With multi-threaded decoding we need to be careful to flush

[dcpomatic.git] / src / lib / util.cc

diff --git a/src/lib/util.cc b/src/lib/util.cc
index 78fac8b3a101b44f4dc0e44ee2d9bacca4860713..2c01678296f489c8038c39160f83ebba6dfc7baf 100644 (file)
--- a/src/lib/util.cc
+++ b/src/lib/util.cc
@@ -62,6 +62,9 @@ extern "C" {
 #include <curl/curl.h>
 #include <glib.h>
 #include <pangomm/init.h>
+#include <unicode/utypes.h>
+#include <unicode/unistr.h>
+#include <unicode/translit.h>
 #include <boost/algorithm/string.hpp>
 #include <boost/range/algorithm/replace_if.hpp>
 #include <boost/thread.hpp>
@@ -395,8 +398,8 @@ DCPOMATIC_ENABLE_WARNINGS
        /* Render something to fontconfig to create its cache */
        list<StringText> subs;
        dcp::SubtitleString ss(
-               optional<string>(), false, false, false, dcp::Colour(), 42, 1, dcp::Time(), dcp::Time(), 0, dcp::HAlign::CENTER, 0, dcp::VAlign::CENTER, dcp::DIRECTION_LTR,
-               "Hello dolly", dcp::NONE, dcp::Colour(), dcp::Time(), dcp::Time()
+               optional<string>(), false, false, false, dcp::Colour(), 42, 1, dcp::Time(), dcp::Time(), 0, dcp::HAlign::CENTER, 0, dcp::VAlign::CENTER, dcp::Direction::LTR,
+               "Hello dolly", dcp::Effect::NONE, dcp::Colour(), dcp::Time(), dcp::Time()
                );
        subs.push_back (StringText(ss, 0));
        render_text (subs, list<shared_ptr<Font>>(), dcp::Size(640, 480), DCPTime(), 24);
@@ -771,28 +774,20 @@ careful_string_filter (string s)
           Safety first and all that.
        */
 
-       wstring ws = boost::locale::conv::utf_to_utf<wchar_t>(s);
+       /* First transliterate using libicu to try to remove accents in a "nice" way */
+       auto icu_utf16 = icu::UnicodeString::fromUTF8(icu::StringPiece(s));
+       auto status = U_ZERO_ERROR;
+       auto transliterator = icu::Transliterator::createInstance("NFD; [:M:] Remove; NFC", UTRANS_FORWARD, status);
+       transliterator->transliterate(icu_utf16);
+       s.clear ();
+       icu_utf16.toUTF8String(s);
 
+       /* Then remove anything that's not in a very limited character set */
+       wstring ws = boost::locale::conv::utf_to_utf<wchar_t>(s);
        string out;
        string const allowed = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-_%.+";
        for (size_t i = 0; i < ws.size(); ++i) {
-
                wchar_t c = ws[i];
-
-               /* Remove some accents */
-               if (wstring(L"áàâ").find(c) != string::npos) {
-                       c = 'a';
-               }
-               if (wstring(L"éèêë").find(c) != string::npos) {
-                       c = 'e';
-               }
-               if (wstring(L"ö").find(c) != string::npos) {
-                       c = 'o';
-               }
-               if (wstring(L"ü").find(c) != string::npos) {
-                       c = 'u';
-               }
-
                if (allowed.find(c) != string::npos) {
                        out += c;
                }
@@ -870,11 +865,11 @@ remap (shared_ptr<const AudioBuffers> input, int output_channels, AudioMapping m
 Eyes
 increment_eyes (Eyes e)
 {
-       if (e == EYES_LEFT) {
-               return EYES_RIGHT;
+       if (e == Eyes::LEFT) {
+               return Eyes::RIGHT;
        }
 
-       return EYES_LEFT;
+       return Eyes::LEFT;
 }
 
 void
@@ -954,7 +949,7 @@ void
 emit_subtitle_image (ContentTimePeriod period, dcp::SubtitleImage sub, dcp::Size size, shared_ptr<TextDecoder> decoder)
 {
        /* XXX: this is rather inefficient; decoding the image just to get its size */
-       FFmpegImageProxy proxy (sub.png_image(), VIDEO_RANGE_FULL);
+       FFmpegImageProxy proxy (sub.png_image(), VideoRange::FULL);
        auto image = proxy.image().image;
        /* set up rect with height and width */
        dcpomatic::Rect<double> rect(0, 0, image->size().width / double(size.width), image->size().height / double(size.height));