[SOLVED] Replace a pattern range line by line with sed

Issue

I need to replace the beginning of each line of a pattern range sorted by ID, I am using sed but other languages are welcome!

This is the sample text with the objectives

...
==OPEN==
  data: blabla
  id: class1
  moredata: blabla
==CLOSE==

==OPEN==
  id: class2
  boringdata: blabla
==CLOSE==

...extra info

==OPEN==
  id: class8
  data: ...
==CLOSE==

...more info

==OPEN==
  data: ...
  boringdata: ...
  id: class10
==CLOSE==
...

If I were to comment out the pattern with id 8, the expected output would be:

...
==OPEN==
  data: blabla
  id: class1
  moredata: blabla
==CLOSE==

==OPEN==
  id: class2
  boringdata: blabla
==CLOSE==

...extra info

// ==OPEN==
//   id: class8
//   data: ...
// ==CLOSE==

...more info

==OPEN==
  data: ...
  boringdata: ...
  id: class10
==CLOSE==
...

The closest code I have gotten is this, but I have to rewrite the entire range and it is not affordable:

sed -e '/==OPEN==/ {:loop; N; /==CLOSE==/! b loop; /id: class8/ {s/.*/NEEDS REWRITE/}}' /example

If I tell it to rewrite the beginning (^), it rewrites only the first line of the range, I think it is because it considers the entire pattern as one line.

Solution

The closest code I have gotten is this, but I have to rewrite the entire range and it is not affordable:

sed -e '/==OPEN==/ {:loop; N; /==CLOSE==/! b loop; /id: class8/ {s/.*/NEEDS REWRITE/}}' /example

That's actually not bad.

If I tell it to rewrite the beginning (^), it rewrites only the first line of the range, I think it is because it considers the entire pattern as one line.

Yes, with POSIX sed, and with GNU sed by default, the ^ matches the beginning of the pattern space only. However, you can match the newlines themselves:

sed -e '/==OPEN==/ {:loop; N; /==CLOSE==/! b loop; /id: class8/ {s,\(^\|\n\),\1// ,g}}' \
  /example

Note in particular:

• The text being replaced is expressed as the group \(^\|\n\), meaning either the zero-length substring at the beginning of the pattern space or a newline, which is captured as a group.
• The matched text is echoed back into the replacement via \1. That has no visible effect when it is the ^ alternative that matches, but it avoids eliminating newlines when the other alternative matches.
• The comma (,) is used as pattern delimiter, so that the slashes (/) in the replacement text don't need to be escaped.
• The g flag is used to cause all appearances of the pattern to be replaced, not just the first.

---

If you are willing to rely on GNU extensions, then you can do it a bit more simply:

sed -e '/==OPEN==/ {:loop; N; /==CLOSE==/! b loop; /id: class8/ {s,^,// ,gm}}' \
  /example

With GNU sed, the m flag in the s command causes ^ to match both at the beginning of the pattern space and immediately after each newline within. This flag is not specified by POSIX.

Answered By - John Bollinger