| Re: Make_array() and using arrays as subscripts [message #46848] |
Tue, 10 January 2006 01:09  |
Liberum
Messages: 48
Registered: September 2005
|
Member |
|
|
Thanks Brad,
Around midnight last night I solved the problem by simply do this (
my_array = my_array[*, good_indices] ). I didn't think about the
asterisk *, i.e. cutting all dimensions with the one dimension array:
good_indices.
Thanks,
Sheldon
|
|
|
|
|
|
|
|
| Re: Make_array() and using arrays as subscripts [message #46851 is a reply to message #46850] |
Tue, 10 January 2006 00:13  |
peter.albert@gmx.de
Messages: 108
Registered: July 2005
|
Senior Member |
|
|
Hi Sheldon,
well, reading the title of this posting I don't see where Make_array()
comes into your question, but anyway ... :-)
Like said already, WEHRE always returns a 1-dimensional vector, so here
you loose your dimensions in my_array. If you can be sure that
good_indices always is a factor of 3 (i.e., if a line contains '9999',
all 3 elements are equal to ''9999'), then you can use a simple
my_array = reform(my_array, 3, n_elements(good_indices)/3)
to get things right again.
However, if this is not the case, I can suggest yet another approach
without loops:
good_indices = where(strpos(strjoin(my_array), '9999')) eq -1)
Mind that in this case good_indices is only "valid" on the second
dimension, so at maximum it can have 344 elements. I am just
concatenating each line into one string each and look for the occurence
of '9999' within the string. If this substring does not occur at all,
it's a good line.
Then you can use
my_array = my_array[*, good_indices]
to shrink your array to the good lines without the nodata value.
Cheers,
Peter
|
|
|
|
| Re: Make_array() and using arrays as subscripts [message #46856 is a reply to message #46851] |
Mon, 09 January 2006 13:24 |
b_gom
Messages: 105
Registered: April 2003
|
Senior Member |
|
|
I think you're looking for the ARRAY_INDICES function (This will
convert the 1-D indices returned by WHERE into a multidimensional
array), but it would be useful if you gave a more specific example of
your problem. For the details you gave, If you have a (3,344) array and
you only need the first 120 elements, then just use REFORM to change
my_array[good_indices] into a (3,40) array.
But, I think you have a more general problem. The result of your WHERE
call is not guaranteed to always fit into an array of the same
dimensionality as your original array, so
you would have to figure out what to do with 'empty' elements in your
final array.
IDL> x=findgen(3,4,5)
IDL> inds=where(x gt 30)
IDL> inds2=array_indices(x,inds)
IDL> help,inds2
INDS2 LONG = Array[3, 29]
IDL> xind=inds2[0,*]
IDL> yind=inds2[1,*]
IDL> zind=inds2[2,*]
IDL> print,n_elements(uniq(xind,sort(xind)))
3
IDL> print,n_elements(uniq(yind,sort(yind)))
4
IDL> print,n_elements(uniq(zind,sort(zind)))
3
So, we have 29 matches, but if we wanted the result in a 3-D array, we
would have 3*4*3-29=7 empty elements!
Brad
Sheldon wrote:
> Hi everyone,
>
> I have a problem with the function MAKE_ARRAY() that I hope someone can
> help me with.
> I created a string array with the dimension (3, 344) and performed some
> basic assignment operations.
> Now I know that when I am finished with the operations I will perform
> on the array, the length 344 will be too long. Let's say that only the
> first 120 elements will be needed. Now I assigned that entire array the
> string '9999' and then when I am ready to reduce the size of the array
> I used the WHERE() function to find the indices where each element is
> not equal to '9999'. All went well until I form this array-as-subscript
> operation:
>
> my_array = my_array[good_indices]
>
> The dimensions all disappear and all I get is an array with the
> elements equalling the number of elements of good_indices, i.e.,
>
> print, size(my_array,/dimensions)
>
> IDL> 120
>
> Now I have used this type of array-as-subscript before on other types
> of array and I have never had this problem before. Does anyone know why
> this happens or how can I cut my_array without losing my 3 dimensions?
>
> Sincerely,
> Sheldon
|
|
|
|
| Re: Make_array() and using arrays as subscripts [message #46857 is a reply to message #46856] |
Mon, 09 January 2006 13:19 |
btt
Messages: 345
Registered: December 2000
|
Senior Member |
|
|
Sheldon wrote:
> Hi everyone,
>
> I have a problem with the function MAKE_ARRAY() that I hope someone can
> help me with.
> I created a string array with the dimension (3, 344) and performed some
> basic assignment operations.
> Now I know that when I am finished with the operations I will perform
> on the array, the length 344 will be too long. Let's say that only the
> first 120 elements will be needed. Now I assigned that entire array the
> string '9999' and then when I am ready to reduce the size of the array
> I used the WHERE() function to find the indices where each element is
> not equal to '9999'. All went well until I form this array-as-subscript
> operation:
>
> my_array = my_array[good_indices]
>
> The dimensions all disappear and all I get is an array with the
> elements equalling the number of elements of good_indices, i.e.,
>
> print, size(my_array,/dimensions)
>
> IDL> 120
>
> Now I have used this type of array-as-subscript before on other types
> of array and I have never had this problem before. Does anyone know why
> this happens or how can I cut my_array without losing my 3 dimensions?
>
Hi,
You will want to keep tabs on the dimension(s) that WHERE operates since
a call to WHERE operates on the entire array unless restricted. For the
sake of simplicity you might try a loop. You haven't said if the 99999
flag could be in any one of the columns. If true then you must check
each column for the 99999 flag.
dims = SIZE(my_array, /DIM)
flag = MAKE_ARRAY(dims[1], VALUE = 0B)
For i = 0L, dims[0]-1 Do Begin
A = WHERE(my_array[i,*] NE 9999, nA) ;check the ith column
if nA GT 0 then flag[A] = flag[A] + 1B ;increment the ok values
EndFor
;any flag LT 3 must have had a 99999 somewhere
A = WHERE(flag EQ 3B, nA)
if nA GT 0 then my_array = my_array[*,A]
Hope that helps,
Ben
PS In about 2 minutes you'll hear a chorus of other (better) methods
from people more knowledgable than I about these things.
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
