fmp4mux: Add language from tags

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1557>

mux/fmp4/examples/hls_vod.rs

@@ -360,6 +360,10 @@ impl AudioStream {
             .property("samplesperbuffer", 4410)
             .property_from_str("wave", &self.wave)
             .build()?;
+        let taginject = gst::ElementFactory::make("taginject")
+            .property_from_str("tags", &format!("language-code={}", self.lang))
+            .property_from_str("scope", "stream")
+            .build()?;
         let raw_capsfilter = gst::ElementFactory::make("capsfilter")
             .property(
                 "caps",
@@ -374,9 +378,23 @@ impl AudioStream {
             .build()?;
         let appsink = gst_app::AppSink::builder().buffer_list(true).build();
 
-        pipeline.add_many([&src, &raw_capsfilter, &enc, &mux, appsink.upcast_ref()])?;
+        pipeline.add_many([
+            &src,
+            &taginject,
+            &raw_capsfilter,
+            &enc,
+            &mux,
+            appsink.upcast_ref(),
+        ])?;
 
-        gst::Element::link_many([&src, &raw_capsfilter, &enc, &mux, appsink.upcast_ref()])?;
+        gst::Element::link_many([
+            &src,
+            &taginject,
+            &raw_capsfilter,
+            &enc,
+            &mux,
+            appsink.upcast_ref(),
+        ])?;
 
         probe_encoder(state, enc);
 
@@ -416,7 +434,7 @@ fn main() -> Result<(), Error> {
             },
             AudioStream {
                 name: "audio_1".to_string(),
-                lang: "fre".to_string(),
+                lang: "fra".to_string(),
                 default: false,
                 wave: "white-noise".to_string(),
             },

mux/fmp4/src/fmp4mux/boxes.rs

@@ -700,7 +700,6 @@ fn write_tref(
 fn language_code(lang: impl std::borrow::Borrow<[u8; 3]>) -> u16 {
     let lang = lang.borrow();
 
-    // TODO: Need to relax this once we get the language code from tags
     assert!(lang.iter().all(u8::is_ascii_lowercase));
 
     (((lang[0] as u16 - 0x60) & 0x1F) << 10)
@@ -710,7 +709,7 @@ fn language_code(lang: impl std::borrow::Borrow<[u8; 3]>) -> u16 {
 
 fn write_mdhd(
     v: &mut Vec<u8>,
-    _cfg: &super::HeaderConfiguration,
+    cfg: &super::HeaderConfiguration,
     stream: &super::HeaderStream,
     creation_time: u64,
 ) -> Result<(), Error> {
@@ -724,8 +723,11 @@ fn write_mdhd(
     v.extend(0u64.to_be_bytes());
 
     // Language as ISO-639-2/T
-    // TODO: get actual language from the tags
+    if let Some(lang) = cfg.language_code {
+        v.extend(language_code(lang).to_be_bytes());
+    } else {
         v.extend(language_code(b"und").to_be_bytes());
+    }
 
     // Pre-defined
     v.extend([0u8; 2]);

mux/fmp4/src/fmp4mux/imp.rs

@@ -250,6 +250,8 @@ struct State {
     end_pts: Option<gst::ClockTime>,
     /// Start DTS of the whole stream
     start_dts: Option<gst::ClockTime>,
+    /// Language code from tags
+    language_code: Option<[u8; 3]>,
 
     /// Start PTS of the current fragment
     fragment_start_pts: Option<gst::ClockTime>,
@@ -2707,6 +2709,7 @@ impl FMP4Mux {
             streams,
             write_mehd: settings.write_mehd,
             duration: if at_eos { duration } else { None },
+            language_code: state.language_code,
             start_utc_time: if variant == super::Variant::ONVIF {
                 state
                     .earliest_pts
@@ -3140,8 +3143,22 @@ impl AggregatorImpl for FMP4Mux {
 
                 self.parent_sink_event(aggregator_pad, event)
             }
-            EventView::Tag(_ev) => {
+            EventView::Tag(ev) => {
-                // TODO: Maybe store for putting into the headers of the next fragment?
+                if let Some(tag_value) = ev.tag().get::<gst::tags::LanguageCode>() {
+                    let lang = tag_value.get();
+                    gst::trace!(CAT, imp: self, "Received language code from tags: {:?}", lang);
+
+                    // Language as ISO-639-2/T
+                    if lang.len() == 3 && lang.chars().all(|c| c.is_ascii_lowercase()) {
+                        let mut state = self.state.lock().unwrap();
+
+                        let mut language_code: [u8; 3] = [0; 3];
+                        for (out, c) in Iterator::zip(language_code.iter_mut(), lang.chars()) {
+                            *out = c as u8;
+                        }
+                        state.language_code = Some(language_code);
+                    }
+                }
 
                 self.parent_sink_event(aggregator_pad, event)
             }

mux/fmp4/src/fmp4mux/mod.rs

@@ -85,6 +85,7 @@ pub(crate) struct HeaderConfiguration {
 
     write_mehd: bool,
     duration: Option<gst::ClockTime>,
+    language_code: Option<[u8; 3]>,
 
     /// Start UTC time in ONVIF mode.
     /// Since Jan 1 1601 in 100ns units.

mux/mp4/src/mp4mux/boxes.rs

@@ -459,7 +459,6 @@ fn write_mdia(
 fn language_code(lang: impl std::borrow::Borrow<[u8; 3]>) -> u16 {
     let lang = lang.borrow();
 
-    // TODO: Need to relax this once we get the language code from tags
     assert!(lang.iter().all(u8::is_ascii_lowercase));
 
     (((lang[0] as u16 - 0x60) & 0x1F) << 10)
@@ -469,7 +468,7 @@ fn language_code(lang: impl std::borrow::Borrow<[u8; 3]>) -> u16 {
 
 fn write_mdhd(
     v: &mut Vec<u8>,
-    _header: &super::Header,
+    header: &super::Header,
     stream: &super::Stream,
     creation_time: u64,
 ) -> Result<(), Error> {
@@ -492,8 +491,11 @@ fn write_mdhd(
     v.extend(duration.to_be_bytes());
 
     // Language as ISO-639-2/T
-    // TODO: get actual language from the tags
+    if let Some(lang) = header.language_code {
+        v.extend(language_code(lang).to_be_bytes());
+    } else {
         v.extend(language_code(b"und").to_be_bytes());
+    }
 
     // Pre-defined
     v.extend([0u8; 2]);

mux/mp4/src/mp4mux/imp.rs

@@ -153,6 +153,9 @@ struct State {
 
     /// Size of the `mdat` as written so far.
     mdat_size: u64,
+
+    /// Language code from tags
+    language_code: Option<[u8; 3]>,
 }
 
 #[derive(Default)]
@@ -1166,6 +1169,25 @@ impl AggregatorImpl for MP4Mux {
                 }
                 self.parent_sink_event_pre_queue(aggregator_pad, event)
             }
+            EventView::Tag(ev) => {
+                if let Some(tag_value) = ev.tag().get::<gst::tags::LanguageCode>() {
+                    let lang = tag_value.get();
+                    gst::trace!(CAT, imp: self, "Received language code from tags: {:?}", lang);
+
+                    // Language as ISO-639-2/T
+                    if lang.len() == 3 && lang.chars().all(|c| c.is_ascii_lowercase()) {
+                        let mut state = self.state.lock().unwrap();
+
+                        let mut language_code: [u8; 3] = [0; 3];
+                        for (out, c) in Iterator::zip(language_code.iter_mut(), lang.chars()) {
+                            *out = c as u8;
+                        }
+                        state.language_code = Some(language_code);
+                    }
+                }
+
+                self.parent_sink_event_pre_queue(aggregator_pad, event)
+            }
             _ => self.parent_sink_event_pre_queue(aggregator_pad, event),
         }
     }
@@ -1365,6 +1387,7 @@ impl AggregatorImpl for MP4Mux {
                 variant: self.obj().class().as_ref().variant,
                 movie_timescale: settings.movie_timescale,
                 streams,
+                language_code: state.language_code,
             })
             .map_err(|err| {
                 gst::error!(CAT, imp: self, "Failed to create moov box: {err}");

mux/mp4/src/mp4mux/mod.rs

@@ -135,6 +135,7 @@ pub(crate) struct Header {
     /// Pre-defined movie timescale if not 0.
     movie_timescale: u32,
     streams: Vec<Stream>,
+    language_code: Option<[u8; 3]>,
 }
 
 #[allow(clippy::upper_case_acronyms)]