comp.lang.idl-pvwave archive: archive

Home » Public Forums » archive » Arrays of Structures

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: Arrays of Structures [message #52580 is a reply to message #20472]

Fri, 09 February 2007 12:51

JD Smith
Messages: 850
Registered: December 1999

Senior Member

On Fri, 09 Feb 2007 02:13:56 -0800, Mick Brooks wrote:

> If I try, I get the following:
>
> IDL> struct=replicate({a:lindgen(100,100,100)},100)
> IDL> print,(m=memory(/HIGHWATER))/1024/1024.,"MB"
> 385.836MB
> IDL> val=struct[10].a[4]
> IDL> print,((n=memory(/HIGHWATER))-m)/1024/1024.,"MB Extra"
> -3.81445MB Extra
> IDL> val3=struct.a[10,4]
> IDL> print,(memory(/HIGHWATER)-n)/1024/1024.,"MB Extra"
> 0.00000MB Extra
> IDL> HELP, val, val3
> VAL LONG = 4
> VAL3 LONG = Array[100]
> IDL> PRINT, val3
> 410 [+ another 99 of the same]
>
> The problem case doesn't use any extra memory (great!), but it gives a
> different result (boo!).

I probably should have used a better example, that was confusing. The
reason val3 "works" is because "a", the field being de-referenced,
already has 3 dimensions. So it grabs the [10,4] element of all a's,
and then proceeds to thread that across all 100 structures in the
structure array, creating a new array in the process. Here's a better
example illustrating this issue:

IDL> val4=(struct.a)[10,4,1,1]
IDL> val5=struct[10].a[4,1,1]
IDL> print,val4,val5
10410 10104
IDL> val6=struct.a[10,4,1,1]
% Subscript range values of the form low:high must be >= 0, < size, with low
<= high: <No name>.

Here's the real issue: since "struct.a" doesn't exist as an array
anywhere in memory, but instead is a composite entity, it must be
formed anew by:

a) allocating a new array of enough memory to hold all of
n_elements(struct) x size(a,/dimensions) values.
b) going through each structure in the structure array, and copying its
copy of "a" into the new array.

I agree the precedence table is misleading on this point. I do think
there is a reasonable argument that IDL should understand
struct.a[10,4,1,1] and not need to first create a 325MB array to
de-reference it. Here's an easy rule of thumb though: in order to
avoid the "new array creation" process described above, just attach
all indices as close as possible to the quantity they are indexing.
"10" goes with struct (we want the 10th struct). [4,1,1] goes with
"a" (we want element [4,1,1] of a).

JD

Report message to a moderator

[Message index]

		Arrays of Structures By: Ben Tupper on Wed, 21 June 2000 00:00
		Re: Arrays of Structures By: news.qwest.net on Thu, 08 February 2007 12:52
		Re: Arrays of Structures By: JD Smith on Thu, 08 February 2007 12:42
		Re: Arrays of Structures By: Paul Van Delst[1] on Fri, 09 February 2007 08:01
		Re: Arrays of Structures By: Mick Brooks on Fri, 09 February 2007 02:13
		Re: Arrays of Structures By: Michael Galloy on Thu, 08 February 2007 12:21
		Re: Arrays of Structures By: news.qwest.net on Thu, 08 February 2007 11:44
		Re: Arrays of Structures By: Paul Van Delst[1] on Thu, 08 February 2007 11:32
		Re: Arrays of Structures By: Mick Brooks on Thu, 08 February 2007 11:27
		Re: Arrays of Structures By: Brian Larsen on Thu, 08 February 2007 11:11
		Re: Arrays of Structures By: news.qwest.net on Thu, 08 February 2007 11:09
		Re: Arrays of Structures By: Paul Van Delst[1] on Thu, 08 February 2007 11:02
		Re: Arrays of Structures By: Brian Larsen on Thu, 08 February 2007 10:59
		Re: Arrays of Structures By: Mick Brooks on Thu, 08 February 2007 10:04
		Re: Arrays of Structures By: Brian Larsen on Thu, 08 February 2007 09:41
		Re: Arrays of Structures By: JD Smith on Fri, 09 February 2007 12:51

Previous Topic:	image rotation Fits headers?
Next Topic:	.sav format

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Oct 08 18:10:50 PDT 2025

Total time taken to generate the page: 0.00553 seconds