Feature Request: "Don't care" parameters

[togakangaroo]

I'd like to submit this feature for discussion and if I get positive feedback might take a crack at implementing it.

Occasionally I get into situations where I don't care about a particular parameter but I have to include it because I use a positional parameter to its right. The most obvious example is receiving data into a jquery event callback

$element.on 'click', (e, thingIActuallyCareAbout) -> ...

I don't care about e at all but have to include it.

In pattern-matching languages there's the catchall _ pattern, so I would just use that, but I can't repeat it in a parameter list and with the prevalence of underscore/lodash it would actually be more confusing.

I propose a new symbol that you can use in parameter lists to indicate you don't care something like

doStuffAndCall (*, *, data) -> ....

It could generate something like

doStuffAndCall(function(__1, __2, data) { ... })

What does everyone think?

[davidchambers]

Perhaps a coding convention is sufficient:

doStuffAndCall (x, xx, data) ->
  ...

Another option:

doStuffAndCall ->
  data = arguments[2]
  ...

[vendethiel]

Duplicate parameter names (as _) were shot down in a previous ticket.

#1002

[michaelficarra]

The problem still exists. You're forced to name parameters you won't reference in the function body.

[lydell]

There’s always the old “destructure into nothing” trick:

doStuffAndCall ([], [], data) ->
  ...
# alternatively:
doStuffAndCall ({}, {}, data) ->
  ...

@michaelficarra said in #1558 (comment):

A parameter list its just an implicit positional destructuring assignment of arguments. It should support all of the LHSs assignment does.

Which seems to be true, and sane. Using ? as a placeholder more or less anywhere has been talked about before (including the following comment), such as for ?, index in array. The “destructure into nothing” trick works there, too: for [], index in array.

So, IMO, this issue could be generalized into adding a single-item placeholder syntax for destructuring in general, which has been talked about before. The latest discussion being #3268 (comment), I believe. (Using the “destructure into nothing” trick feels like a hack to me, isn’t very intuitive, is a bit ugly and hasn’t got an ideal compilation.)

[michaelficarra]

Ah, I had forgotten about that. Thanks, @lydell. Looks like we already decided that [] or {} would be the accepted standard in #3268.

[xixixao]

Hmm, I see I didn't get around to it. But clearly ? is not a good placeholder and I think we're out of characters (_ obviously being unusable given its popularity as valid identifier).

[satyr]

we're out of characters

We don't need a character for this. Just allow: (, , data) ->

[davidchambers]

Great idea, @satyr. It's clear to me what this means, as it resembles JavaScript's [,,,42].

[lydell]

I thought about that, too, some time ago. I really like it. The reason I didn’t propose it was that I couldn’t make up my mind about when newlines are used instead of commas.

[uninteresting, interesting] = foo
[
  uninteresting
  interesting
] = foo

[ , interesting] = foo
[
  # ???
  interesting
] = foo

Should a comma be required in that last case?

[
  ,
  uninteresting
] = foo

[vendethiel]

Another funny one is : should [1,,] be the same as [1,] ? (array trailing commas)

[michaelficarra]

I was just going to bring that up, @Nami-Doc. Array and object literals (in CS and ES5) have both holes and elisions. If parameter lists are allowed to have holes, do they also have elisions? Basically, what is the arity of (a, b,) ->?

[lydell]

Another problem is that [a,]=b is currently allowed. (It is equivalent to [a]=b.)

[satyr]

when newlines are used instead of commas

Should a comma be required in that last case?

It'd have to.

In case that feels too awkward, we could also allow null as placeholder:

[
  null
  interesting
] = foo

what is the arity of (a, b,) ->?

Should be same as [a, b,].length.