regex in sed not working as in PCRE

Question

I have this source file.c:

#include <dir/header1>
#include <dir/header2>
#include <dir/header3>
...

I would like to match only those headers after dir, (e.g. header1). I do this:

$ sed -En 's/\/(.*)>/\1/p' file.c

but the will output:

#include <dirheader1, so it matched from the beginning of line, But I wanted to group from / character (therefor s/\/...). So I have tried:

$ sed -En 's/<.*\/(.*)>/\1/p' file.c

Neither that helped. So how to achieve that?

Answer 1

To extract only the headerx bit of the three lines that you show with the s/// command in sed, you will have to remove all bits of the line that you don't want.

The bits of the line that you don't want is the initial #include <dir/ string at the start, and the final > string at the end. This must all be matched by your regular expression. Then also use a capture group for the bit in-between the initial and final strings to replace the whole line with the captured string:

sed -n 's,.*/\(.*\)>,\1,p' file.c

or, more explicitly,

sed -n 's,^#include <dir/\(.*\)>,\1,p' file.c

Answer 2

You are not discarding the first part of the match so you could correct your code to do that.. as per @kusalananada

Alternatively a generic match using a negated pattern

sed -En "s|[^/]+/([^>]+)>|\1|p" file

For a specific match for #include

sed -En "s|^#include <dir/([^>]+)>|\1|p" file

Or with grep though you

grep -Po "(?<=#include <dir/)[^>]+" file

Or with gawk

awk -F"/|>" '$1~include{print $(NF-1)}' file