And yet, I still can't quite grasp at how I can solve my for loop problem. I think it might involve the use of modulo (MOD), but I'm not sure how. My question is, how can you grab the indices (i,j,k) of a 3D array in real space, and throw them into basically 3 1D arrays that is just a list of all the cells in the "proper" order (column-major).
index=0L
for k=0,zcells do begin
for j=0,ycells do begin
for i = 0,xcells do begin
coord(index,0)=i
coord(index,1)=j
coord(index,2)=k
index=index+1
endfor
endfor
endfor
end
This is a really simple version of a complex problem I have. I have sets of different size boxes from an AMR MHD code, and I need to keep track of their indices, but I just want a list of all of the cells, not to drag around a bunch of smaller arrays or try to concatenate them into one giant sparse array (waste of space). I'm certain that someone must have had this problem before, but I can't find any other suggestions on this forum.
> And yet, I still can't quite grasp at how I can solve my for loop problem. I think it might involve the use of modulo (MOD), but I'm not sure how. My question is, how can you grab the indices (i,j,k) of a 3D array in real space, and throw them into basically 3 1D arrays that is just a list of all the cells in the "proper" order (column-major).
> Here is an example of what I mean:
> pro testreader
> xcells=15
> ycells=10
> zcells=20
> ncells=xcells*ycells*zcells
> data=dindgen(xcells,ycells,zcells)
> coord=intarr(ncells,3)
> index=0L
> for k=0,zcells do begin
> for j=0,ycells do begin
> for i = 0,xcells do begin
> coord(index,0)=i
> coord(index,1)=j
> coord(index,2)=k
> index=index+1
> endfor
> endfor
> endfor
> end
> This is a really simple version of a complex problem I have. I have sets of different size boxes from an AMR MHD code, and I need to keep track of their indices, but I just want a list of all of the cells, not to drag around a bunch of smaller arrays or try to concatenate them into one giant sparse array (waste of space). I'm certain that someone must have had this problem before, but I can't find any other suggestions on this forum.
If I understand well your problem, a solution might be:
On Friday, August 24, 2012 5:32:05 AM UTC-4, alx wrote:
> Le jeudi 23 août 2012 22:58:48 UTC+2, simu...@gmail.com a écrit :
> If I understand well your problem, a solution might be:
> And yet, I still can't quite grasp at how I can solve my for loop problem. I think it might involve the use of modulo (MOD), but I'm not sure how. My question is, how can you grab the indices (i,j,k) of a 3D array in real space, and throw them into basically 3 1D arrays that is just a list of all the cells in the "proper" order (column-major).
There is a perfectly good builtin function in IDL to do this: ARRAY_INDICES.
IDL> nx = 15L
IDL> ny = 10L
IDL> nz = 20L
IDL> data = randomn(seed, nx,ny,nz)
IDL> ai = array_indices(data, lindgen(nx*ny*nz))
IDL> IDL> help, ai
AI LONG = Array[3, 3000]
If you really need dimensions [3000,3], you can add
IDL> ai = transpose(temporary(ai))
> Just for the record, that particular post was made almost entirely in jest: sometimes a for loop is indeed what you want.
> JD
*Almost* entirely you say, but in my case entirely accurate. I reduced the runtime of my code from 10 minutes to 2 minutes by using the above arrays to find my indices instead of for loops, which seems significant to me, considering I need to run it over and over again. I was getting really bored.
In general I find that it's easy to loose focus on a problem if your runtime is approximately between 4 min - 2 hours. You feel like it's too short of a time to switch focus to another project and be productive at all. So, yay for code optimization.
In response to Yngvar:
I also had the problem that I had to add a constant value to each i,j,k specified by another array xBounds that gives the lower left corner of each grid. Since I was basically concatenating all the data from each grid, I could figure out how to add this lower left corner value to each data point before the reform.
