match and replace multiple newlines with a SED or PERL one-liner

Question

I have an input C file (myfile.c) that looks like this :

void func_foo();
void func_bar();

//supercrazytag

I want to use a shell command to insert new function prototypes, such that the output becomes:

void func_foo();
void func_bar();
void func_new();

//supercrazytag

So far I've been unsuccessful using SED or PERL. What didn't work:

sed 's|\n\n//supercrazytag|void func_new();\n\n//supercrazytag|g' < myfile.c
sed 's|(\n\n//supercrazytag)|void func_new();\1|g' < myfile.c

Using the same patterns with perl -pe "....." didn't work either.

What am I missing ? I've tried many different approaches, including this and this and that.

Answer 1

This is actually an adaptation of an answer I just gave here, on detecting a certain point in a file.

In pseudocode:
1. start reading and writing lines from the file
2. when we find the end of the prototype section, insert some new text

use strict;
use warnings;

my $new_prototype = 'void func_new();';
my $seen_proto;

while (<>)
{
    if (/^void \w+\(\);$/ .. /^$/)
    {
        # in the middle of the prototype section
        $seen_proto = 1;
    }
    else
    {
        # this code is run when either before or after the prototype section

        # if we have seen prototypes, we're just after that section - print
        # out the new prototype and then reset our flag so we don't do it twice
        print "$new_prototype\n" and $seen_proto-- if $seen_proto;
    }

    # print out the original line
    print;
}

Put this code in process.pl and run via: perl process.pl < mycode.cpp > mycode_new.cpp

Answer 2

For "perl -pe", your problem is that it is processing line by line, so there is no way it will find "\n\n". If you add the -0777 flag to Perl (to make it process the whole file at once) it will work:

perl -0777 -pe "s|(\n\n//supercrazytag)|\nvoid func_new();$1|g" myfile.c

I also changed the (deprecated for this usage) \1 to $1 and added an extra "\n" to beginning of the replacement for readability.

See perlrun (Command Switches) for an explanation of the odd-looking "-0777"

Answer 3

This will work:

sed '/^$/N;s|\n//supercrazytag|void func_new();\n\n//supercrazytag|' myfile.c

EDIT:
Or more concisely:

sed '/^$/N;s|\(\n//supercrazytag\)|void func_new();\n\1|' myfile.c

Answer 4

awk  '/supercrazy/{$0="void func_new()\n\n"$0}1'  file