Give emoji-only post titles distinct filenames (#52)

lib/Plerd/Post.pm

@@ -11,6 +11,8 @@ use HTML::SocialMeta;
 use Try::Tiny;
 use JSON;
 use Path::Class::File;
+use Digest::MD5 qw( md5_hex );
+use Encode qw( encode_utf8 );
 
 use Plerd::SmartyPants;
 use Web::Mention;
@@ -194,15 +196,22 @@ sub _build_published_filename {
         $filename =~ s/\..*$/.html/;
     }
     else {
-        $filename = $self->title;
         my $stripper = HTML::Strip->new( emit_spaces => 0 );
-        $filename = $stripper->parse( $filename );
-        $filename =~ s/\s+/-/g;
-        $filename =~ s/--+/-/g;
-        $filename =~ s/[^\w\-]+//g;
-        $filename = lc $filename;
-        $filename = $self->date->ymd( q{-} ) . q{-} . $filename;
-        $filename .= '.html';
+        my $slug = $stripper->parse( $self->title );
+        $slug =~ s/\s+/-/g;
+        $slug =~ s/--+/-/g;
+        $slug =~ s/[^\w\-]+//g;
+        $slug = lc $slug;
+
+        # A title made entirely of characters that can't appear in a filename
+        # (e.g. emoji) slugs down to nothing, which would make every such post
+        # on a given day collide on "YYYY-MM-DD-.html". Fall back to a short
+        # hash of the title, so distinct titles get distinct, stable filenames.
+        unless ( length $slug ) {
+            $slug = substr( md5_hex( encode_utf8( $self->title ) ), 0, 10 );
+        }
+
+        $filename = $self->date->ymd( q{-} ) . q{-} . $slug . '.html';
     }
 
     return $filename;

t/emoji_title.t

@@ -0,0 +1,71 @@
+use warnings;
+use strict;
+use utf8;
+use Test::More;
+use Path::Class::Dir;
+use Path::Class::File;
+use URI;
+
+use FindBin;
+use lib "$FindBin::Bin/../lib";
+
+use_ok( 'Plerd' );
+use Plerd::Init;
+
+# Regression test for GitHub issue #52: a post titled entirely with
+# characters that can't appear in a filename (e.g. emoji) used to slug down
+# to nothing, so every such post on a given day collided on
+# "YYYY-MM-DD-.html" and overwrote its predecessor. Distinct titles must
+# produce distinct, non-empty published filenames.
+
+my $blog_dir = Path::Class::Dir->new( "$FindBin::Bin/emoji_title_blog" );
+$blog_dir->rmtree;
+Plerd::Init::initialize( $blog_dir->stringify, 0 );
+
+my $source_dir = Path::Class::Dir->new( $blog_dir, 'source' );
+
+# Two emoji-only titles, same publication day, source filenames carry no date
+# prefix so the title-based filename logic runs.
+Path::Class::File->new( $source_dir, 'smile.md' )->spew(
+    iomode => '>:encoding(utf8)',
+    "title: \x{1F600}\ntime: 2021-01-19T10:00:00\n\nBody one.\n",
+);
+Path::Class::File->new( $source_dir, 'hats.md' )->spew(
+    iomode => '>:encoding(utf8)',
+    "title: \x{1F3A9}\x{1F3A9}\ntime: 2021-01-19T11:00:00\n\nBody two.\n",
+);
+
+my $plerd = Plerd->new(
+    path         => $blog_dir->stringify,
+    title        => 'Test Blog',
+    author_name  => 'Nobody',
+    author_email => 'nobody@example.com',
+    base_uri     => URI->new( 'http://blog.example.com/' ),
+);
+
+$plerd->publish_all;
+
+my %filename;
+for my $post ( @{ $plerd->posts } ) {
+    $filename{ $post->title } = $post->published_filename;
+}
+
+my @names = values %filename;
+isnt( $names[0], $names[1],
+    'Two emoji-only titles get distinct published filenames.' );
+
+unlike( $_, qr/-\.html$/, 'Published filename is not empty-slugged.' )
+    for @names;
+
+# Both posts' HTML files should actually exist on disk (neither overwrote the
+# other).
+for my $title ( keys %filename ) {
+    ok(
+        -e Path::Class::File->new( $blog_dir, 'docroot', $filename{ $title } ),
+        "HTML file exists for title '$title' ($filename{ $title })."
+    );
+}
+
+$blog_dir->rmtree;
+
+done_testing();