Px = n_add Py, Pz # create new Px as sum(Py, Pz)
and that overloaded infix functions have to return the result.
For Python, Lisp and probably more HLLs the same is of course needed for unary opcodes:
abs Px, Py # use existing abs, Px exists Px = n_abs Py # create new abs result PMC
This is of course just a fancy way to write
n_abs Px, Py # same
but it makes it more clear that a new value is assigned.
For the existing unary opcodes with PMC variants:
abs, neg, not, nots (don't we have more?)
the same scheme as to infix opcodes apply: if the passed in C<dest> PMC is NULL, a new result PMC is created. I don't know, if we should have distinct inplace variants too:
abs Px # VTABLE_i_absolute()
I'd like additionally to have PMC variants of mostly existing vtables that currently return native types only:
PMC *get_str() # string context PMC *get_num() # numeric context PMC *get_int() # get integer (maybe) PMC *get_list() # list context PMC *hash() # get hash value - we have INTVAL hash()
The exsiting C<INTVAL get_bool()> should better be C<get_bool_native> so that we have too:
PMC *get_bool() PMC *get_bignum() # exists PMC *get_complex() # return a new complex with SELF.value
I think these all should just return a new result PMC.
Comments welcome, leo
