On 12/07/2012 07:57 PM, Danny Yoo wrote:
Or maybe using a : or something to separate rows for 2D arrays?
(array 1 2 : 3 4)
(array 1 0 0 0 0 0
: 0 1 0 0 0 0
: 0 0 1 0 0 0
: 0 0 0 1 0 0
: 0 0 0 0 1 0
: 0 0 0 0 0 1)
What about something like:
(array #:cols 6
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1)
where an extra #:cols argument lets the system know now many columns
there are per row?
This works right now, and is pretty close:
(make-mutable-array
#(6 6)
#(1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1))
It gets a little uglier when you want to override TR's generalization
for polymorphic, mutable types. For example, you'd have to apply (inst
make-mutable-array Byte) to get a (Mutable-Array Byte) instead of a
(Mutable-Array Integer). No problem in untyped Racket, of course.
There's no shape check on the data, though, so it'd be really easy to
make an array with values in the wrong places.
As far as using (array [[...]]) - no hashes - it used to work like that.
The `array' macro would check the shape of the parens to determine
whether (e ...) was meant to be a row like [list 1 2 3] or an element
like (list 1 2 3). The problem is, that's a syntax property, which
macros rarely preserve, and Scribble couldn't render them properly. In
general, it would be too easy for array literals computed or transformed
at syntax time to lose their shape and meaning.
Using commas, dots and semicolons is out of the question. Colons would
clash with TR's use in types. Other punctuation is sinfully ugly. I'm
afraid the only real option for something with a little less visual
clutter is a reader extension or a custom language. I considered it, but
in the end I rejected #lang math. I just didn't have the hubris for it. :D
Neil ⊥
____________________
Racket Users list:
http://lists.racket-lang.org/users
