Merge 4f4ab53252 into 3c8ae0ae42

Fixed a panic in SpeechDispatcher.voices()

.github/workflows/release.yml

@@ -12,7 +12,7 @@ jobs:
     env:
       CARGO_TOKEN: ${{ secrets.CARGO_TOKEN }}
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - run: |
           sudo apt-get update
           sudo apt-get install -y libspeechd-dev

.github/workflows/test.yml

@@ -5,6 +5,17 @@ on:
   pull_request:
 
 jobs:
+  check_formatting:
+    name: Check Formatting
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - run: |
+          rustup toolchain install stable
+          cargo fmt --all --check
+          cd examples/web
+          cargo fmt --all --check
+
   check:
     name: Check
     strategy:
@@ -12,41 +23,38 @@ jobs:
         os: [windows-latest, ubuntu-22.04, macos-latest]
     runs-on: ${{ matrix.os }}
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - run: sudo apt-get update; sudo apt-get install -y libspeechd-dev
         if: ${{ runner.os == 'Linux' }}
       - run: |
           rustup toolchain install stable
-          cargo fmt --check
+          cargo clippy --all-targets
-          cargo clippy
 
   check_web:
     name: Check Web
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - run: |
           rustup target add wasm32-unknown-unknown
           rustup toolchain install stable
-          cargo fmt --all --check
+          cargo clippy --all-targets --target wasm32-unknown-unknown
-          cargo clippy --target wasm32-unknown-unknown
 
   check_android:
     name: Check Android
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - run: |
+          rustup target add aarch64-linux-android
           rustup toolchain install stable
-          rustup target add aarch64-linux-android armv7-linux-androideabi i686-linux-android x86_64-linux-android
+          cargo clippy --all-targets --target aarch64-linux-android
-          cargo install -f cargo-apk
-          cargo apk build
 
   check_web_example:
     name: Check Web Example
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - run: |
           rustup target add wasm32-unknown-unknown
           rustup toolchain install stable

Cargo.toml

@@ -1,9 +1,10 @@
 [package]
 name = "tts"
-version = "0.25.6"
+version = "0.26.0"
 authors = ["Nolan Darilek <nolan@thewordnerd.info>"]
 repository = "https://github.com/ndarilek/tts-rs"
 description = "High-level Text-To-Speech (TTS) interface"
+documentation = "https://docs.rs/tts"
 license = "MIT"
 exclude = ["*.cfg", "*.yml"]
 edition = "2021"
@@ -26,11 +27,11 @@ serde = { version = "1", optional = true, features = ["derive"] }
 thiserror = "1"
 
 [dev-dependencies]
-env_logger = "0.10"
+env_logger = "0.11"
 
 [target.'cfg(windows)'.dependencies]
 tolk = { version = "0.5", optional = true }
-windows = { version = "0.51", features = [
+windows = { version = "0.52", features = [
     "Foundation",
     "Foundation_Collections",
     "Media_Core",
@@ -64,7 +65,6 @@ web-sys = { version = "0.3", features = [
 [target.'cfg(target_os="android")'.dependencies]
 jni = "0.21"
 ndk-context = "0.1"
-ndk-glue = "0.7"
 
 [package.metadata.docs.rs]
 no-default-features = true

src/backends/android.rs

@@ -250,6 +250,7 @@ impl Backend for Android {
             utterance_callbacks: true,
             voice: false,
             get_voice: false,
+            synthesize: false,
         }
     }
 
@@ -284,6 +285,10 @@ impl Backend for Android {
         }
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         let vm = Self::vm()?;
         let mut env = vm.get_env()?;

src/backends/appkit.rs

@@ -125,6 +125,10 @@ impl Backend for AppKit {
         Ok(None)
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         trace!("stop()");
         unsafe {

src/backends/av_foundation.rs

@@ -170,6 +170,7 @@ impl Backend for AvFoundation {
             voice: true,
             get_voice: false,
             utterance_callbacks: true,
+            synthesize: false,
         }
     }
 
@@ -206,6 +207,10 @@ impl Backend for AvFoundation {
         Ok(Some(UtteranceId::AvFoundation(utterance)))
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         trace!("stop()");
         unsafe {

src/backends/speech_dispatcher.rs

@@ -85,6 +85,7 @@ impl Backend for SpeechDispatcher {
             voice: true,
             get_voice: false,
             utterance_callbacks: true,
+            synthesize: false,
         }
     }
 
@@ -108,6 +109,10 @@ impl Backend for SpeechDispatcher {
         }
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         trace!("stop()");
         self.0.cancel()?;
@@ -188,6 +193,7 @@ impl Backend for SpeechDispatcher {
             .0
             .list_synthesis_voices()?
             .iter()
+            .filter(|v| LanguageTag::parse(v.language.clone()).is_ok())
             .map(|v| Voice {
                 id: v.name.clone(),
                 name: v.name.clone(),

src/backends/tolk.rs

@@ -39,6 +39,10 @@ impl Backend for Tolk {
         Ok(None)
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         trace!("stop()");
         self.0.silence();

src/backends/web.rs

@@ -59,6 +59,7 @@ impl Backend for Web {
             voice: true,
             get_voice: true,
             utterance_callbacks: true,
+            synthesize: false,
         }
     }
 
@@ -121,6 +122,10 @@ impl Backend for Web {
         }
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        unimplemented!();
+    }
+
     fn stop(&mut self) -> Result<(), Error> {
         trace!("stop()");
         if let Some(window) = web_sys::window() {

src/backends/winrt.rs

@@ -12,8 +12,11 @@ use windows::{
     Media::{
         Core::MediaSource,
         Playback::{MediaPlayer, MediaPlayerAudioCategory},
-        SpeechSynthesis::{SpeechSynthesizer, VoiceGender, VoiceInformation},
+        SpeechSynthesis::{
+            SpeechSynthesisStream, SpeechSynthesizer, VoiceGender, VoiceInformation,
+        },
     },
+    Storage::Streams::DataReader,
 };
 
 use crate::{Backend, BackendId, Error, Features, Gender, UtteranceId, Voice, CALLBACKS};
@@ -138,6 +141,20 @@ impl WinRt {
             voice: SpeechSynthesizer::DefaultVoice()?,
         })
     }
+
+    fn create_synthesis_stream(&mut self, text: &str) -> Result<SpeechSynthesisStream, Error> {
+        self.synth.Options()?.SetSpeakingRate(self.rate.into())?;
+        self.synth.Options()?.SetAudioPitch(self.pitch.into())?;
+        self.synth.Options()?.SetAudioVolume(self.volume.into())?;
+
+        self.synth.SetVoice(&self.voice)?;
+        let stream = self
+            .synth
+            .SynthesizeTextToStreamAsync(&text.into())?
+            .get()?;
+
+        Ok(stream)
+    }
 }
 
 impl Backend for WinRt {
@@ -155,6 +172,7 @@ impl Backend for WinRt {
             voice: true,
             get_voice: true,
             utterance_callbacks: true,
+            synthesize: true,
         }
     }
 
@@ -189,18 +207,12 @@ impl Backend for WinRt {
             }
         }
         if no_utterances {
-            self.synth.Options()?.SetSpeakingRate(self.rate.into())?;
+            let stream = self.create_synthesis_stream(text)?;
-            self.synth.Options()?.SetAudioPitch(self.pitch.into())?;
+
-            self.synth.Options()?.SetAudioVolume(self.volume.into())?;
+            let media_source = MediaSource::CreateFromStream(&stream, &stream.ContentType()?)?;
-            self.synth.SetVoice(&self.voice)?;
+            self.player.SetSource(&media_source)?;
-            let stream = self
-                .synth
-                .SynthesizeTextToStreamAsync(&text.into())?
-                .get()?;
-            let content_type = stream.ContentType()?;
-            let source = MediaSource::CreateFromStream(&stream, &content_type)?;
-            self.player.SetSource(&source)?;
             self.player.Play()?;
+
             let mut callbacks = CALLBACKS.lock().unwrap();
             let callbacks = callbacks.get_mut(&self.id).unwrap();
             if let Some(callback) = callbacks.utterance_begin.as_mut() {
@@ -210,6 +222,18 @@ impl Backend for WinRt {
         Ok(Some(utterance_id))
     }
 
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error> {
+        let stream = self.create_synthesis_stream(text)?;
+
+        let size = stream.Size()?;
+        let data_reader = DataReader::CreateDataReader(&stream.GetInputStreamAt(0)?)?;
+        let mut bytes = vec![0; size as usize];
+        data_reader.LoadAsync(size as u32)?;
+        data_reader.ReadBytes(&mut bytes)?;
+
+        Ok(bytes)
+    }
+
     fn stop(&mut self) -> std::result::Result<(), Error> {
         trace!("stop()");
         if !self.is_speaking()? {

src/lib.rs

@@ -14,9 +14,10 @@ use std::collections::HashMap;
 #[cfg(target_os = "macos")]
 use std::ffi::CStr;
 use std::fmt;
+use std::rc::Rc;
 #[cfg(windows)]
 use std::string::FromUtf16Error;
-use std::sync::{Arc, Mutex};
+use std::sync::Mutex;
 use std::{boxed::Box, sync::RwLock};
 
 #[cfg(any(target_os = "macos", target_os = "ios"))]
@@ -162,6 +163,7 @@ unsafe impl Sync for UtteranceId {}
 #[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
 pub struct Features {
     pub is_speaking: bool,
+    pub synthesize: bool,
     pub pitch: bool,
     pub rate: bool,
     pub stop: bool,
@@ -217,6 +219,7 @@ pub trait Backend: Clone {
     fn id(&self) -> Option<BackendId>;
     fn supported_features(&self) -> Features;
     fn speak(&mut self, text: &str, interrupt: bool) -> Result<Option<UtteranceId>, Error>;
+    fn synthesize(&mut self, text: &str) -> Result<Vec<u8>, Error>;
     fn stop(&mut self) -> Result<(), Error>;
     fn min_rate(&self) -> f32;
     fn max_rate(&self) -> f32;
@@ -258,7 +261,7 @@ lazy_static! {
 }
 
 #[derive(Clone)]
-pub struct Tts(Arc<RwLock<Box<dyn Backend>>>);
+pub struct Tts(Rc<RwLock<Box<dyn Backend>>>);
 
 unsafe impl Send for Tts {}
 
@@ -271,18 +274,18 @@ impl Tts {
             #[cfg(target_os = "linux")]
             Backends::SpeechDispatcher => {
                 let tts = backends::SpeechDispatcher::new()?;
-                Ok(Tts(Arc::new(RwLock::new(Box::new(tts)))))
+                Ok(Tts(Rc::new(RwLock::new(Box::new(tts)))))
             }
             #[cfg(target_arch = "wasm32")]
             Backends::Web => {
                 let tts = backends::Web::new()?;
-                Ok(Tts(Arc::new(RwLock::new(Box::new(tts)))))
+                Ok(Tts(Rc::new(RwLock::new(Box::new(tts)))))
             }
             #[cfg(all(windows, feature = "tolk"))]
             Backends::Tolk => {
                 let tts = backends::Tolk::new();
                 if let Some(tts) = tts {
-                    Ok(Tts(Arc::new(RwLock::new(Box::new(tts)))))
+                    Ok(Tts(Rc::new(RwLock::new(Box::new(tts)))))
                 } else {
                     Err(Error::NoneError)
                 }
@@ -290,20 +293,20 @@ impl Tts {
             #[cfg(windows)]
             Backends::WinRt => {
                 let tts = backends::WinRt::new()?;
-                Ok(Tts(Arc::new(RwLock::new(Box::new(tts)))))
+                Ok(Tts(Rc::new(RwLock::new(Box::new(tts)))))
             }
             #[cfg(target_os = "macos")]
-            Backends::AppKit => Ok(Tts(Arc::new(RwLock::new(Box::new(
+            Backends::AppKit => Ok(Tts(Rc::new(RwLock::new(
-                backends::AppKit::new()?
+                Box::new(backends::AppKit::new()?),
-            ))))),
+            )))),
             #[cfg(any(target_os = "macos", target_os = "ios"))]
-            Backends::AvFoundation => Ok(Tts(Arc::new(RwLock::new(Box::new(
+            Backends::AvFoundation => Ok(Tts(Rc::new(RwLock::new(Box::new(
                 backends::AvFoundation::new()?,
             ))))),
             #[cfg(target_os = "android")]
             Backends::Android => {
                 let tts = backends::Android::new()?;
-                Ok(Tts(Arc::new(RwLock::new(Box::new(tts)))))
+                Ok(Tts(Rc::new(RwLock::new(Box::new(tts)))))
             }
         };
         if let Ok(backend) = backend {
@@ -374,6 +377,15 @@ impl Tts {
             .speak(text.into().as_str(), interrupt)
     }
 
+    pub fn synthesize<S: Into<String>>(&mut self, text: S) -> Result<Vec<u8>, Error> {
+        let Features { synthesize, .. } = self.supported_features();
+        if synthesize {
+            self.0.write().unwrap().synthesize(text.into().as_str())
+        } else {
+            Err(Error::UnsupportedFeature)
+        }
+    }
+
     /// Stops current speech.
     pub fn stop(&mut self) -> Result<&Self, Error> {
         let Features { stop, .. } = self.supported_features();
@@ -640,7 +652,7 @@ impl Tts {
 
 impl Drop for Tts {
     fn drop(&mut self) {
-        if Arc::strong_count(&self.0) <= 1 {
+        if Rc::strong_count(&self.0) <= 1 {
             if let Some(id) = self.0.read().unwrap().id() {
                 let mut callbacks = CALLBACKS.lock().unwrap();
                 callbacks.remove(&id);