Legacy implementation of strtrim prior to R2016a

It looks like strtrim changed in R2016a to not trim significant whitespace characters. However, I have a need for a legacy implementation. Is it possible to provide one that works in R2016a, but also trims significant whitespace characters?

Answer 1

[Guillaume]

regexp(yourstring, '(?<=^\W*)\w.*?(?=\W*$)', 'match', 'once')

should do the trick

[sst]

I had tried using regular expressions myself, but the result is well over 10 times slower than the built-in strtrim function. I had forgotten to mention that I had a speed constraint.

[Guillaume]

Then, how about?

notspace = ~isspace(yourstring);
trimmed = yourstring(cumsum(notspace) & cumsum(notspace, 'reverse'))

or

notspace = ~isspace(yourstring);
trimmed = yourstring(find(notspace, 1):find(notspace, 1, 'last'))

[sst]

The first implementation using regexp turned out to be the fastest alternative approach, though still well over ten times slower than the built-in strtrim. I was really hoping TMW would have provided a 'legacy' option for strtrim, like they did for functions like ismember, union, and intersect :)

[Guillaume]

First of all, I made a mistake in the regex, it should have been \s instead of \W and \S instead of \w (yes upper and lower case should be swapped as well), so:

regexp(yourstring, '(?<=^\s*)\S.*?(?=\s*$)', 'match', 'once')

This will not speed up the regex, the slow down is probably caused by the non-greedy all capturing star .*? followed by the greedy capturing \s*, which causes a lot of backtracking.

However, this regex will do the same and should be significantly faster:

regexp(yourstring, '(?<=^\s*)\S.*\S(?=\s*$)', 'match', 'once')

As for a legacy option for strtrim, this is something you would have to ask Mathworks through a service request.