Nicholas Clark <[EMAIL PROTECTED]> wrote: > Can 2 different bytecode segments each try to define a new infix operator? > If so, how do they number their infix operators to avoid a clash?
The same problem arises with user defined opcodes or generally for a name => index mapping for which the assembler and the Parrot run cores need the same view. I see three possible solutions: 1) demand predeclaration of such resources in main (at compile time): $I0 = register_infix "__hyper_add" # assign next infix op number or a variation of this theme: demand that the assembler and runtime "executes" such registration in the same order. 2) do a runtime lookup infix "__hyper_add", Pd, Pl, Pr 3) treat unknown infix ops as ordinary multi sub calls Pd = "__hyper_add"(Pl, Pr) With runcores that can rewrite the bytecode all three boil down to the same and fast PIC-based operation. > Nicholas Clark leo
