On 04/08/2014 08:56 PM, Stefan Monnier wrote:
>>> Why limit this to strings?  It makes just as much sense to do it for
>>> comments, doesn't it?
>> It's a bit harder to implement for comments (as in, rewrite up-list
>> using parse-partial-sexp), and I'm not sure it's as useful. Probably not
>> too bad though.
> 
> I don't follow you.  It doesn't seem hard: just use forward-comment to
> skip over the comment instead of forward-sexp to skip over the string.
> Otherwise, the code should be pretty much identical for comments as
> for strings.

Right now, the main case is based on scan-lists. Say we have something
like this JavaScript code:

  [ 1, 2, /* NO: [ 3.1, 3.3, 3.3 ] */ 3, 4 ]
  abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOP

If I understand what you mean by "do it for comments", if point begins
at `D' (well, between `C' and `D'), up-list should take us to `p', then
`i', then `a'. The problem is that `scan-lists' called with FROM `p',
COUNT -1, and DEPTH 1 will move directly from `p' to `a', bypassing
position `i'. We're going to need fancier logic to make that work correctly.

Anyway, even if you did stop at position `i', you couldn't do much: you
can't manipulate that comment as a sexp. The use case that motivated my
change is quickly replacing the string that happens to contain point,
like vim can with `di"' command. Since we don't have any commands that
manipulate comments as first-class lexical objects, making up-list stop
on comments seems less useful to me. I'd rather just treat comments as
pseudo-whitespace like the rest of Emacs does.

It'd be nice to treat comments as sexps, but that change is out of scope.