Shell Programming and Scripting

Hello all! Please help me with the following complex regexp which works in egrep:

But silently does not work in sed:

or

Just can't figure out what's wrong...

Try:

Code:

sed 's|"http://ccc\.bbb\.com/documents/0000/[0-9]\{4\}/[^.][^.]*\.[a-z]\{3\}[^"]*"|aaa|g' infile

It works!!! Thank you!

Could you give a sample of your input? (I adjusted my post, because you changed the name to ccc.bbb.com later BTW)

Code:

sed 's/"http:\/\/ccc\.bbb\.com\/documents\/0000\/[0-9][0-9][0-9][0-9]\/[^.][^.]*\.[a-z][a-z][a-z][^"]*"/aaa/g' my-file.html

tyler_durden

I suppose your problem are some unescaped grouping constructs:

Your regex:

Code:

"/\"http:\/\/ccc\.bbb\.com\/documents\/0000\/[0-9]{4}\/([^\.]+\.[a-z]{3})[^\"] \"/

probably correct:

Code:

"/\"http:\/\/ccc\.bbb\.com\/documents\/0000\/[0-9]\{4\}\/\([^\.]+\.[a-z]\{3\}\)[^\"] \"/

If you want to repeat the previous regex part several times you have to use "regex\{<n>\}", not "regex{<n>}" and grouping has to be done with "\(<regex>\)", not "(<regex>)".

Furthermore, the "+" is not part of POSIX regexps, but some (usually GNU/Linux-) utilities understand it as "one or more", similar to "*" meaning "zero or more". I am not entirely sure how you meant it, but you would be probably better off making this clear by either escaping the "+" our using "<regex><regex>*" instead of "<regex>+" if you want to match one or more instances of <regex>.

I hope this helps.

bakunin

Thank you all guys. You are really fast.