Thread Subject: REGEXP: no AND in sight?

.'Logical operations' page of REGEXP documentation lists OR, but not AND! Is it true that one cannot combine patterns in AND (hence AND NOT) fashion in Matlab ? If not, can anybody suggest how I could code a (silly but simple) pattern like 'a or b or c, but not c'? [abc]..?

Thank you.

Dimitri Shvorob wrote:
>
>
> .'Logical operations' page of REGEXP documentation lists OR, but
> not AND! Is it true that one cannot combine patterns in AND (hence
> AND NOT) fashion in Matlab ? If not, can anybody suggest how I
> could code a (silly but simple) pattern like 'a or b or c, but not
> c'? [abc]..?
>
> Thank you.
  

Regular Expressions are not a Matlab invention but a well-known
standard for STRING-based (=expression) processing, existing in the
programming world for several decades. Being string-based, the
standard does not need the AND construct and therefore has one
(namely, '|') defined only for OR. For AND, there's simple
concatenation with wildcards. Here's an example:

str = 'your name is Dimitri Shvorob';
index = regexp(str,'Dimitri|Dima'); % => index=14, OR
index = regexp(str,'name.*Dimitri'); % => index=6, AND

Yair Altman
 <http://www.ymasoftware.com>

You can simulate AND in regular expressions using lookaround operators.

For example, to match consonant letters, you could list them
individually in a set:

>> regexp('this is text', '[bcdfghjklmnpqrstvwxyz]', 'match')

ans =

    't' 'h' 's' 's' 't' 'x' 't'



Which is clunky. It is easier to match not vowels, since there are less
of them:

>> regexp('this is text', '[^aeiou]', 'match')

ans =

    't' 'h' 's' ' ' 's' ' ' 't' 'x' 't'

But that doesn't work, since it includes characters that are not
letters. So use lookahead to ensure that the next character is a
letter, before checking to see that it isn't a vowel:

>> regexp('this is text', '(?=[a-z])[^aeiou]', 'match')

ans =

    't' 'h' 's' 's' 't' 'x' 't'

Hope that helps,

-=>J

Yair, thank you for help. I must admit I am not sure in what sense the second example
index = regexp(str,'name.*Dimitri');
implements AND.
I can check a string's correspondence to n > 1 patterns with something like
if regexp(string,pattern1) & regexp(string,pattern2) ...
The difficulty is with *selecting* text, as in
regexp(string,pattern,'match')
The task that I am working on is selecting all e-mail addresses, except, say, bill.gates@microsoft.com, from a text file. The command immediately above, supplied with a half-decent pattern, gives me a cell array of strings, containing the matches. I need to loop through it, and eliminate any occurrences of bill.gates@microsoft.com. No big deal, but it would be much nicer to skip them in the first place.

No need to loop. Here's a simple vectorized one-line filter (assume your positive matches are stored in cell array c):

c(~cellfun('isempty',regexpi(c,'bill.gates@microsoft.com')))=[];

Yair Altman
http://www.ymasoftware.com

:) Thank you very much.