Discussions on the type system

Discussions on the type system

Sometimes it is not immediately clear whether we should allow or disallow a kind of expression. Such cases are discussed here. Feedback is very welcome.

Comparison operators <, <=, >, >=

Should these cases be allowed or disallowed?

• 1 < 2 => allowed (standard use)
• [1, 2, 3] < 4 => disallowed (use all or any)
• [1, 2, 3] < [4, 5] => disallowed (makes no sense)
• [1, 2, 3] < [4, 5, 6] => disallowed (in principle we could compare them pairwise, but as far as we know, there is no use case for that. Furthermore, allowing this would prove difficult for expressions that return a list with undetermined length, e.g., comparing a list with constraint (1..3) with another with constraint (3..5))
• 1 all < 2 => disallowed (the cardinality operator all is redundant)
• [1, 2, 3] all < 2 => allowed (standard use)
• 2 all > [1, 2, 3] => disallowed (see section Operators with all and any)

Comparing date, zonedDateTime and time

Should values of all of these types be comparable with = and <? => yes (same rules as discussed for numbers in previous section apply)

Should they be allowed to be mixed (e.g., comparing a date with a zonedDateTime, or a date with a time)? => no (there is no good way to compare them; users should explicitly convert between them, e.g., by adding a time zone to a date object)

Operators with all and any

When using the prefixes all and any with operators such as =, <> and >=, one of the arguments should be a single value. In principle, this could be either the left or the right operand, but I think it is better to have a consistent position for the single item, i.e., by always forcing the right operand to be singular. E.g.,

• this is allowed: [1, 2, 3] all < 2
• this is disallowed: 2 all > [1, 2, 3]

Furthermore, we disallow using these prefixes unnecessarily, such as in 5 any = 3 and 2 all <= 10.