#!/usr/bin/perl -CIOESA -w

# Download quotes from noobfarm.org, format into a fortune file.

($SELF = $0) =~ s,.*/,,;
$VERSION = "0.9.99";

$fortunefile = 'noobfarm';
$jsonfile = 'noobfarm.json';
$indent = " " x 30;

use JSON;
use Getopt::Long;

sub usage {
	print <<EOF;
$SELF [-d] [-n] [-p] [-o dir]

$SELF v$VERSION by B. Watson (urchlay\@slackware.uk), released
under the WTFPL. Do WTF you like with this. For full license, see
http://sam.zoy.org/wtfpl/

Downloads quotes from noobfarm.org, formats them as a fortune
database. Output is a text file called "noobfarm" and an index called
"noobfarm.dat", created in the current directory (or wherever the -o
option says).

Options:
 -h, --help
   This usage message.
 -o, --output-dir=DIR
   Write the noobfarm and noobfarm.dat files here. This could be
   /usr/share/games/fortunes to write directly to the system-wide
   fortune database. Default is the current directory.
 -d, --download-only
   Download the quotes, but do not parse or create a fortune file.
   The downloaded file will be saved as "$jsonfile" in the
   current directory. This option automatically enables -k/--keep-json.
 -n, --no-format
   Do not format the output with the fmt(1) command. Default is to use
   "fmt -s" to break up long lines in the fortune file.
 -p, --process-only
   Do not download the quotes. Instead, use the file "$jsonfile"
   in the current directory (possibly created by a previous run with
   the -d, --download-only option, and possibly edited since then).
   This option automatically enables -k/--keep-json.
 -k, --keep-json
   Do not delete "$jsonfile" before exiting. Normally, without the
   -d/--download-only or -p/--process-only options, this gets deleted.

Notes:
 - This script is NOT supported by the owners of noobfarm.org. Do NOT
   contact them if you have problems.
 - Single-quotes and/or backslashes are silently removed from the
   directory name for --output-dir. WONTFIX. Spaces are actually OK :)
 - Exit status is generally 0 (success) or non-zero (failure). But
   every possible failure mode hasn't been tested, so YMMV.
 - I wouldn't run this with elevated privileges, or from something
   like a CGI script, if I were you. It hasn't been audited for
   security and never will be (use or use not, there is no warranty).
EOF
	exit ($_[0] || 0);
}

# Download all the quotes in one request. As of this writing, it's a
# 1.1MB json file.
# wget warns about wildcards in HTTP (the * in the URL), but changing
# it to an escaped form, %2a, breaks the search (0 results). So let
# it warn.
# We ask for the first billion quotes all at once. I think it'll be
# several eons before there are more than a billion quotes...
sub download {
	unlink($jsonfile);
	system(
		"wget " .
		"-O$jsonfile " .
		"--header='Accept: application/json' " .
		"https://noobfarm.org/search/Approved:T*/1/999999999"
	) && die "wget failed"; # the && is not a typo! non-zero means error.
}

# Read JSON from a filehandle. Return the entire fortune file as a string,
# ready to use except for breaking up long lines (caller does that).
sub format_quotes {
	my $fh = shift;
	local $/;
	undef $/;

	$input = <$fh>;
	# from_json will die() on invalid input. Let it.
	my $j = from_json($input);

	for(@{$j->{Quotes}}) {
		my $quote = $_->{Quote};
		my $date = $_->{Submitted};

		# \ are doubled
		$quote =~ s,\\\\,\\,g;

		# some (but not all!) newlines are encoded as \\n instead of \n.
		# this is a potentially lossy transform: if the user in the quote
		# actually said a literal '\n', this will turn it into a newline.
		# in practice, this doesn't seem to be a problem.
		$quote =~ s,\\\\?n,\n,g;

		# turn \uXXXX hex escapes back into utf8 sequences.
		$quote =~ s,\\u([0-9a-fA-F]{4}),chr(hex $1),ge,;

		# dates look like: 2020-10-09T22:38:39.404662072Z
		# we only want 2020-10-09.
		$date = substr($date, 0, 10);

		$quotes{$_->{ID}} = $quote;
		$dates{$_->{ID}} = $date;
	}

	for(sort { $a <=> $b } keys %quotes) {
		push @output, $quotes{$_} . "\n" .
			"$indent-- noobfarm.org, quote #$_, $dates{$_}\n";
	}

	close $fh;
	return join "%\n", @output;
}

# main()

$help = $dlonly = $process_only = $outdir = 0;

### Parse options
GetOptions(
	'help'          => \$help,
	'download-only' => \$dlonly,
	'output-dir=s'  => \$outdir,
	'no-format'     => \$no_format,
   'process-only'  => \$process_only,
	'keep-json'     => \$keep_json,
) || usage(1);

$keep_json++ if $dlonly | $process_only;

### Check options
die "--download-only and --process-only can't both be given\n"
	if $dlonly && $process_only;
die "Unknown argument(s): @ARGV" if @ARGV;
usage(0) if $help;

### Create output dir, if needed
if($outdir) {
	$outdir =~ s/['\\]//g;
	system("mkdir", "-p", $outdir);
	exit 1 unless -d $outdir && -w _;
	$fortunefile = "$outdir/$fortunefile";
}

### Download the JSON file (and exit, if --download-only)
download unless $process_only;
exit 0 if $dlonly;

### Open input & output filehandles
open($infh, "<", $jsonfile) or die $!;
if($no_format) {
	open(OUT, ">", $fortunefile) or die $!;
} else {
	# using a shell pipe like this is the reason for not allowing
	# single quotes or backslashes in the output dir. I know it's not
	# "modern" or "best practice" but it's *so* convenient.
	open(OUT, "|fmt -s > '$fortunefile'") or die $!;
}

### Write output, close filehandles
print OUT format_quotes($infh);
close $infh;
close OUT;

### Clean up, unless asked not to
if($keep_json) {
	print STDERR "Keeping JSON file $jsonfile\n";
} else {
	unlink($jsonfile);
}

### Create the index fortune needs. Let strfile's exit status be ours.
exec("strfile '$fortunefile'");