If there was ever a a candidate for "lazy evaluation", "catalog" would be
it.



On Fri, May 9, 2014 at 1:00 PM, Devon McCormick <[email protected]> wrote:

> I've used "catalog" in real-world apps but, as Dan indicates, it was with
> arguments that gave fairly small results.
>
>
>
> On Thu, May 8, 2014 at 8:30 PM, 'Pascal Jasmin' via Programming <
> [email protected]> wrote:
>
>> > one reason for its rarity is, at least, that in any real-world context,
>> a complete MxN catalog would be prohibitively expensive
>>
>> its very cheap compared to A. and filtering permutations after you've
>> generated them.  In the resource matching application I linked to, where 1
>> user ends up with 1 resource, its also possible to prefilter the arguments
>> to catalogue removing options where forced choices exist, and thus
>> significantly reducing an exponential algorithm to allow solutions to
>> moderately large problem spaces.
>>
>> Its fast compared to backtracking loop solutions in other languages even
>> when they abort after finding first valid set of matches.  It seems like a
>> very clean J way of approaching fairly hard problems.  Polynomial and
>> higher problems benefit from strategies that reduce search space.
>>
>> Using catalogue to analyze n moves with an average x possibilities per
>> move has an expensive x ^ n search space.  But reducing any move
>> possibility to 1 reduces both average x, and reduces n by 1
>>
>>
>> ----- Original Message -----
>> From: Dan Bron <[email protected]>
>> To: "[email protected]" <[email protected]>
>> Cc:
>> Sent: Thursday, May 8, 2014 6:39:14 PM
>> Subject: Re: [Jprogramming] a cool permutation verb
>>
>> I don't see catalog used frequently in natural J code. It does come up on
>> these Forums from time to time, mostly in contrived situations. In my own
>> code, I sometimes use it to generate J sentences in order to compare
>> performance statistics for semantically-equivalent phrases. But outside of
>> that, I don't see it too often. I suspect one reason for its rarity is, at
>> least, that in any real-world context, a complete MxN catalog would be
>> prohibitively expensive.
>>
>> So it's unsurprising that APL didn't have a primitive catalog function.
>> Of course, you could express it as {jot} {dot} {comma}, dot as being outer
>> product (and jot in this context a placeholder, not unlike [:) with the
>> usual caveats about the differences between J's rank and APL's
>> multidimensionality (ie you could expect catalog to work for 1D vectors in
>> APL and generalized lists in J).
>>
>> So, when he got his chance to "rationalize" APL and rebuild it from
>> the ground up as J, did Ken have a change of heart about catalog's
>> centrality to the discipline of computer science? Is that why he assigned
>> to it that scarce and desirable resource, a primitive spelling? A
>> one-character primitive, no less?
>>
>> I suspect not. Recall the other valence of {, to wit: from. You might
>> think having this kind of partner is an even greater endorsement of
>> catalog's virtues ... until you realize:
>>
>>    row_2n3n6_col_7n13 =: 2 3 6 ; 7 13
>>    table =: i. 10 20
>>
>>    (< row_2n3n6_col_7n13) { table NB. "from"
>> 47  53
>> 67  73
>> 127 133
>>
>>    ({ row_2n3n6_col_7n13)         NB. "catalog"
>> +---+----+
>> |2 7|2 13|
>> +---+----+
>> |3 7|3 13|
>> +---+----+
>> |6 7|6 13|
>> +---+----+
>>
>>    ({ row_2n3n6_col_7n13) { table NB. "catalog from"
>> 47  53
>> 67  73
>> 127 133
>>
>> Ah, "from" *is* "catalog". Catalog is a primitive because from needed a
>> thematically related definition for monad { (and from, of course,  got a
>> precious one-character primitive because it *is* central to array
>> programming). And, incidentally, this is why the results of catalog are
>> boxed, even though they don't need to be, and we end up writing >@:{
>> instead of plain { every time we use it.
>>
>> That is, by definition, the results of { will be homogenous in both type
>> and shape, and so we know for certain boxes aren't required ... except that
>> dyad { has a scalar left rank, and so to completely address an atom (or
>> anything more granular than an item, ie a Cartesian point in the frame
>> specified as a vector of indices, one per axis), each selection must be
>> boxed.
>>
>> True story.
>>
>> -Dan
>>
>>
>>
>> Sent from my iPhone
>>
>>
>> > On May 8, 2014, at 5:34 PM, "'Pascal Jasmin' via Programming" <
>> [email protected]> wrote:
>> >
>> > Its the same.  Just more descriptive name than catalogue perhaps.
>> Google seems to only find J references to query "catalogue function
>> permutations".  I feel as though catalogue and my use of it is a
>> generalizable pattern, and the J designers agreed, but I don't know if
>> other languages did something similar, and just used a different name?
>> >
>> > Was catalogue in APL?
>> >
>> >
>> > ----- Original Message -----
>> > From: David Lambert <[email protected]>
>> > To: programming <[email protected]>
>> > Cc:
>> > Sent: Thursday, May 8, 2014 4:50:40 PM
>> > Subject: Re: [Jprogramming] a cool permutation verb
>> >
>> > Indeed cool, does it differ from ,@{ :[:
>> >> Date: Thu, 8 May 2014 10:53:53 -0700 (PDT)
>> >> From: "'Pascal Jasmin' via Programming"<[email protected]>
>> >> To: Programming forum<[email protected]>
>> >> Subject: [Jprogramming] a cool permutation verb
>> >> Message-ID:
>> >>     <[email protected]>
>> >> Content-Type: text/plain; charset=utf-8
>> >>
>> >> cperm =: [: , [: < "1 [: > [: ,"0 1/ each/ (_2&}. , [: ,"0 0/ each/
>> _2&{.)
>> >>
>> >> this creates constrained permutations where each position in the list
>> can only be drawn from its own set:
>> >>
>> >> ? ?cperm 3 4 ; 1 2
>> >> ?????????????????
>> >> ?3 1?3 2?4 1?4 2?
>> >> ?????????????????
>> >>
>> >> ? ?cperm 0 ; 3 4 ; 1 2; 5
>> >> ?????????????????????????????????
>> >> ?0 3 1 5?0 3 2 5?0 4 1 5?0 4 2 5?
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
>
>
>
> --
> Devon McCormick, CFA
>
>


-- 
Devon McCormick, CFA
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to