Sed delete everything between 2 patterns, but not including the patterns

Question

I have found a few examples on this, but none that work exactly how I would like.

I would like to delete everything between 1 and several other possible patterns, but not including the patterns themselves. The pattern pairs are per line only, not across multiple lines.

eg

:1543453 Brown Fox
:789 123456 Cat
:abcdef Yellow Duck

to

:Brown Fox
:Cat
:Yellow Duck

So the first pattern to match is the ":" and the second being "Brown" OR "Cat" OR "Yellow"

Answer 1

There's brute force and ignorance, which works well on occasion:

sed -e 's/^:.* Brown/:Brown/' \
    -e 's/^:.* Cat/:Cat/' \
    -e 's/^:.* Yellow/:Yellow/' \
    data-file.txt

You might be able to use 'extended regular expressions' with the -E (BSD, Mac, Linux) or -r (Linux only) options:

sed -E 's/^:.* (Brown|Cat|Yellow)/:\1/' data-file.txt

Both produce the desired output on the sample data.

Note that the .* used is 'greedy'. Given the input file:

:1543453 Brown Fox
:789 123456 Cat
:abcdef Yellow Duck
:quantum mechanics eat Yellow Ducks for being yellow (but leave Yellow Dafodils alone)

both the scripts produce:

:Brown Fox
:Cat
:Yellow Duck
:Yellow Dafodils alone)

You'd need Perl or a sed enhanced with PCRE (Perl-Compatible Regular Expressions), or some other program, to avoid the greediness. For example:

$ perl -n -e 'print if s/^:.*? (Brown|Cat|Yellow)/:\1/' data-file.txt
:Brown Fox
:Cat
:Yellow Duck
:Yellow Ducks for being yellow (but leave Yellow Dafodils alone)
$