| Why does IDL strip unary dimensions from structure elements? [message #5240] |
Thu, 26 October 1995 00:00 |
Thomas A McGlynn
Messages: 3
Registered: August 1995
|
Junior Member |
|
|
I've been writing some IDL code which does a lot of generic
stuff with dyamically defined structures, and I've run into
a bug/feature that is causing some problems.
Suppose you have an array of structures, s, with an element
e which is itself an array, e.g,
s = replicate({e:intarr(e1,e2,e3), s1)
If I now extract e from the structure, e.g.,
var = s.(0)
I'll get var with dimensions, e1,e2,e3,s1. So far so good.
However, if the last dimension, or the last several dimensions
of this list are 1, then these dimensions will be stripped
so that the dimensionality of var will be less than expected.
E.g.,
s = replicate({e:intarr(10,1,1)},1)
b = s.(0)
gives b which is a one dimensional vector, not four dimensions.
IDL is consistent: print, size(s.(0)) yields
1 10 2 10
Since it's also an error to have more dimensions on the right
side of an equation than on the left, this means that I have
to keep track of the fact that IDL silently drops these dimensions.
I am also a bit concerned if this behavior may change in different
versions of IDL since my code is intended to be very portable.
Any ideas on clever ways to deal with this? I've seen this with
IDL 3.6 and 4.0.1 on OSF and Ultrix. Is this behavior standard?
Thanks for your help,
Tom McGlynn
mcglynn@grossc.gsfc.nasa.gov
|
|
|
|
| Re: Why does IDL strip unary dimensions from structure elements? [message #5242 is a reply to message #5240] |
Thu, 26 October 1995 00:00  |
thompson
Messages: 584
Registered: August 1991
|
Senior Member |
|
|
Thomas A McGlynn <mcglynn@grossc.gsfc.nasa.gov> writes:
> I've been writing some IDL code which does a lot of generic
> stuff with dyamically defined structures, and I've run into
> a bug/feature that is causing some problems.
Stuff deleted, about the fact that IDL strips off trailing unary
dimensions.
We've also run into this "feature", and it's a big pain. I'm sure it's there
because somebody wanted it somewhere along the line, or thought it was a good
idea.
It appears in many places, not only when working with structures. For example
IDL> a = indgen(3,1,1) & help,a
A INT = Array(3)
or
IDL> a = reform(a,3,1,1) & help,a
A INT = Array(3, 1, 1)
IDL> a = float(a) & help,a
A FLOAT = Array(3)
or
IDL> a = reform(a,3,1,1) & help,a
A FLOAT = Array(3, 1, 1)
IDL> a = a /3 & help,a
A FLOAT = Array(3)
The only way I've come across to get around this problem is to use the SIZE
function to get the original dimensions, and then REFORM to put them back, e.g.
SZ = SIZE(A)
A = A / 3
A = REFORM(A, SZ(1:SZ(0)), /OVERWRITE)
As I said, it's a real pain.
Bill Thompson
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
