[ruby-talk:03729] Re: Why it's quiet -- Ruby features -- Arrays

Hi,

Dave Thomas wrote:

> Conrad Schneiker <[e-mail]> writes:
>
> > Any chance that Ruby 1.6+ will have a built-in standard
> > multidimensional array?  Especially since 2-D arrays are a very
> > common data type that seem to turn up frequently in tables for
> > general data processing, plus for graphics, math, engineering, and
> > physics?
>
> Apart from performance, are there any advantages to building this in,
> rather than using the array and matrix library modules?

Apart from performance, is there any advantage to flying 1,000+ miles
across the ocean rather than swimming? :-) I wasn't thinking about
performance differences, but if it was a 5x difference for the 2-D case, I
think that would be sufficient reason.

Please excuse my imprecision with regard to "built in"--I tend think of
the standard (i.e. as shipped, out-of-the-(tar-file)-box) Ruby library as
an integral (i.e. official) part of Ruby. My primary concern is that
multidimensional arrays be supported "out of the box" as a standard
class/objects/datatype that everyone refers to the same way.

> In general, I'd suggest that one of the charms of Ruby is its small
> conceptual size.

Well, it's _relatively_ small conceptual size compared to (say) C++ is
certainly nice, but it's one of several charms. If the conceptual size of
Ruby were any smaller, I think many people would regard this as
suboptimal. For example, Tcl can make the same claim about small
conceptual size, but this has previously led to a variety of incompatible
extensions. What are the odds that Ruby has reached the optimum conceptual
size, all things considered? It may be pretty close, but I don't think it
is has yet reached that size. From a user's point of view, not having
something as common as multidimensional arrays available "out of the box"
in a way that is the most natural generalization of 1-D arrays makes the
use of Ruby somewhat more conceptually complex when they need arrays.

> Even though many of the basic types are built in, we can pretend that
> they're not, that instead they are add-in library modules. The built in
> types do not add any new syntax to the language. My vote for future
> enhancements would be that if it _can_ be done outside the core of the
> language, it _should_ be done outside the core.

Well, obviously you wan't to be careful about what you add to the core,
but I think it is going too far to always do this when possible. After
all, we want Ruby to be supremely _pragmatic_! :-) ... especially for
students, scientists, engineers, and currency crunchers!  In cases where
there is a syntactic advantage in terms of greater clarity and consistency
and naturalness of expression for very commonly used items (and/or
substantial performance advantages), I think they should be done in the
core--if the feature in question is widely used. I don't know to what
extent these considerations apply to multidimensional arrays, but I do
know that 2-D arrays are very widely used. Multidimensional arrays are not
some exotic or comparatively rare data type, and I don't think adding them
to Ruby is going to open Pandora's box of conceptual feature bloat.

In any case, I am all for multidimensional arrays being part of standard
Ruby (as the out-of-the-box programming language), regardless of whether
it is built into the core for performance/syntactic reasons or implemented
as a standard Ruby module.

--
Conrad Schneiker
(This note is unofficial and subject to improvement without notice.)