On 5/9/07, Allison Randal <[EMAIL PROTECTED]> wrote:
Patrick R. Michaud wrote:
>
> Personally, I would think that the standard approach for going from
> a String PMC to a method invocation would be via the find_method
> opcode. But that's just me.
That goes back to the philosophical question of "Is PIR a language to be
generated by compilers, or is it a language for humans to use?" It's
about half-way in between at the moment, and the conflict shows.
true. In a way, PIR already has some double syntax sugar: 1 for generating,
1 for humans; consider the sub invocation (or method for that matter)
syntax: PIRallows both the "long" version (.pcc_begin/_end) and the "short"
version (" foo() ")
For generated code, it's no problem to require a methodname lookup for
every call. For hand-written code it's insanely annoying (which is the
reason the syntactic sugar was introduced in the first place).
Perhaps PIR's next stage of evolution is to split off into two
languages, one for generation and one for humans.
Allison
kjs
