[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: library separation



Alex Shinn scripsit:

> Could we have a (srfi 114 base) library that provides only the
> predicates and accessors, and possibly some subset of default
> predicates?  (notably eq? is special)

I'm okay with this idea in the abstract, but I think that list is
*too* short.  If there's to be a base library, it should contain what a
client of comparators needs:  predicates (3), accessors (4), primitive
applicators (5), which are often more useful than accessors, comparison
syntax (7), comparison predicates (5), and maybe interval predicates (4)
and min/max (2).  Plus eq, eqv, and equal (3), which need to be special
cased by some libraries.  That's 27 to 33 identifiers.  By comparison,
the full library has 76 identifiers.

> Nitpick: comparator-comparison-procedure? would be better named
> comparator-has-comparison-procedure?.

Excellent point.  Fixed.

> But do we need this given the accessor?

Yes.  The accessor always returns a procedure, even if it's an
automatically generated error procedure; if you don't want to call it
blindly and try to catch the error, you need the predicate.

> You're overloading "compare" to mean both the abstraction and ordered
> comparisons.  Could we rename the ordering to be something like "less"
> or "order"?

The abstraction is always called a comparator.

-- 
One Word to write them all,             John Cowan <cowan@xxxxxxxx>
  One Access to find them,              http://www.ccil.org/~cowan
One Excel to count them all,
  And thus to Windows bind them.                --Mike Champion