Fix crashes when using templates in some cases (#2491).
[dcpomatic.git] / src / lib / text_decoder.cc
index c655fb5b58d5e5b2e208931f27d4ae0e753d9632..1ecdcd34783ae8050fbb6c282d27a511aabcef44 100644 (file)
@@ -41,12 +41,10 @@ using namespace dcpomatic;
 
 TextDecoder::TextDecoder (
        Decoder* parent,
-       shared_ptr<const TextContent> content,
-       ContentTime first
+       shared_ptr<const TextContent> content
        )
        : DecoderPart (parent)
        , _content (content)
-       , _position (first)
 {
 
 }
@@ -63,27 +61,13 @@ void
 TextDecoder::emit_bitmap_start (ContentBitmapText const& bitmap)
 {
        BitmapStart (bitmap);
-       _position = bitmap.from();
-}
-
-
-static
-string
-escape_text (string text)
-{
-       /* We must escape some things, otherwise they might confuse our subtitle
-          renderer (which uses entities and some HTML-esque markup to do bold/italic etc.)
-       */
-       boost::algorithm::replace_all(text, "&", "&amp;");
-       boost::algorithm::replace_all(text, "<", "&lt;");
-       boost::algorithm::replace_all(text, ">", "&gt;");
-       return text;
+       maybe_set_position(bitmap.from());
 }
 
 
 static
 void
-set_forced_appearance(shared_ptr<const TextContent> content, dcp::SubtitleString& subtitle)
+set_forced_appearance(shared_ptr<const TextContent> content, StringText& subtitle)
 {
        if (content->colour()) {
                subtitle.set_colour(*content->colour());
@@ -103,16 +87,64 @@ set_forced_appearance(shared_ptr<const TextContent> content, dcp::SubtitleString
 }
 
 
+string
+TextDecoder::remove_invalid_characters_for_xml(string text)
+{
+       string output;
+
+       /* https://www.w3.org/TR/REC-xml/#charsets says that XML may only contain 0x9, 0xa, 0xd below 0x32.
+        * Not sure if we should be doing direct UTF-8 manipulation here.
+        */
+       for (size_t i = 0; i < text.length(); ++i) {
+               auto const c = text[i];
+               if ((c & 0xe0) == 0xc0) {
+                       // start of 2-byte code point
+                       output += c;
+                       output += text[i + 1];
+                       ++i;
+               } else if ((c & 0xf0) == 0xe0) {
+                       // start of 3-byte code point
+                       output += c;
+                       output += text[i + 1];
+                       output += text[i + 2];
+                       i += 2;
+               } else if ((c & 0xf8) == 0xf0) {
+                       // start of 4-byte code point
+                       output += c;
+                       output += text[i + 1];
+                       output += text[i + 2];
+                       output += text[i + 3];
+                       i += 3;
+               } else {
+                       if (c >= 0x20 || c == 0x9 || c == 0xa || c == 0xd) {
+                               output += c;
+                       }
+               }
+       }
+
+       return output;
+}
+
+
 void
-TextDecoder::emit_plain_start (ContentTime from, vector<dcp::SubtitleString> subtitles)
+TextDecoder::emit_plain_start(ContentTime from, vector<dcp::SubtitleString> subtitles, dcp::SubtitleStandard valign_standard)
 {
+       vector<StringText> string_texts;
+
        for (auto& subtitle: subtitles) {
-               subtitle.set_text(escape_text(subtitle.text()));
-               set_forced_appearance(content(), subtitle);
+               auto string_text = StringText(
+                       subtitle,
+                       content()->outline_width(),
+                       content()->get_font(subtitle.font().get_value_or("")),
+                       valign_standard
+                       );
+               string_text.set_text(remove_invalid_characters_for_xml(string_text.text()));
+               set_forced_appearance(content(), string_text);
+               string_texts.push_back(string_text);
        }
 
-       PlainStart(ContentStringText(from, subtitles));
-       _position = from;
+       PlainStart(ContentStringText(from, string_texts));
+       maybe_set_position(from);
 }
 
 
@@ -143,7 +175,7 @@ TextDecoder::emit_plain_start (ContentTime from, sub::Subtitle const & sub_subti
                }
        }
 
-       vector<dcp::SubtitleString> dcp_subtitles;
+       vector<StringText> string_texts;
        for (auto line: sub_subtitle.lines) {
                for (auto block: line.blocks) {
 
@@ -160,13 +192,20 @@ TextDecoder::emit_plain_start (ContentTime from, sub::Subtitle const & sub_subti
                                switch (line.vertical_position.reference.get_value_or(sub::BOTTOM_OF_SCREEN)) {
                                case sub::BOTTOM_OF_SCREEN:
                                case sub::TOP_OF_SUBTITLE:
-                                       /* This 1.015 is an arbitrary value to lift the bottom sub off the bottom
+                                       /* This 0.1 is an arbitrary value to lift the bottom sub off the bottom
                                           of the screen a bit to a pleasing degree.
                                           */
-                                       v_position = 1.015 -
+                                       v_position = 0.1 +
                                                (1 + bottom_line.get() - line.vertical_position.line.get()) * multiplier;
 
-                                       v_align = dcp::VAlign::TOP;
+                                       /* Align our subtitles to the bottom of the screen, because if we are making a SMPTE
+                                        * DCP and the projection system uses the wrong standard to interpret vertical position,
+                                        * a bottom-aligned subtitle will be less wrong than a top-aligned one.  This is because
+                                        * in the top-aligned case the difference will be the distance between bbox top an
+                                        * baseline, but in the bottom-aligned case the difference will be between bbox bottom
+                                        * and baseline (which is shorter).
+                                        */
+                                       v_align = dcp::VAlign::BOTTOM;
                                        break;
                                case sub::TOP_OF_SCREEN:
                                        /* This 0.1 is another fudge factor to bring the top line away from the top of the screen a little */
@@ -233,47 +272,54 @@ TextDecoder::emit_plain_start (ContentTime from, sub::Subtitle const & sub_subti
                           content by the other emit_plain_start() above.
                        */
 
-                       dcp_subtitles.push_back(
-                               dcp::SubtitleString (
-                                       string(TEXT_FONT_ID),
-                                       block.italic,
-                                       block.bold,
-                                       block.underline,
-                                       block.colour.dcp(),
-                                       block.font_size.points (72 * 11),
-                                       1.0,
-                                       dcp::Time (from.seconds(), 1000),
-                                       /* XXX: hmm; this is a bit ugly (we don't know the to time yet) */
-                                       dcp::Time (),
-                                       h_position,
-                                       h_align,
-                                       v_position,
-                                       v_align,
-                                       dcp::Direction::LTR,
-                                       block.text,
-                                       dcp::Effect::NONE,
-                                       block.effect_colour.get_value_or(sub::Colour(0, 0, 0)).dcp(),
-                                       /* Hack: we should use subtitle.fade_up and subtitle.fade_down here
-                                          but the times of these often don't have a frame rate associated
-                                          with them so the sub::Time won't convert them to milliseconds without
-                                          throwing an exception.  Since only DCP subs fill those in (and we don't
-                                          use libsub for DCP subs) we can cheat by just putting 0 in here.
-                                       */
-                                       dcp::Time (),
-                                       dcp::Time (),
-                                       0
-                                       )
+                       auto dcp_colour = [](sub::Colour const& c) {
+                               return dcp::Colour(lrintf(c.r * 255), lrintf(c.g * 255), lrintf(c.b * 255));
+                               };
+
+                       auto dcp_subtitle = dcp::SubtitleString(
+                               optional<string>(),
+                               block.italic,
+                               block.bold,
+                               block.underline,
+                               dcp_colour(block.colour),
+                               block.font_size.points (72 * 11),
+                               1.0,
+                               dcp::Time (from.seconds(), 1000),
+                               /* XXX: hmm; this is a bit ugly (we don't know the to time yet) */
+                               dcp::Time (),
+                               h_position,
+                               h_align,
+                               v_position,
+                               v_align,
+                               0,
+                               dcp::Direction::LTR,
+                               remove_invalid_characters_for_xml(block.text),
+                               dcp::Effect::NONE,
+                               dcp_colour(block.effect_colour.get_value_or(sub::Colour(0, 0, 0))),
+                               /* Hack: we should use subtitle.fade_up and subtitle.fade_down here
+                                  but the times of these often don't have a frame rate associated
+                                  with them so the sub::Time won't convert them to milliseconds without
+                                  throwing an exception.  Since only DCP subs fill those in (and we don't
+                                  use libsub for DCP subs) we can cheat by just putting 0 in here.
+                               */
+                               dcp::Time (),
+                               dcp::Time (),
+                               0
                                );
-               }
-       }
 
-       for (auto& subtitle: dcp_subtitles) {
-               subtitle.set_text(escape_text(subtitle.text()));
-               set_forced_appearance(content(), subtitle);
+                       auto string_text = StringText(
+                               dcp_subtitle,
+                               content()->outline_width(),
+                               content()->get_font(block.font.get_value_or("")),
+                               dcp::SubtitleStandard::SMPTE_2014
+                               );
+                       set_forced_appearance(content(), string_text);
+                       string_texts.push_back(string_text);
+               }
        }
 
-       PlainStart(ContentStringText(from, dcp_subtitles));
-       _position = from;
+       PlainStart(ContentStringText(from, string_texts));
+       maybe_set_position(from);
 }
 
 
@@ -285,9 +331,9 @@ TextDecoder::emit_stop (ContentTime to)
 
 
 void
-TextDecoder::emit_plain (ContentTimePeriod period, vector<dcp::SubtitleString> subtitles)
+TextDecoder::emit_plain(ContentTimePeriod period, vector<dcp::SubtitleString> subtitles, dcp::SubtitleStandard valign_standard)
 {
-       emit_plain_start (period.from, subtitles);
+       emit_plain_start (period.from, subtitles, valign_standard);
        emit_stop (period.to);
 }
 
@@ -317,3 +363,13 @@ TextDecoder::seek ()
 {
        _position = ContentTime ();
 }
+
+
+void
+TextDecoder::maybe_set_position (dcpomatic::ContentTime position)
+{
+       if (!_position || position > *_position) {
+               _position = position;
+       }
+}
+