Crystal: Multiple constant assignment regression
FOO, BAR = 1, 2
FOO # => undefined constant FOO
This was fixed after being reported reported in #958 but it's failing again with a different error message.
Exilor
All 4 comments
Good catch. I will remove this "feature" from the language
asterite
on 25 Oct 2016
@asterite i agree with you. Removing this will be better 馃憤
sdogruyol
on 25 Oct 2016
Why remove it?
Exilor
on 25 Oct 2016
Because this:
FOO, BAR = 1, 2
Is equivalent to this:
tmp1 = 1
tmp2 = 2
FOO = tmp1
BAR = tmp2
Multi-assign is expanded this way to make this possible:
a, b = b, a
and you need to store the right-hand side in a temporary variable first.
But that doesn't work with constants, because constants don't have access to local variables, they have their own scope.
I also think that this actually makes the code a bit harder to read. And constants are used much less frequently than local variables. We _could_ change the rule and simply assign values one by one in a multi-assign, just for constants, but I'm not sure it's worth it. So the easiest thing to do is to remove this from the language.
asterite
on 25 Oct 2016
Related issues
straight-shoota
路
91Comments
xtagon
路
132Comments
asterite
路
71Comments
RX14
路
62Comments
asterite
路
70Comments
Most helpful comment
Because this:
Is equivalent to this:
Multi-assign is expanded this way to make this possible:
and you need to store the right-hand side in a temporary variable first.
But that doesn't work with constants, because constants don't have access to local variables, they have their own scope.
I also think that this actually makes the code a bit harder to read. And constants are used much less frequently than local variables. We _could_ change the rule and simply assign values one by one in a multi-assign, just for constants, but I'm not sure it's worth it. So the easiest thing to do is to remove this from the language.