[CS2] Refine #4666: add parens to chained do IIFE with params

[helixbass]

@connec pointed out that #4666 didn't handle do <IIFE> with arguments

So this extends #4666 to handle arguments. Added tests where the arguments are on separate lines and contain nested PARAM_END token

[jashkenas]

And if the arguments have default values or are destructuring?

(Let's revert the previous, and do this in the grammar — instead of continuing to tweak this rule.)

[helixbass]

@jashkenas added a test including destructured param (it was already working) - it's using the rewriter's detectEnd(), which takes nesting into account, so I don't think there will be weird specific types of params that this won't handle

My only argument for keeping it this way for now is that the pre-#4666 behavior shouldn't go into 2.0.0. It's obvious that a grammar-based solution will be superior but unless someone wants to get it done pre-2.0.0 (which seems impending?), I think this is ok as a stopgap. I was tempted to go ahead and try and tackle it myself now but don't think I'm quite there yet

[jashkenas]

Cool!

[GeoffreyBooth]

@connec this look good to you?

@helixbass I would say go ahead and try putting this into the grammar as a follow-up PR. Even if it slips into 2.0.1, that would be fine. Just because we release 2.0.0 doesn't mean development will stop.

[connec]

My lexer-foo is still not great, but it passes the most convoluted test I can think of so 👍

do (a = (b) ->) ->
  a
.c

((function(a) {
  return a;
})(function(b) {})).c;

[GeoffreyBooth]

If you wanted to pass arguments into an IIFE, wouldn't you need to not use do?

((a) ->
  {b: a}
)(1).b

[helixbass]

@GeoffreyBooth you can do it via default-valued args to do, at least:

do (a=1) ->
  {b: a}
.b