|
|
|
|
| Re: array chicanery [message #28048 is a reply to message #28047] |
Thu, 15 November 2001 17:29  |
nrh
Messages: 19
Registered: September 2000
|
Junior Member |
|
|
Mark Hadfield wrote:
>
> From: "Rochelle Hatton" <nrh@imag.wsahs.nsw.gov.au>
>> I've experienced some weird business with arrays, and I'm
>> hoping somebody can explain.
>
> Sure can!
>
>> I have been using the IDL function PCOMP, calculating the principal
>> components of some data. If I feed in a float array to the function,
>> eg.
>> result=PCOMP(data, coefficients=eigenvectors, $
>> eigenvalues=eigenvalues, /covariance, /standardize)
>>
>> I get a result, but when I redisplay the data array, it has changed.
>> Since PCOMP only returns a result, what is it doing to my original
>> array, and why?
>
> PCOMP doesn't "only return a result".
Thanks - and yes I have had my doubts about PCOMP. I have used SVDC with
much more success, although I am even a bit sceptical about that, seeing it
uses code from Numerical Recipes...
> Looking at the source code for PCOMP and searching for "Array" (the name of
> the first argument) one sees...
>
> ;Standardize the columns of the input array with a mean of 0.0
> ;and a variance of 1.0
> if KEYWORD_SET(Standardize) ne 0 then $
> Array = STANDARDVAR(Array, Dimension, Double = Double)
>
> So "Array" is modified and if this corresponds to an argument passed by
> reference then the original will be modified. In your case you have passed
> it an array ("data") so it is fair game. For more info on this issue see
> "Procedures and Functions/Parameter Passing Mechanisms" in "Building IDL
> Applications". This is a common source of confusion, usually in the opposite
> sense: people pass array segments and structure tags to routines and are
> surprised when they *aren't* modified.
>
> The easiest way of protecting your original array is to pass PCOMP a copy of
> the data. This should work
>
> result=PCOMP(reform(data),...)
>
> BTW can anyone suggest a better general purpose function for this purpose?
>
> I will refrain from commenting on whether PCOMP's behaviour represents good
> programming practice... But I do think they could have warned you.
>
> ---
> Mark Hadfield
> m.hadfield@niwa.cri.nz http://katipo.niwa.cri.nz/~hadfield
> National Institute for Water and Atmospheric Research
--
"For every complex problem, there is a solution that is simple, neat, and wrong."-H. L Mencken (1880-1956)
Rochelle Hatton
Department of Nuclear Medicine and Ultrasound
Westmead Hospital ph(02) 9845 7223
|
|
|
|
| Re: array chicanery [message #28050 is a reply to message #28048] |
Thu, 15 November 2001 16:35 |
Mark Hadfield
Messages: 783
Registered: May 1995
|
Senior Member |
|
|
From: "Rochelle Hatton" <nrh@imag.wsahs.nsw.gov.au>
> I've experienced some weird business with arrays, and I'm
> hoping somebody can explain.
Sure can!
> I have been using the IDL function PCOMP, calculating the principal
> components of some data. If I feed in a float array to the function,
> eg.
> result=PCOMP(data, coefficients=eigenvectors, $
> eigenvalues=eigenvalues, /covariance, /standardize)
>
> I get a result, but when I redisplay the data array, it has changed.
> Since PCOMP only returns a result, what is it doing to my original
> array, and why?
PCOMP doesn't "only return a result".
Looking at the source code for PCOMP and searching for "Array" (the name of
the first argument) one sees...
;Standardize the columns of the input array with a mean of 0.0
;and a variance of 1.0
if KEYWORD_SET(Standardize) ne 0 then $
Array = STANDARDVAR(Array, Dimension, Double = Double)
So "Array" is modified and if this corresponds to an argument passed by
reference then the original will be modified. In your case you have passed
it an array ("data") so it is fair game. For more info on this issue see
"Procedures and Functions/Parameter Passing Mechanisms" in "Building IDL
Applications". This is a common source of confusion, usually in the opposite
sense: people pass array segments and structure tags to routines and are
surprised when they *aren't* modified.
The easiest way of protecting your original array is to pass PCOMP a copy of
the data. This should work
result=PCOMP(reform(data),...)
BTW can anyone suggest a better general purpose function for this purpose?
I will refrain from commenting on whether PCOMP's behaviour represents good
programming practice... But I do think they could have warned you.
---
Mark Hadfield
m.hadfield@niwa.cri.nz http://katipo.niwa.cri.nz/~hadfield
National Institute for Water and Atmospheric Research
--
Posted from clam.niwa.cri.nz [202.36.29.1]
via Mailgate.ORG Server - http://www.Mailgate.ORG
|
|
|
|
| Re: array chicanery [message #28051 is a reply to message #28050] |
Thu, 15 November 2001 16:28 |
nrh
Messages: 19
Registered: September 2000
|
Junior Member |
|
|
That works. Thankyou. I think I know what happens now.
> Lousy programming. On someone's part other than yours. :-)
>
> Try this:
>
> IDL> result=PCOMP(data+0,coefficients=eigenvectors,$
> eigenvalues=eigenvalues,/covariance,/standardize)
> IDL> Help, data
>
> Does that help?
>
> Cheers,
>
> David
>
-
"For every complex problem, there is a solution that is simple, neat, and wrong."-H. L Mencken (1880-1956)
Rochelle Hatton
Department of Nuclear Medicine and Ultrasound
Westmead Hospital
|
|
|
|
| Re: array chicanery [message #28052 is a reply to message #28051] |
Thu, 15 November 2001 16:20 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Rochelle Hatton (nrh@imag.wsahs.nsw.gov.au) writes:
> I've experienced some wierd business with arrays, and I'm hoping somebody can explain.
> I have been using the IDL function PCOMP, calculating the pricipal components of some data.
> If I feed in a float array to the function,
> eg.
> result=PCOMP(data,coefficients=eigenvectors,eigenvalues=eige nvalues,/covariance,/standardize)
>
> I get a result, but when I redisplay the data array, it has changed. Since PCOMP only returns
> a result, what is it doing to my original array, and why? There are no common variables, as far
> as I know, so the original data should not change. I apologise in advance if this is blindingly
> obvious, and of course, I can just make a copy of the original data into another variable to use
> later, I am just curious as to what is going on.
> Any thoughts?
Lousy programming. On someone's part other than yours. :-)
Try this:
IDL> result=PCOMP(data+0,coefficients=eigenvectors,$
eigenvalues=eigenvalues,/covariance,/standardize)
IDL> Help, data
Does that help?
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438, E-mail: david@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
