Sed: regular expression match lines without <!--

Question

I have a sed command to comment out xml commands

sed 's/^\([ \t]*\)\(.*[0-9a-zA-Z<].*\)$/\1<!-- Security: \2 -->/' web.xml 

Takes:

<a>

  <!-- Comment -->
  <b>
    bla
  </b>

</a>

Produces:

<!-- Security: <a> -->

  <!-- Security: <!-- Comment --> -->    // NOTE: there are two end comments.  
  <!-- Security: <b> -->
    <!-- Security: bla -->
  <!-- Security: </b> -->

<!-- Security: </a> -->

Ideally I would like to not use my sed script to comment things that are already commented.

Ie:

<!-- Security: <a> -->

  <!-- Comment -->
  <!-- Security: <b> -->
    <!-- Security: bla -->
  <!-- Security: </b> -->

<!-- Security: </a> -->

I could do something like this:

sed 's/^\([ \t]*\)\(.*[0-9a-zA-Z<].*\)$/\1<!-- Security: \2 -->/' web.xml
sed 's/^[ \t]*<!-- Security: \(<!--.*-->\) -->/\1/' web.xml

but I think a one liner is cleaner (?)

This is pretty similar: http://stackoverflow.com/questions/436850/matching-a-line-that-doesnt-contain-specific-text-with-regular-expressions

Answer 1

This seems to work.

^([ \t]*)(?!<!--)([0-9a-zA-Z<].*)$

You have to escape the parenteses and other stuff in order to actually make it work with sed.

Tested the expression in an online regular expression tester.

Unfortunately it seems that sed does not support lookahead (but ssed does).
maibe you should try awk or ssed.

Answer 2

you can use awk, If i get your requirement correct:

awk -F"," '!/<!--/&&NF{$0="<!-- Security: "$0"-->"}1' file

Answer 3

Ended up going with this:

startInsert="<!-- Security: "
endInsert=" -->"

whiteSpace="[ \t]*"
char="[0-9a-zA-Z^_)!-/:-@[-\{-~]"

sed "s/^\($whiteSpace\)\(.*$char.*\)$/\1$startInsert\2$endInsert/" $file 
sed "s/^\($whiteSpace\)$startInsert\(<!--.*\-->\)$endInsert/\1\2/" $file 

Not as graceful as I was hoping for but works like a charm.