Much discussion has been made on IRC concerning symbol names.
The request, mainly, is for imcc to handle sigil characters from other languages which basically equates to exposing a lot to imcc from the high-level language. I won't argue how much of that is good or bad; I'd rather just try to make imcc as friendly as possible.
The state of things:
1) Declared symbols can be handled pretty easily with any character we want to support, imcc just has to track it. It just so happens that we don't allow many non alpha characters at this time.
2) $ is currently used to denote a symbolic register ($I[0-9]+ is an int register).
which is not pre-declared. It just pops up in the instruction stream and
imcc assigns a register.
It is possible that we can stick with $ for temporaries, but make imcc check symbol tables first, and allow people to declare symbols with $ as well. This would solve some issues but might make for some confusing looking code.
In reality it would not really be that confusing if you don't know your variables the same convention as temporaries, but who can guarantee that.
.local PerlString $str .local PerlString $I123 $I123 = "help" $str = $I123 $I124 = 1 + 2 $125 = $124 * 3
The $ no longer stands out as a temporary, so maybe we choose a different character for temp registers (I have suggested using a period C<.>)
.local PerlString $str .local PerlString $I123 $I123 = "help" $str = $I123 .i124 = 1 + 2 .i125 = .i124 * 3
Another option is to use quotes for symbols with sigils, but since most of our code will end up coming from Perl6, that won't be optimizing for the common case.
Finally, I for one, support name mangling. Its arguable how much high level compilers should expose to the back-end compiler. I think, though, that most people prefer to be able to debug PIR code and see the original symbols, and I sympathize.
This is just an example to stimulate discussion. I'd like to hear all sides before making any decisions.
(And remember namespaces when considering solutions)
-Melvin
