If you've used sed for a while, you've probably encountered the
problem that all your (group) matches are greedy. This is easy to
illustrate with an example:
Say you have this text, where you want to convert the java classes
surrounded with @-signs into HTML <code/> elements:
The @HashMap@ and @ArrayList@ classes are great!
To convert this into HTML, you'd typically create groups and wrap
these two inside <code> elements:
echo 'The @HashMap@ and @ArrayList@ classes are great!' | sed 's#@\(.*\)@#<code>\1</code>#g'
However, sed will always do a greedy search, matching the longest
possible match, not quite yielding the result you want:
The <code>HashMap@ and @ArrayList</code> classes are great!
The solution
What you need to do, is to match against everything but the closing
delimiter, in our case, the '@'. This is done with the character group
notation and the negator ^ in front of the closing delimiter:
@\([^@]*\)@
Now, we get what we want with both java classes wrapped in HTML
<code> elements:
echo 'The @HashMap@ and @ArrayList@ classes are great!' | sed 's#@\([^@]*\)@#<code>\1</code>#g'
The <code>HashMap</code> and <code>ArrayList</code> classes are great!
There, another sed mystery solved. Yeah!
