Return to Answer

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n//g; s/\n/<!!!>/g'

To support multi-line matching (as per your edit), with recent versions of GNU sed, you can use the first one with -z option (assuming the file doesn't contain NUL characters) or use something like:

sed ':1
     $!{
       N
       b1
     }
     s/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

AST-OPEN sed (not commonly found in the wild) supports augmented regexps that support a negation and conjunction operators. There, you can do:

sed -A ':1
       $!{
         N
         b1
       }
       s/===(.*&(.*\<\!\!\!\>.*)!)\<\!\!\!\>//g'

However, that regexp is quite expensive, so will probably not be usable for large inputs.

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n//g; s/\n/<!!!>/g'

To support multi-line matching (as per your edit), with recent versions of GNU sed, you can use the first one with -z option (assuming the file doesn't contain NUL characters) or use something like:

sed ':1
     $!{
       N
       b1
     }
     s/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n//g; s/\n/<!!!>/g'

To support multi-line matching (as per your edit), with recent versions of GNU sed, you can use the first one with -z option (assuming the file doesn't contain NUL characters).

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n//g; s/\n/<!!!>/g'

To support multi-line matching (as per your edit), with recent versions of GNU sed, you can use the first one with -z option (assuming the file doesn't contain NUL characters) or use something like:

sed ':1
     $!{
       N
       b1
     }
     s/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

AST-OPEN sed (not commonly found in the wild) supports augmented regexps that support a negation and conjunction operators. There, you can do:

sed -A ':1
       $!{
         N
         b1
       }
       s/===(.*&(.*\<\!\!\!\>.*)!)\<\!\!\!\>//g'

However, that regexp is quite expensive, so will probably not be usable for large inputs.

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|/whatever/g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n/whatever/g; s/\n/<!!!>/g'

With sed which doesn't support non-greedy *, you'd need to use tricks like:

sed 's/_/_u/g;s/|/_p/g;s/<!!!>/|/g
     s/===[^|]*|//g
     s/|/<!!!>/g;s/_p/|/g;s/_u/_/g'

Or with some sed implementations:

sed 's/<!!!>/\
/g; s/===[^\n]*\n//g; s/\n/<!!!>/g'

To support multi-line matching (as per your edit), with recent versions of GNU sed, you can use the first one with -z option (assuming the file doesn't contain NUL characters).