Re: [Chicken-users] sequences egg

[Alan Post]

On Fri, Nov 19, 2010 at 03:20:42AM -0500, Felix wrote:
> From: Alan Post <address@hidden>
> Subject: Re: [Chicken-users] sequences egg
> Date: Thu, 18 Nov 2010 08:37:35 -0700
> 
> > On Thu, Nov 18, 2010 at 10:03:44AM -0500, Felix wrote:
> >> Hello!
> >> 
> >> 
> >> I've put together a little library of generic "sequence" operations,
> >> and would like to get some feedback, since I'm not sure about the
> >> nomenclature and API. Find it here:
> >> 
> >>   http://wiki.call-cc.org/eggref/4/sequences
> >> 
> >> And in the repo, usual place. It has a test-suite, but surely contains
> >> some bugs, and is a bit of a test for the "fast-generic" extension.
> >> 
> >> Comments, suggestions, rants or critique is welcome.
> >> 
> >> (I think it needs a very recent chicken, but am not quite sure,
> >> it may work with older versions)
> >> 
> > 
> > When I write programs with iterators, I find myself using:
> > 
> >   map
> >   zip/unzip
> >   chain
> >   remove/filter
> > 
> > nearly every time.  I use:
> > 
> >   tee (copy)
> >   pairwise
> >   count
> >   repeat
> 
> Could you provide examples for `chain', `pairwise' and `repeat'?
> 

Chain is like append, it takes a list of sequences and returns all
of sequence 1, all of sequence 2, &c.

I use chain when my data comes from different sources (say two
separate files) but I can treat it as a single stream.


pairwise could be (now that #!optional handling is fixed!):

  (use test)

  (define (pairwise a #!optional (b (cdr a)))
    (if (null? (cdr b))
        `((,(car a) ,(car b)))
        (cons `(,(car a) ,(car b)) (pairwise (cdr a) (cdr b)))))

  (test '((a b) (b c) (c d)) (pairwise '(a b c d)))

I have used pairwise when I have a file I need to break into smaller
pieces, I'll take the size of the file, divide it into chunks, say
1000 lines, and then use pairwise to generate my boundaries:

  (pairwise (0 1000 2000 2256))

So each unit processing a portion of the file gets a start (inclusive)
and end (exclusive) boundary.

I'll also us it when I need to assert some condition about my data,
say that each number is greater than the number before it in a
stream.


repeat acts like circular-list from srfi-1:

  (repeat 'a) => (a a a ...)
  (repeat '(a b c)) => (a b c a b c ...)

or is that last one?:

  (repeat 'a 'b 'c) => (a b c a b c ...)

I use repeat along with zip to append information to a stream that
I'll later modify, say if I'm deciling a bunch of data, I'll provide
a default decile of 0 with repeat before performing the calculation.


> > One thing I notice missing is that you have linear-sequence and
> > random-access-sequence, but what about something like a pipe, which
> > is a linear-sequence with (say) the additional constraint of only
> > being able to traverse it once?  Similarly, if you had a routine
> > like repeat, it is a linear-sequence that is non-finite. 
> 
> Good points. How would you express that operationally?
> 

pipe I would say acts like a dynamic variable.  Any reference to the
iterator operates on the same stream, and once the iteration is
exhausted it can't be used again.  If pairwise worked, it would
return:

  ((a b) (c d))
 
because the separate references to |a| and |b| are using the same
underlying source.

non-finite sequences only make sense to me when paired with a finite
sequence using zip.  The idea is that the shortest sequence in a set
causes iteration to step for all other sequences in that set:

  (zip '(foo bar baz)
       '(1 2 3 4)
       (repeat 'a))
 
  => (foo 1 a) (bar 2 a) (baz 3 a)

-Alan
-- 
.i ko djuno fi le do sevzi