Untested

You hadn't specified the arguments were regex patterns, not text.

perl -le'while(<STDIN>) { print $1 if /^$ARGV[0]:([^:]*):$ARGV[1]:/ }'
[download]

If they are actually text, then you can easily solve that for the Perl version.

perl -le'while(<STDIN>) { print $1 if /^\Q$ARGV[0]\E:([^:]*):\Q$ARGV[1
+]\E:/ }'
[download]

Using split is bit longer, but more readable.

I forgot to include the file name, but I think you get the idea.

And you're complaining about the code I gave? I can't wait for the day you create a user name, password, database name or server name that's a subset of a user name, password, database name or server name (which doesn't seem unlikely).

The direct equivalent of:

database=...
user=...
PASSWD=`cat password_file | grep $database | grep $user | cut -d: -f2`
[download]

is

my $database = '...';
my $user = '...';
my $PASSWD = '';
open(my $fh, '<', 'password_file') or die("open: $!\n");
while (<$fh>) {
   chomp;
   $PASSWD = (split /:/)[2] if /$database/ && /$user/;
}
[download]

Not bad. Replacing 4 entire programs with 5 statements. (Although cat is totally useless in your code.)

If you want an equivalent drop-in replacement, you can use

PASSWD=`perl -e"chomp; print((split /:/)[2]) if /$database/ && /$user/
+" < password_file`
[download]

If you want a better drop-in replacement, you can use

PASSWD=`perl -F: -lane'print $F[1] if $F[2] eq $ARGV[0] && $F[0] eq $A
+RGV[1]' $database $user < password_file`
[download]