I have a working perl regex using grep. I am trying to understand how it works.

Here is the command command.

grep -oP '(?<=location>)[^<]+' testFile1.xml

Here are the contents of testFile1.xml


And this is the result


I am trying to understand the regex, i.e. this part (?<=location>)[^<]+


1 Answer 1


(?<=...) is a look-behind PCRE operator. By itself, it doesn't match anything but acts as a condition (that what's on the left matches ...).

(?<=X)Y matches Y provided that what's on the left matches X. In blahYfooXYbar, that matches the second Y, the X is not part of what is being matched. The (?<=X) itself matches the zero-width (imaginary) spot just before that Y. Here illustrated:

$ echo X-RAY THE FOX | perl -lpe 's/(?<=X)/<there>/g'
X<there>-RAY THE FOX<there>

Because with -o, grep only prints the matched portion, that's a way to make it print what's after the location> (here what matches [^>]+: one or more (+) non-< characters ([^>]) so everything up to (but not included) the next < character or the end of the line provided it's not empty).

Another approach is to use \K (in newer versions of PCRE) to reset the start of the matched portion:

grep -Po 'location>\K[^>]+'

Note that -P and -o are GNU extensions. With recent versions (8.11 or over) of pcregrep (another grep implementation that uses PCRE), you can also do:

pcregrep -o1 'location>([^>]+)'

(-o1 prints what's captured by the 1st (here unique) (...))

  • I under stand the look-behind part. I am still not sure what does "[^<]+" mean? does it mean anything starts with "<" symbol?
    – Sas
    Oct 20, 2016 at 16:00
  • also I tried the pcregrep and it's says "pcregrep: Unknown option letter '1' in "-o1"
    – Sas
    Oct 20, 2016 at 16:04
  • @Sas, See if my latest edit make it clearer. Oct 20, 2016 at 16:04
  • @Sas, -o<n> was added in 8.11 (Dec 2010). You probably have an older version. (I do mention recent versions in my answer though I hadn't realised it was almost 6 years ago it was added. Time flies...). Oct 20, 2016 at 16:11
  • 1
    @Sas If you don't understand basics like [^abc], then I recommend reading through the docs once or twice. Especially for PCRE, this is time well spent, as almost every programming language that implements regexes will use syntax & semantics similar to this. (But, keep in mind that POSIX/shell regexes are fairly different; those are the two most common variants one needs to use.)
    – jpaugh
    Oct 20, 2016 at 21:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.