Hi Slava, Hi Factor list,
I'm sure this must have been discussed before, but I was wondering if
the sequence combinators would be better served by an underlying
iterator protocol in addition to the current random access one. I can
see a couple of advantages of this:
- sequence combinators could be used with streams
- sequence combinators could be used with lazy lists
- outsources iteration to sequence classes, affording opportunites for
optimization
(By the latter I mean that e.g. a fixnum-length container could optimize
to fixnum+fast)
I had a quick go at an iterator protocol by way of proof-of-concept.
(old tuple notation I'm afraid). This protocol is more like python's
than STL or java. The 'each' combinator was pretty easy:
GENERIC: next ( iterator -- end? )
GENERIC: elem ( iterator -- element )
GENERIC: iter ( seq -- iterator )
USING: math.private ;
TUPLE: fixnum-iterator num end ;
C: <fixnum-iterator> fixnum-iterator
M: fixnum iter ( seq -- iterator ) 0 swap <fixnum-iterator> ;
M: fixnum-iterator elem ( iterator -- element ) fixnum-iterator-num ;
M: fixnum-iterator next ( iterator -- end? )
dup fixnum-iterator-num 1 fixnum+fast
dup pick fixnum-iterator-end fixnum<
[ swap set-fixnum-iterator-num t ] [ 2drop f ] if ;
: each-iter-loop ( iter quot -- )
2dup [ elem ] dip call
over next [ each-iter-loop ] [ 2drop ] if ; inline
: each-iter ( seq quot -- )
[ iter ] dip each-iter-loop ; inline
( scratchpad ) 3 [ . ] each-iter
0
1
2
What do you think?
Cheers,
Phil
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Factor-talk mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/factor-talk
