On Thu, July 7, 2005 11:40 am, Roger Browne said: > Matt Diephouse wrote: >> Should as much functionality as possible be put into the core PMCs? > > I'd like to see parrot include a set of core PMCs that implement fairly > pure abstractions, without any language-specific stuff (such as > automatic conversions, preconceived notions of which string values are > true/false etc). It would be easier for a language implementor to add > functionality to these, than to modify PMCs that include some perl-isms > or python-isms. >
One specific issue this brought to mind is the Pair PMC. The pair pmc is not really a pair. It is what I would call a dictionary entry. The design and writing of the pair assumes the pair will be used for a key->value type entry. But I just want to have a pair that stores two PMCs. Thus, I currently have the option of using FixedPMCArray and setting the size to 2, or writing a "real" pair pmc. Using the FixedPMCArray has a performance hit, so I would like to use a "real" pair pmc. The problem with the pair pmc is the inability to change the key. So the existing pair PMC should be modified to use indexing to access elements 0 and 1 (get_pmc_keyed_int and set_pmc_keyed_int). This is what you think of when you think of a pair. Then a DictionaryEntry PMC can be derived from this that protects the key from changing and also includes the set_pmc_keyed and is_equal functions. John
