As many of us, I recurrently need to extract multi line blocks of text from, let say, logs files. Depending of the structure of the block of text, a simple egrep "regexp" -Bn -An could be sufficient, or not. When it comes to match more complex regular expression over several lines of text, all traditional unix tools fails (well, I fail miserably) miserably to offer a simple solution.

The idea here is to provide a simple tool for matching complex regular expression over multiple lines of text.

mgrep -b "Warning" -e "ignored"

Will match a block of text where the first line contain the word "Warning" and the last line the word "ignored". The idea here is that such an expressions will be auto-anchored like this:

mgrep -b "^.*Warning.*$" -e "^.*ignored.*$"

One could give a block body constraint:

mgrep -b "Warning" -i "Unknow (media|mime) type" -e "ignored"

Will match the given text block only if one of it's inners lines match against "^.*Unknow (media|mime) type.*$"

Parts of the auto-anchoring could be dissabled by anchoring manually:

mgrep -b "^Warning" -e "ignored\.$"

Will be auto-anchored like this:

mgrep -b "^Warning.*$" -e "^.*ignored\.$"

Consecutive lines could be matched by providing several time the same switch:

mgrep -b "Warning" -b "^File" -e "ignored"

Will match a text block where the first line match "^.*Warning.*$", the second "^File.*$" and the last "^.*ignored.*$"

The last idea is to provide a simple scripting language:

Warning
^File
*
+MIME type
*(^Unknow media)|(Bad file type)
!Configuring NAT
+
ignored\.$

Here,

?, +, * and ! in front of the line have the same meaning as in regular expressions, but they match against new lines.

So, the following script match a text block which follow thoses rules:

The first line match

^.*Warning.*$

The second

^File.*$

Followed by zero or many lines

Followed by one or many lines matching

^MIME type.*$

Followed by zero or many lines matching

(^Unknow media.*$)|(^.*Bad file type.*$)

Followed by any line not matching

^.*Configuring NAT.*$

Followed by one or many lines

Finally, the last line in the pattern should match

^.*ignored\.$

That's it, let's back to work!