discussed in #125, Dedup/Distinct currently does not handle the de-deduplication of error values which is important for handling COUNT(DISTINCT *) . This case could be special-cased in the text, but custom set functions. I'd like to see Dedup updated to handle error values natively as it allows custom aggregates to work over error values while having (what I think of as the expected) DISTINCT handling handled before the aggregation.

[hartig]

As I just wrote in #125 (comment), the question of error values is irrelevant for COUNT(DISTINCT *) (and also for COUNT(*)) because these are special cases for which the ListEval function is not even used. Instead, the expected result for these two special cases is defined purely based on the given sequence Ψ of solution mappings, rather than on the sequence M(Ψ) of lists as produced by ListEval.

[kasei]

You're right, ListEval isn't used in the case of COUNT(DISTINCT *). Including it in the discussion here is a mistake. I think the example you gave in #125 is a good summary of the issue using some arbitrary SomeSetFct set function.

[pfps]

Is this issue then resolved by #125 being merged?

[hartig]

Yes, I would say the issue is resolved. @kasei ?