> I'm thinking that we're going to have one cmp-style vtable comparison
> function for strings and one for numbers. Anything else and people can go
> override the parser if they really need to.
Good. We don't need to burden the low-level interface with twelve
vtable methods when two will do for the vast majority of the cases.
> >One additional datapoint to overload your brain with is to consider
> >the ambiguity of equality and comparison. Unicode normalization:
> >is A + grave equal to Agrave? Is Agrave less than Aacute? Unicode
> >collation combined with language/locale-specific rules.
>
> Comparisons on Unicode data will do it on the Unicode collation version of
> the string data. Equality checking will be done either on normalized data
We need to include in our design a spot for the customization hooks, though.
> or whatever representation it's in, depending on Larry's call. (I'd prefer
> normalization form C, but I'm not sure the regularity's worth the CPU cost.
> Telling the programmer to beware might be sufficient)
The NFC seems like the way to go.
--
$jhi++; # http://www.iki.fi/jhi/
# There is this special biologist word we use for 'stable'.
# It is 'dead'. -- Jack Cohen