Question

I have to write a program in Matlab, the performance is quite important since it handles a lot of files.

I have to find the last file extension of a filename and split it there.

e.g. file.name.tar.gz should be split into file.name.tar and gz

I identified 3 methods to do this but I don’t know which is the fastest.

1.

javaFilename = java.lang.String(fileName);
lastDot = javaFilename.lastIndexOf('.');
name = char(javaFilename.substring(0, lastDot-1));    
ext = char(javaFilename.substring(lastDot+1));

2.

dots = findstr(fileName, '.');
lastDot = dots(length(dots));
name = fileName(1:lastDot-1);
ext = fileName(lastDot+1:end);    

3.

[name tempExt] = strtok(fileName, '.');
while tempExt
    [temp2 tempExt] = strtok(fileName, '.');
    if tempExt
       name = strcat(name, '.', temp2);
    end
end
ext = temp2(2:end);

I think the third one is the worst. But how about the other methods?
Can you come up with anything that is faster than the methods described by me?

Answer 1

You can test these by doing something like:

tic;
fileName = 'testFileName.ext';
for i=1:100000
  <find last dot>
end
toc;

I benchmarked yours and found the second one was much faster.

My solution would be

for index=length(fileName):-1:1
  if (fileName(index) == '.')
    break;
  end
end

The times I got were:

Java Solution: 23 seconds

findstr Solution: 0.4 seconds

strtok Solution: (didn’t terminate; maybe I copied it wrong)

reverse for-loop Solution: 0.01 seconds

The advantage of my solution over findstr and strtok is that:
a) it only looks for one dot, rather than all of them
b) it starts from the end of the string (presumably most file extensions will be 2-4 letters long).