using regexp with negative and positive floating numbers

data = 

      '0.00640869<'
      '0.00640869<'
      '-0.00598145'
      '-0.00598145'
      '-0.0010376<'
      '-0.0010376<'
      '-0.00604248'
      '-0.00213623'
      '0.00567627<'
      '-0.00219727'
      '0.00567627<'
% 11x1 cell

to remove the non-numeric character;

data_num = regexp(data,'[0-9]+.[0-9]+', 'match');

In this case, negative signs (-) are also removed. How can I modify this code to get - signs?

Answer 1

[Stephen23]

regexp(data,'[+-]?\d+\.?\d*', 'match');

Notes:

1. important: . does NOT match the period character, but in fact it matches any character. To only match the period, you need to escape it: \.
2. \d matches any digit, and is simpler than writing [0-9]
3. by making the decimal fraction optional my answer will also match integer values

Read the MATLAB documentation for more information:

https://www.mathworks.com/help/matlab/matlab_prog/regular-expressions.html

You might also like to try my FEX submission, which lets you experiment with regular expressions and see their output in real time (as you type):

https://www.mathworks.com/matlabcentral/fileexchange/48930-interactive-regular-expression-maker

[Walter Roberson]

The \.? should not have the ? because if the . is missing, that would be like \d+\d* in a row. Instead, if the . is optional, then

'[+-]?\d+(\.\d*)?'

[Stephen23]

Why would \d+\d* be a problem? My tests showed that it correctly matches any integer, exactly as I intended, and I don't see how it would cause any problems for this use case. Can you please explain.

[Walter Roberson]

It is redundant -- you already matched all the possible digits in a row with the \d+ and then you spin looking for optional digits.

[Stephen23]

@Walter Roberson: the tests I just ran showed that your code is actually slightly slower (1e5 iterations):

Elapsed time is 3.67537 seconds. % \.?\d*
Elapsed time is 3.91039 seconds. % (\.\d*)?

I have no idea how TMW implements this, or whether it would be stable between versions...

[Alberto Fernandez Osorio]

Works great to me! Thanks @Stephen23