Question

I’m writing a regexp to pick out punctuation in strings and I’m getting some behavior I don’t expect:

ix = regexp('FGFR4','[~!@#$%^&*()-=+{}\|;:''",<.>/?\[]')
ix =

     [5]

ix = regexp('FGFR4','[~!@#$%^&*()-+{}\|;:''",<.>/?\[]') %note, the '=' is gone
ix =

     []

So it appears that ‘=’ is matching the number 4. What I expect is it only to match the ‘=’ sign as so:

ix = regexp('FOO=SPAM','[~!@#$%^&*()-=+{}\|;:''",<.>/?\[]')
ix =

     [4]

What’s going on here?

Answer 1

The problem is not the = but the - in front of it. It creates a range of all characters from ) to = (in ASCII order). The reason why this is not a problem if you remove the = is that + comes before 4 in ASCII order, so the range does not include the 4 (in fact it only includes ), * and +, and since you have * anyway, this would have never mattered.

Three solutions:

1. escape the hyphen:

   [~!@#$%^&*()\-=+{}\|;:''",<.>/?\[]
   

2. or put it at the end of the character class:

   [~!@#$%^&*()=+{}\|;:''",<.>/?\[-]
   

3. unless you want to make sure that you use exactly this set of characters… so if you would be alright with matching anything that is not a space, a letter or a digit (or an underscore), you could just as well use this:

   [^\w\s]
   

   Matches any non-underscore, non-letter, non-digit, non-whitespace character.