`?` at the end of an instance variable name

Help & Support
1

Sometimes I’d like to put a ? at the end of an instance variable name. For a simple example,

struct Foo
  @happy? : Bool
  
  def initialize(@happy?)
  end
end

But it gives a parsing error: Carcin

syntax error in eval:2
Error: unexpected token: ":"

Can the parser be made to allow that?

1 Like
2

I don’t think, this was explicitly disallowed as per Disallow question mark (?) and exclamation mark (!) at the end of the LHS in an assignment. · Issue #6685 · crystal-lang/crystal · GitHub.

You may be interested in the getter? macro tho if your intent is to want to be able to do something like foo.happy? outside the type itself. Otherwise the ivar itself will need to be named @happy and referenced as such.

3

Thanks. Yeah, getter? would definitely work. But this was a little surprise to me as ? is allowed in all other places as var names.

4

as ? is allowed in all other places as var names.

It is not allowed as a variable, unfortunately. I think that would be awesome, but the most of the core members disagree, but I digress.

So in that regard, instance variables not being allowed ? is consistent.

5

True. But method arguments can have ? in their names. I find it a bit odd because method arguments are essentially variables.

1 Like
6

Good catch, this should be a bug in the parser, for the ? end of the method, Crystal follow the rule come from Ruby, in the Ruby, following code raised.

 ╰──➤ $ ruby 1.cr

1.cr: --> 1.cr
unexpected name for a parameter
> 1  def foo(a? = 100)
> 3  end

1.cr:1: syntax error, unexpected method, expecting ')' (SyntaxError)
def foo(a? = 100)
        ^~
7

Aha. If we want to be consistent with Ruby, I’m fine with that. Hopefully the parser will be changed to enforce that. Summing up,

  • Variables and method arguments cannot end with ?
  • Only method names can end with ?

Does this sound correct?

1 Like
8

Does this sound correct?

IIRC, Only method names can end with ?, no more rule, Crystal added the macro, it usage almost same as method.

9

Whoa, TIL

def foo(a? = 100)
  pp a? * 2
end

foo 21 # => 42

I’m legitimately surprised at this. I had no idea this was valid.

1 Like
10

Also related: Def, macro, and block parameters could end with `?` or `!` · Issue #10917 · crystal-lang/crystal · GitHub

1 Like
11

Wow！ I just found out that ! and ? it can also be used as a local variable.

true? : Bool = true
fail! : Bool = false
p true?
p fail!

If I knew it when I was started use Crystal, I will use variable name like this.

Any disadvantage there?

If no, why not? Crystal is not Ruby anyway.

12

It’s likely this is a missed context from when true? = true was disallowed. I don’t think it’s expected.

1 Like
13

The tree sitter will highlight them as methods instead of variables currently.

14

The fact that these things work is a bug that will eventually be fixed, so please don’t use this :slightly_smiling_face:

Def, macro, and block parameters could end with `?` or `!` · Issue #10917 · crystal-lang/crystal · GitHub is accepted as being a bug

15

I’ve yet to read an actual argument against ? in variables though. ! makes sense as it could potentially conflict with stuff like != too, but ? has no such downside.

The argument that’s usually made is “? denotes a method call.”. But why? there’s nothing about ? that’s inherently related to methods. If it’s because of how ruby does it, I don’t think that Crystal has to follow Ruby in everything.

? for boolean values is objectively useful for readability.

16

As the person who started this thread, I’m giving my humble opinion. I’m not coming from Ruby. But building a Ruby-like language that is also fast was one of the motivations of Crystal IIUC. So it’d be best to not deviate too much from Ruby’s syntax unless there’s obvious benefits, in my opinion.

1 Like