UNIX for Dummies Questions & Answers

Hi,

I'm curious about how to do a very simple thing with regular expressions that I'm unable to figure out.

If I want to find out if a string contains 'a' AND 'b' AND 'c' it can be very easily done with grep:

Code:

echo $STRING|grep a|grep b|grep c

but, how would you do that in a single regexp?

A possible solution would be:

Code:

/a.*b.*c|a.*c.*b|b.*c.*a|b.*a.*c|c.*b.*a|c.*a.*b/

but it's so ugly it's nasty!! (imagine if instead of 3 chars we have 10!)

any gurus out there know how to do this?

cheers

Code:

awk '/a/&&/b/&&/c/'

Scrutinizer: that's very cool, it's better than chaining up grep's But I would like to do that in a single regexp i.e. without the use of shell tools like grep, awk, sed...

Well, awk has not one regex but one command, which is nicer formatted, and you can use sed, which is faster, generally and because it doe not evaluate regex if the line is done=dead, and the list of patterns can be of any length, easily viewed:

Code:

sed '
  /a/!d
  /b/!d
  /c/!d
 '

---------- Post updated at 10:45 AM ---------- Previous update was at 10:42 AM ----------

If you want just regex not commands, you are probably out of luck. The searches are too unrelated for one regex. What context do you want to use it in, if not a command?

If you want to use return codes (like grep -q)

Code:

awk '/a/&&/b/&&/c/{f=1;exit}END{if(!f){exit 1}}'

For example in any programming language that supports pcre: C, perl, python, Ruby... of course every programming language has other ways to check this, for example in python:

Code:

>>> s = "axbxc"
>>> 'a' in s and 'b' in s and 'c' in s
True

I just want to know if it's possible to do that in a single regular expression, just out of curiousity and simply to get a better understanding of regexps.

I tried to do it like this:

Code:

/([abc]).*([^\1]).*[^\2]/

But of course that doesn't work because [^\1] matches *all* the characters except the character that matched in the first parenthesis set... I think this should be done with some kind of backtracking.

And BTW I'm sure that it can be done with regexps! I mean, if you can test if a number is a primer number with regular expressions, I refuse to believe this simple thing can't be done

Well, if you dislike but must apply the three regex in sequence, you might try this simple heurism: Put the three regex in as list, and apply them in the current order; if one misses, removing a line from contention, move it to the top of the list if not already, sliding the others down. For instance, consider c e q as three regex. The q will reject more lines than c, usually, and the e less, but by letting the best rejecters float to the top, you save a lot of second and third regex searches.

I have a name for this, but it is not politically correct, something about how a dictator selects a military commander -- death at first failure. It came to me one day as a text editor took very long to find instances of 'equal': it did a character scan and for every first character, it stopped and did a sting compare, tragically missing the filtering power of q. If I searched for 'qual', it was quick (Borland Sprint on I386 dos emulation under UNIX SVR3).