Predicate definitions with Arity above 0 should be able to end in a '?' without generating a warning to maintain consistency with simple boolean return value getters.

[peter-wasilko]

Imba Version: 2.0.0-alpha.202

Details

In trying to define a predicate to return whether a collection contains a reference to an element, I want to end the function name with a question mark to flag its boolean result type (ie. invoke the predicate with : collection.contains? element)

Expected Behavior

This should work without generating a warning unless the function definition doesn't return a boolean value. Otherwise, someone reading imba code will not be able to rely on the question mark as a marker for all predicates.

Restricting the '?' notation to getters/setters seems arbitrary and counter intuitive and would require a preceding function call to store any arguments in an element's properties or blackboard construct before invoking an arity 0 getter proper. (e.g. element.test-collecition = $list followed by element.in-test-collection?)

Actual Behavior

The code runs but the compiler warns:

        warning: [imba] Only getters/setters should end with ?

Possible Fix

If the compiler can not infer that a boolean is being return, change the warning message to:

        warning: [imba] Cannot infer a boolean return value, only functions returning true or false should end with ? If that is the case, you may safely disregard this warning.

Otherwise, permit getters to accept arguments.

[somebee]

I see. I've always thought that predicates should not support arguments, mostly from an aesthetic viewpoint (with parens it is easy to misread as a ternary etc item.render?(something)). But I see the argument against this. Another thing I'm worried about is that if people define predicates as methods, item.contains? will always be truthy since it is just a reference to a function, and this may be very hard to catch.

My proposal would be to still allow them as methods, but warn (or disallow) predicate methods without parameters. What do you think?

[peter-wasilko]

I think that is an ideal solution, since a parameterless predicate should be expressed as a getter rather than a method.

[somebee]

@haikyuu Let's implement this together and ship it as part of the next alpha tomorrow!