| Sum along diagonals [message #21460] |
Fri, 25 August 2000 00:00  |
mole6e23
Messages: 31
Registered: December 1998
|
Member |
|
|
Hi there..
I thought I'd ask this question, more out of curiosity as to a possible
result than for really needing to optimize it (this code is used
approximately three times for every blue moon). I'm always impressed with
some of the results from this group, most recently:
a=a[*,where(histogram(b,MIN=0,MAX=(size(a,/DIMENSIONS))[1]-1 ,BINSIZE=1) eq 0)]
I wouldn't have thought of compacting all that into one line!
Every once in a while (not often enough to make me worry about optimizing
too much), I want to take a not necessarily square matrix and get the sum
along the diagonals, such as the following, with the theoretical function
sum_diag:
IDL> blah = indgen( 4, 4 )
IDL> print, blah
0 1 2 3
4 5 6 7
8 9 10 11
12 13 14 15
IDL> print, sum_diag( blah )
0 5 15 30 30 25 15
which is the series [0, 4+1, 8+5+2, 12+9+6+3, ... ]
Of course, to be difficult, I'd like it to work for non-square matrices as well:
IDL> blah = indgen( 5, 3 )
IDL> print,blah
0 1 2 3 4
5 6 7 8 9
10 11 12 13 14
and the result would be the series [0, 5+1, 10+6+2, 11+7+3, ... ]
The best I've found is to count the diagonals, loop through those, and
then loop again through the possible row and column indices in that array.
It's pretty ugly, and very time consuming for a 512x512 array (most of
what I want). It just occured to me now that for the squre matrix, you
could take "blah + transpose( blah )" and now you only have to worry about
the lower half of the matrix (and the doubled diagonal elements), so it
might be a bit better, but still messy.
So if you're bored, I'd love to see a better solution than a bunch of
long, nested loops, even if just only for the square case (since that is
90% of the cases in my data)
Enjoy!
Todd
|
|
|
|
| Re: Sum along diagonals [message #21524 is a reply to message #21460] |
Mon, 28 August 2000 00:00  |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
colinr@toliman.uio.no (Colin Rosenthal) writes:
> On Fri, 25 Aug 2000 17:57:01 -0700,
> Todd Clements <mole6e23@hotmail.com> wrote:
>> craigmnet@cow.physics.wisc.edu wrote:
>>> How about this solution. It's not a one-liner, and it uses two loops,
>>> [snip]
>>
>> That works just fine! Once again, I am duly impressed. I'm sure it took me
>> longer to figure out your code than it did for you to write it! Absolutely
>> amazing. Having read this newsgroup for a while, I'm beginning to wonder
>> if there is an IDL question you can't answer.
>
> Can he write a Stein-Vidar emulator in IDL? Or maybe David could write a
> Craig emulator?
Actually Stein-Vidar is at Goddard where I work too [ in case people
didn't know. ] I made a promise long ago to meet him for lunch, but
never did. I guess now's my chance to collect data on that
emulator...
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
