Duplicate selected words with sed and replace it

Question

I have to modify a document containing xml data; one modification I couldn't set up with sed.

I have to modify the following expressions from:

<Bild href_fmt="Bilder/Bildschirmfoto%202012-06_fmt.jpeg" 
href="file:///user/folder/Desktop/Bildschirmfoto%202012-06-18%20um%2013.17.45.png"></Bild>

to:

<a href="http://test.server.at/media/somefolder/Bildschirmfoto%202012-06_fmt.jpeg" 
target="_blank"><img src="http://test.server.at/media/somefolder/Bildschirmfoto%202012-06_fmt.jpeg" 
alt="" width="350" height="" /></a>

My problem: I couldn't duplicate the Bildschirmfoto%202012-06_fmt.jpeg to generate the correct new entry. The HTML tag is somewhere in the line, therefore I can't duplicate the whole line...

How can I do this with sed or other Linux commands?

Answer 1

sed 's@.*href_fmt="[^/]*/\([^"]*\)".*@<a href="http://test.server.at/media/somefolder/\1target="_blank"><img src="http://test.server.at/media/somefolder/\1" alt="" width="350" height="" /></a>@'

We need to extract image name which is inside href_fmt="Bilder/Bildschirmfoto%202012-06_fmt.jpeg" . therefore first of all we'll write pattern for this :

.*href_fmt="[^/]*/\([^"]*\)".*

where .* matches anything before href_fmt="
href_fmt=" matches href_fmt=" =) [^/]*/ matches dir name Bilder/ \([^"]*\) matches our image name. \(\) is used to make group and use it as \1 in the future ".* matches the rest of the line and now we will replace everything in the line accordinog to next expression:

<a href="http://test.server.at/media/somefolder/\1target="_blank"><img src="http://test.server.at/media/somefolder/\1" alt="" width="350" height="" /></a>

where \1 equals to image name as mentioned above.

Hope now it is clear how does it work.