How to do it more efficiently?

Hi,

In my script, this piece of code is taking about 70% of processing time.

The idea is to create uniform timestamp, it searches the time array and compare it with the Tsynced. if there is no similar entry, the other parameters to NaN.

Is there anyway to change the script below without using find function? Even better if no for-loop is required.

Code:

for ix = 1:length(time_synced)
  [r,~,~] = find(cCells(:,1)==time_synced(ix),1,'first');
    if isempty(r)
      curArr(ix,:) = NaN;
    else
      curArr(ix,:) = cCells(r,:);
    end
end

Thanks.

Answer 1

[Roger Stafford]

See if this is faster:

   [t,ic] = ismember(time_synced,cCells(:,1));
   curArr(t) = cCells(ic);
   curArr(~t) = NaN;

Note: This assumes that 'time_synced' and 'curArr' are vectors of the same length.

[Ilham Hardy]

Ah, yes. It is way faster than the previous find+for-loop method.

Thanks.

[Ilham Hardy]

A bit correction for other fellow readers,

   [t,ic] = ismember(time_synced,cCells(:,1));
   curArr(t,:) = cCells(ic,:);
   curArr(~t,:) = NaN;

PS: The curArr is preallocated.

Answer 2

[Andrei Bobrov]

[l0,ii] = ismember(time_synced,cCells(:,1));

out = nan(numel(time_synced),size(cCells,2));
out(l0,:) = cCells(ii(l0),:);

[Ilham Hardy]

Many thanks for the answer,