| Re: Incremental variable names? [message #37872] |
Wed, 04 February 2004 01:34 |
Chris Lee
Messages: 101
Registered: August 2003
|
Senior Member |
|
|
In article <BC4475B7.18FD5%greenberg@ucdavis.edu>, "Jonathan Greenberg"
<greenberg@ucdavis.edu> wrote:
> I'm trying to optimize this algorithm for use with large images.
> ENVI/IDL work well if you extract a single line of an image at a time
> (line = y axis, by the way). For a circular mask, however, I'd need to
> extract at least as many lines as the maximum circular diameter (e.g. If
> my circle is 3 pixels in radius, I need to pull out 7 lines of data to
> figure out the values along the ring).
> It occurred to me (and tell me if I'm wrong), that while I CAN extract
> the line I'm working on +/- 3 lines from the image each time, this seems
> like an awful waste of I/O, when in reality the next line will use all
> but one of the previous extracted rows (e.g. I just need to read the
> next line in, and delete the current top line to free up memory).
> Creating arrays of the lines would require rewriting the arrays each
> time (e.g. Arraynew = (arrayold(2:7),newline) -- I know this isn't idl
> code, but you get the idea) which seems slower than just having 7 named
> variables (line1,line2,...line7), doing my calculation, then deleting
> line1, and adding line8 and repeating.
> No? Is there a better way of "preserving" the data I've already read
> from the HD than using incremental variable names? --j
> On 2/2/04 1:45 PM, in article
If I understand your method correctly.....Create a 2D array with the
dimension of (radius, line length). Each time you read in a new line..
array=fltarr(radius, line_length)
input=fltarr(line_length)
openr, file, "filename", /get_lun
counter=0L
cmax=2*radius+1
while(not eof(file)) do begin
readu,file, input
array[counter mod cmax,0]=input
array=shift(array, 1,0) ;;;;This could be wrong, try -1,0...
result=do_stuff(array)
counter=counter+1
endwhile
free_lun, file
; I think that would work..
Of course, if I understand this whole ring thing, your doing this with
every point on your image? If you are, the function DO_STUFF above (your
semi-variance code) had better be vectorised along the line_length.
Chris.
|
|
|
|
| Re: Incremental variable names? [message #37888 is a reply to message #37872] |
Mon, 02 February 2004 21:46  |
Jonathan Greenberg
Messages: 91
Registered: November 2002
|
Member |
|
|
I should get in the practice of not making "examples" of what I want to do
and just come out and say it:
OK, from a previous email we were discussing extracting "rings" of data from
an image to calculate a semivariance for each pixel in the image. The
previous suggestions helped a lot in optimizing how to extract the data from
the array (use where + an array of index values of the ring locations offset
from the center pixel).
I'm trying to optimize this algorithm for use with large images. ENVI/IDL
work well if you extract a single line of an image at a time (line = y axis,
by the way). For a circular mask, however, I'd need to extract at least as
many lines as the maximum circular diameter (e.g. If my circle is 3 pixels
in radius, I need to pull out 7 lines of data to figure out the values along
the ring).
It occurred to me (and tell me if I'm wrong), that while I CAN extract the
line I'm working on +/- 3 lines from the image each time, this seems like an
awful waste of I/O, when in reality the next line will use all but one of
the previous extracted rows (e.g. I just need to read the next line in, and
delete the current top line to free up memory). Creating arrays of the
lines would require rewriting the arrays each time (e.g. Arraynew =
(arrayold(2:7),newline) -- I know this isn't idl code, but you get the idea)
which seems slower than just having 7 named variables
(line1,line2,...line7), doing my calculation, then deleting line1, and
adding line8 and repeating.
No? Is there a better way of "preserving" the data I've already read from
the HD than using incremental variable names?
--j
On 2/2/04 1:45 PM, in article
df160b8f.0402021345.25b58a8@posting.google.com, "Big Bird"
<condor@biosys.net> wrote:
> Jonathan Greenberg <greenberg@ucdavis.edu> wrote in message
> news:<BC430FE2.18E9C%greenberg@ucdavis.edu>...
>> Have a question about variable names. Is there any way to create and then
>> call variable names that have some incremental value in them, e.g.:
>>
>> t0
>> t1
>> t2
>>
>> Etc... I want to be able to first create and assign a value to tn where n =
>> 0 to whatever, and then be able to subsequently call that variable using a
>> loop.
>>
>
>
> OK, now tell us what you REALLY want to do. ;-)
>
> As it is written up there, there doesn't exactly seem to be a reason
> not to use an array for your variables -- instead of tn you write t[n]
> and that's pretty much it.
>
> I assume that there's a reason for you not to do things that way.
>
> Now while there is certainly the possibility to use EXECUTE, I'd be
> surprised if tehre weren't some cleverer way to do whatever you're
> trying to do and I'm sure a lot of people would be happy to tell you
> about that way if we had a little more information at hand ...
|
|
|
|
| Re: Incremental variable names? [message #37891 is a reply to message #37888] |
Mon, 02 February 2004 13:45 |
condor
Messages: 35
Registered: January 2002
|
Member |
|
|
Jonathan Greenberg <greenberg@ucdavis.edu> wrote in message news:<BC430FE2.18E9C%greenberg@ucdavis.edu>...
> Have a question about variable names. Is there any way to create and then
> call variable names that have some incremental value in them, e.g.:
>
> t0
> t1
> t2
>
> Etc... I want to be able to first create and assign a value to tn where n =
> 0 to whatever, and then be able to subsequently call that variable using a
> loop.
>
OK, now tell us what you REALLY want to do. ;-)
As it is written up there, there doesn't exactly seem to be a reason
not to use an array for your variables -- instead of tn you write t[n]
and that's pretty much it.
I assume that there's a reason for you not to do things that way.
Now while there is certainly the possibility to use EXECUTE, I'd be
surprised if tehre weren't some cleverer way to do whatever you're
trying to do and I'm sure a lot of people would be happy to tell you
about that way if we had a little more information at hand ...
|
|
|
|
| Re: Incremental variable names? [message #37903 is a reply to message #37891] |
Mon, 02 February 2004 08:52 |
MKatz843
Messages: 98
Registered: March 2002
|
Member |
|
|
Jonathan, I apologize if this solution is way too obvious and you may
have something else in mind. Have you considered an array?
t = fltarr(10)
t(0)
t(1)
t(2)
If the values all need to be of different types, then pointers come to
the rescue.
t = ptrarr(10)
t(0) = ptr_new(0.5)
t(1) = ptr_new('hello')
t(2) = ptr_new(dblarr(100,400)
print, *t(0)
print, *t(1)
and so forth
Plus you can extend the length at will
t = [t, ptr_new('extension works')]
or remove elements, retaining only some.
t = t([1, 3, 4])
Is there a reason why this wouldn't work for you?
M. Katz.
Jonathan Greenberg <greenberg@ucdavis.edu> wrote in message news:<BC430FE2.18E9C%greenberg@ucdavis.edu>...
> Have a question about variable names. Is there any way to create and then
> call variable names that have some incremental value in them, e.g.:
>
> t0
> t1
> t2
>
> Etc... I want to be able to first create and assign a value to tn where n =
> 0 to whatever, and then be able to subsequently call that variable using a
> loop.
>
> --j
|
|
|
|
| Re: Incremental variable names? [message #37911 is a reply to message #37903] |
Mon, 02 February 2004 05:57 |
btt
Messages: 345
Registered: December 2000
|
Senior Member |
|
|
Jonathan Greenberg wrote:
> Have a question about variable names. Is there any way to create and then
> call variable names that have some incremental value in them, e.g.:
>
> t0
> t1
> t2
>
> Etc... I want to be able to first create and assign a value to tn where n =
> 0 to whatever, and then be able to subsequently call that variable using a
> loop.
Hello,
I'm a bit confused about why you to name the variable incrementally if
what you really want to access them incrementally. It seems to me that
you will not be able to decode the increment of the variable with ease
after you have created it. I would seriously consider using a linked
list (http://www.dfanning.com/documents/programs.html#LINKEDLIST) or
make a variable object that can be stored in a container (which is just
a linked list itself.) Then you simply access the variable in the
order the have been stored.
I was introduced to the latter method when Martin Schultz (where is he,
anyway?) first built a generic object system. I prefer the Martin's
method now, as I can do some spiffy things with the variables without
cluttering up my code.
Cheers,
Ben
|
|
|
|
| Re: Incremental variable names? [message #37913 is a reply to message #37911] |
Sun, 01 February 2004 20:57 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
David Fanning writes:
> Here is how I do this in one of my programs. In this
> case, the files are named image01.tif, image02.tif...
> image53.tif.
>
> FOR j=0, 53 DO BEGIN
> filename = 'image' + String(j, Format='(I2.2)') + '.tif'
> Write_Tiff, filename, Reverse(Reform(head[*,*,j]),2)
> ENDFOR
Whoops! Sorry, I'm still thinking about Janet Jackson and
the Super Bowl.
I'll try to answer the real question next time. :-(
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting, Inc.
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
|
|
|
|
| Re: Incremental variable names? [message #37914 is a reply to message #37913] |
Sun, 01 February 2004 20:55 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Jonathan Greenberg writes:
> Have a question about variable names. Is there any way to create and then
> call variable names that have some incremental value in them, e.g.:
>
> t0
> t1
> t2
>
> Etc... I want to be able to first create and assign a value to tn where n =
> 0 to whatever, and then be able to subsequently call that variable using a
> loop.
Here is how I do this in one of my programs. In this
case, the files are named image01.tif, image02.tif...
image53.tif.
FOR j=0, 53 DO BEGIN
filename = 'image' + String(j, Format='(I2.2)') + '.tif'
Write_Tiff, filename, Reverse(Reform(head[*,*,j]),2)
ENDFOR
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting, Inc.
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
|
|
|
|
| Re: Incremental variable names? [message #37915 is a reply to message #37914] |
Sun, 01 February 2004 20:51 |
mmeron
Messages: 44
Registered: October 2003
|
Member |
|
|
In article <BC430FE2.18E9C%greenberg@ucdavis.edu>, Jonathan Greenberg <greenberg@ucdavis.edu> writes:
> Have a question about variable names. Is there any way to create and then
> call variable names that have some incremental value in them, e.g.:
>
> t0
> t1
> t2
>
> Etc... I want to be able to first create and assign a value to tn where n =
> 0 to whatever, and then be able to subsequently call that variable using a
> loop.
>
Yes, using EXECUTE. You can find some examples in my library (MIDL,
on the RSI users contribution page). Look in the routines HOW_MANY
and OUTPUT, for example.
Mati Meron | "When you argue with a fool,
meron@cars.uchicago.edu | chances are he is doing just the same"
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
