[SOLVED] sed regexp - extra unwanted line in matching output

Issue

I have this file

~/ % cat t
---
    abc
def DEF    
ghi GHI
---
123
456

and I would like to extract the content between the three dashes, so I try

sed -En '{N; /^---\s{5}\w+/,/^---/p}' t

I.e. 3 dashes followed by 5 whitespaces including the newline, followed by one or more word characters and ending with another set of three dashes. This gives me this output

~/ % sed -En '{N; /^---\s{5}\w+/,/^---/p}' t
--- 
    abc
def DEF
ghi GHI
---
123

I don't want the line with "123". Why am I getting that and how do I adjust my expression to get rid of it? [EDIT]: It is important that the four spaces of indentation after the first three dashes are matched in the expression.

Solution

This might work for you (GNU sed):

sed -En '/^---/{:a;N;/^ {4}\S/M!D;/\n---/!ba;p}' file

Turn on extended regexp (-E) and off implicit printing (-n).

If a line begins --- and the following line is indented by 4 spaces, gather up the following lines until another begins --- and print them.

If the following line does not match the above criteria, delete the first and repeat.

All other lines will pass through unprinted.

N.B. The M flag on the second regexp for multiline matching , since the first line already begins --- the next must be indented.

Answered By - potong