[Jprogramming] Verb definition + inadvertent train formation.

[Don & Cathy Kelly]

I just ran this
    y=.i.9000
    (10)6!:2 '=i.# y'
0.178201
    (10)6!:2 '=i.9000'
0.178399
verb variations with &.>9000 are in the same ball park -much faster than 
idMatt2 (0.511) but appreciably slower than idMatt 1 and 3


       idMatt1=: (2$])$1{.~>:           NB. Tacit and fast
     (100) 6!:2 'idMatt1 &.> 9000 '
0.063462

idMatt3=: ,~ $ 1{.~ >:
    (100) 6!:2 'idMatt3 &.> 9000 '
0.0643745


Don


On 20/11/2012 5:48 PM, bob therriault wrote:
> Wow, would not have expected that , Devon :)
>
> After testing a substantially reduced challenge on my laptop (Devon, you have some serious metal behind your J installation!)
>     idMate1=: 3 : '(2$y)$(>:y){.1'   NB. Explicit and fast
>     (10) 6!:2 'idMate1 &.> 9000 '
> 0.120056
>       idMatt1=: (2$])$1{.~>:           NB. Tacit and fast
>     (10) 6!:2 'idMatt1 &.> 9000 '
> 0.11531
>       idMatt2=: ,~ $ >: {. 1:             NB. Tacit and slow
>     (10) 6!:2 'idMatt2 &.> 9000 '
> 0.510053
>
> I am guessing that the 1: is the culprit and that these verbs may be used rarely enough that they don't get the benefit of special code
>
> idMatt3=: ,~ $ 1{.~ >:                NB. Same as idMatt2 without use of 1:
>     idMatt3=: ,~ $ 1{.~ >:
>
> 0.114926
>
> Thanks for the insight Devon, as I await Roger's clarification if I have guessed wrong. :)
>
> Cheers, bob
>
> On 2012-11-20, at 5:12 PM, Devon McCormick wrote:
>
>> Here's what I get for timings for some of these methods:
>>
>> NB. Various ways to create ] x ] identity mat.
>>    idMat0=: =@:i.
>>    idMat1=: 3 : '(2$y)$(>:y){.1'  NB.* faster, scales better
>>    idMat2=: ,~ $ >: {. 1:
>>
>>    (10) 6!:2 'idMat0 &.> 36000 38000 40000 42000 44000 46000 48000 50000'
>> 21.7756
>>    (10) 6!:2 'idMat1 &.> 36000 38000 40000 42000 44000 46000 48000 50000'
>> 9.55495
>>    (10) 6!:2 'idMat2 &.> 36000 38000 40000 42000 44000 46000 48000 50000'
>> 23.0792
>>
>> If you need a *:5e4 identity matrix, you're in trouble.
>>
>>
>> On Tue, Nov 20, 2012 at 6:04 PM, bob therriault <bobtherriault at mac.com>wrote:
>>
>>> Devon,
>>>
>>> I like that, or another tacit option
>>>
>>> idMat1=: ,~ $ >: {. 1:
>>>    idMat1 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
>>>
>>> Cheers, bob
>>>
>>> On 2012-11-20, at 1:33 PM, Devon McCormick wrote:
>>>
>>>> I used to have a named verb "idMat" in my own utilities library that was
>>>> defined this way:
>>>> idMat=: =@:i.
>>>>
>>>> but testing this on large arguments led me to formulate this
>>>> idMat=: 3 : '(2$y)$(>:y){.1'  NB.* faster and scales better
>>>>
>>>> or, the same thing tacitly:
>>>> idMat=: (2 $ ]) $ 1 {.~ >:
>>>>
>>>>
>>>> On Tue, Nov 20, 2012 at 2:29 PM, Don & Cathy Kelly <dhky at shaw.ca> wrote:
>>>>
>>>>> A key point that dawned on me, after reading all the helpful comments
>>> and
>>>>> suggestions that have come in, is that I don't need a verb- just a noun
>>> - so
>>>>>   y=: 2 3 4 2 2 3  (it also works for an n,n array)
>>>>>
>>>>> ] u=:=i.# y
>>>>>
>>>>> 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
>>>>>
>>>>> and I can use   u*y    for different 'y''s ( as long as the # is the
>>> same)
>>>>> to get the desired diagonal  elements with off diagonal elements  0
>>>>> I want this because I want to add to the diagonal elements  of a matrix.
>>>>> This is for a Newton Raphson based power system load flow.
>>>>>
>>>>> Thanks to all,
>>>>> Don
>>>>>
>>>>>
>>>>>
>>>>> On 20/11/2012 8:24 AM, Roger Hui wrote:
>>>>>
>>>>>> http://www.jsoftware.com/**jwiki/Essays/Identity%20Matrix<
>>> http://www.jsoftware.com/jwiki/Essays/Identity%20Matrix>: 34 different
>>> ways
>>>>>> to generate the identity matrix of order n; the first is =@i. and the
>>> last
>>>>>> is _&q:@p:@i.
>>>>>> ------------------------------**------------------------------**
>>>>>> ----------
>>>>>> For information about J forums see
>>> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm>
>>>>>>
>>> ------------------------------**------------------------------**----------
>>>>> For information about J forums see
>>> http://www.jsoftware.com/**forums.htm<http://www.jsoftware.com/forums.htm>
>>>>
>>>>
>>>> --
>>>> Devon McCormick, CFA
>>>> ^me^ at acm.
>>>> org is my
>>>> preferred e-mail
>>>> ----------------------------------------------------------------------
>>>> 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
>> ^me^ at acm.
>> org is my
>> preferred e-mail
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>