Output all the paragraphs that match

print is short for print $, and $ contains the entire file. It's $& that contains the matched substring.

perl -gne'print $& while /^(?:\S.\n)font\.size = 12\.0\n(?:\S.\n)/mg'

---

The following also works:

perl -gne'print /^(?:\S.\n)font\.size = 12\.0\n(?:\S.\n)/mg'

Here the match is evaluated in list context. While a match is scalar context simply returns true or false depending on whether a match occurred, a match in list context returns all of the captured text, or all of the matched substrings if there are no captures. Since I have already eliminated the captures for being a useless drain, we can simply print the scalars returned by the match operator.

---

Note that this format normally accepts blank lines. You should really be searching for a (lack of) header.

perl -gne'print $& while /
   ^
   \[\[ . \n
   (?: (?! \[\[ ) . \n )
   font\.size \s = \s 12\.0 \n
   (?: (?! \[\[ ) . \n )*
/xmg'

If the one-liners work for you, stick with those. But, sometimes the assumptions those require aren't that good. I'm wary of the possibility that paragraph mode might break if there is a blank line under a heading, which means the regex might fail because the "lines" where already broken up on that whitespace.

I might be including to use a positive lookahead assertion in split to break on headers while keeping them as part of the item. The trick is to properly recognized a header, because TOML, for example, has multi-line strings where the [ at the start of a line might be inside a string.

However, once split, the rest of the problem is the same. It's just different chunks of text than what paragraph mode might give you.

use v5.10;
my $config =