| where on indices based on a variable condition without many loops [message #88287] |
Mon, 07 April 2014 15:28  |
havok2063
Messages: 24
Registered: December 2012
|
Junior Member |
|
|
I'm looking for a non-loop or single loop solution to the following problem. I need to select "good" indices of a 2d-array based on two conditions.
The original code was this
good = where(error NE 999 AND abs(diff) LT radius, ngood)
where error and diff are 2d-arrays of rough size 4000x700 and radius is a scalar = 2.0. this returned a reformed 1-d array of indices that satisfied the conditions of the where on the 2-d array.
Now I need to replace the scalar radius with an array of (not all) varying numbers, e.g. [2.0,2.0,3.0,2.0,4.0].
These new values reference certain subsets inside the row dimension [0:699] of the 2-d array, e.g. element 0 in radius applies to elements [*,0:31] in the 2-d array, element 1 applies to [*,32:60] , etc, etc. I do know what the indices are that mark these subsets.
I still need "good" to be a 1-d returned array of the correct indices in the full 2-d array satisfying the above conditions.
I've been testing various loop ideas that could work, but how do I remap the indices in the individual subsets returned in the loop iterations to the right indices in the full array?
I feel like the solution involves value_locate, histogram, and unique , but I haven't quite been able to converge.
Any ideas? Thanks.
|
|
|
|
| Re: where on indices based on a variable condition without many loops [message #88288 is a reply to message #88287] |
Mon, 07 April 2014 19:28  |
Heinz Stege
Messages: 189
Registered: January 2003
|
Senior Member |
|
|
Hi Brian,
if I got it right, you don't need any loop. You can work with an
4000x700 array for radius and use exactly the good=... line as in the
original code:
etc=[0,32,61,100,400]
ii=value_locate(etc,indgen(700))
radius_vs_row=([2.0,2.0,3.0,2.0,4.0])[ii]
radius=rebin(transpose(radius_vs_row),4000,700,/sample)
good = where(error NE 999 AND abs(diff) LT radius, ngood)
Cheers, Heinz
|
|
|
|
| Re: where on indices based on a variable condition without many loops [message #88289 is a reply to message #88288] |
Mon, 07 April 2014 20:50 |
havok2063
Messages: 24
Registered: December 2012
|
Junior Member |
|
|
Excellent. That did the trick and was exactly what I was looking for. Thanks a bunch.
Thanks, Brian
On Monday, April 7, 2014 10:28:04 PM UTC-4, Heinz Stege wrote:
> Hi Brian,
>
>
>
> if I got it right, you don't need any loop. You can work with an
>
> 4000x700 array for radius and use exactly the good=... line as in the
>
> original code:
>
>
>
> etc=[0,32,61,100,400]
>
> ii=value_locate(etc,indgen(700))
>
> radius_vs_row=([2.0,2.0,3.0,2.0,4.0])[ii]
>
> radius=rebin(transpose(radius_vs_row),4000,700,/sample)
>
> good = where(error NE 999 AND abs(diff) LT radius, ngood)
>
>
>
> Cheers, Heinz
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
