[q-lang-users] Q 7.2 RC 1 issue 1: Nomenclature

[John C. <co...@cc...>]

I am not happy with the names of the syntactic vs. semantic number
predicates.  Currently it is isint, isrational, isreal, etc. for syntactic
(representation) predicates, and isintnum, isratnum, iscompnum etc. for
semantic (value) predicates.  This is intended to be consistent with
isbool, isfile, islist etc., which are syntactic; however, it is only
with numbers that this dichotomy between syntactic and semantic types
is really necessary.

I want to argue two things: that semantic predicates are more fundamental
and deserve the shorter names, and that (in the alternative) the
particular names "isintnum", "isratnum", etc.  are very unfortunate.

Fundamentally, a number is a number, and its representation is a
low-level issue.  What counts is that a number actually is or is not
complex, or rational, or an integer; not what its representation may be.
Representations matter only when considering efficiency, which is a
second-order concern in Q.  It is better to use isinteger to determine if
something really is an integer rather than to determine if its internal
representation is a GMP integer.

(By the way, does Q use any sort of fixed-width integer representation
internally, or are all integers in GMP format internally?  I'm just
curious.)

In addition, Q's numeric tower is now modeled on that of Scheme,
but the names are exactly the opposite of Scheme's.  In Scheme, the
predicate "rational?" tests for a rational number, not for a (numerator,
denominator) representation -- and the nickname for such a representation
is "ratnum".  So people like me who use both Scheme and Q will need
to remember that the names have exactly opposite meanings in the two
languages.  This is regrettable.

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