add map.keys()

[ben-albrecht]

I find myself trying to use map.keys() quite often, both because of the symmetry with map.values() and my python background leading me to expect a similar interface to dict. However, this results in a not-so-great error message, because keys is an array field, not a method:

use Map;

var m: map(int, int);

m[1] = 2;

writeln(m.keys());

error: unresolved access of 'DefaultAssociativeDom(int(64),false)' by '()'

I am in favor of adding a keys() iterator to mirror the values() iterator. If not, we should at least add a better error message for this specific case.

[ben-albrecht]

Kind of a follow-on to #13624 after having more map experience under my belt.

[e-kayrakli]

I think we should have keys.

Apparently, I said in #13624:

What should these yield? I am going the python way again and making it yield
keys. And adding items iterator to yield k-v tuples, values iterator to
yield values.

Reading this again, I don't know why I didn't add keys in the first place. I think that it was an oversight on my part to not add an iter keys() that is identical to iter these()

[bradcray]

Only vaguely related, but recall that late discussion on issue #13624 also raised the question as to whether these() should behave more like items() and less like keys().

[e-kayrakli]

@bradcray

Only vaguely related, but recall that late discussion on issue #13624 also raised the question as to whether these() should behave more like items() and less like keys().

I opened #14718 to summarize the discussion under that closed PR.