On Saturday, March 22, 2014 at 12:02 PM, Bozhidar Batsov wrote:
On Friday, March 21, 2014 at 4:53 PM, Bozhidar Batsov wrote:
On Friday, March 21, 2014 at 4:47 PM, Dmitry Gutov wrote:
Bozhidar Batsov <bozhidar@batsov.com> writes:

Here's a few examples:

"this is #$$"

var = $!

Things are different for:

"this is #$1"

var = $1

Have you tried it in the latest Emacs? For me, only one example

"this is #$$"

is not highlighted (I'll fix that).

$1 gets font-locked immediately, but $$ currently gets font-locked only after some “word-boundary” character gets inserted (like space, newline, etc). 
There's another thing to consider - do built-in global vars should be
font-locked like built-ins or like the other (user-defined) global
variables? Personally I'd font-lock them as built-in to underline their
significance.

Hmmm, maybe. But we also highlight nil, self, true, false, __LINE__,
__ENCODING__ and __FILE__ with font-lock-variable-name-face. Should we
change these, too?
Technically speaking all of those are keywords, not variables. Somewhat odd __LINE__ and friends are
treated at string literals by the Ruby parser. As all of those evaluate to some value unlike most other keywords I guess it makes some sense to font-lock them as variables, but I’d prefer if we used font-locking that makes their special status more apparent. 
There’s one more thing to consider - the special variables aliases defined in `English`. Those are considered de facto built-in (and are actually built-in in some implementations like JRuby; will probably be available out-of-the-box in MRI 3 as well), so I’d suggest treating them the same way as vars like $$, etc.