Synthesizerの構造改革をする

VOICEVOX · Nov 18, 2023 · 04b3787 · 04b3787
1 parent ae4a45d
commit 04b3787
Show file tree

Hide file tree

Showing 25 changed files with 1,153 additions and 1,087 deletions.
diff --git a/crates/voicevox_core/src/__internal.rs b/crates/voicevox_core/src/__internal.rs
@@ -1,4 +1,5 @@
 pub mod doctest_fixtures;
+pub mod interp;
 
 // VOICEVOX CORE内のラッパー向けの実装
 // FIXME: 要議論: https://github.com/VOICEVOX/voicevox_core/issues/595

diff --git a/crates/voicevox_core/src/__internal/doctest_fixtures.rs b/crates/voicevox_core/src/__internal/doctest_fixtures.rs
@@ -4,7 +4,7 @@ use crate::{AccelerationMode, InitializeOptions, OpenJtalk, Synthesizer, VoiceMo
 
 pub async fn synthesizer_with_sample_voice_model(
     open_jtalk_dic_dir: impl AsRef<Path>,
-) -> anyhow::Result<Synthesizer> {
+) -> anyhow::Result<Synthesizer<Arc<OpenJtalk>>> {
     let syntesizer = Synthesizer::new(
         Arc::new(OpenJtalk::new(open_jtalk_dic_dir).unwrap()),
         &InitializeOptions {

diff --git a/crates/voicevox_core/src/__internal/interp.rs b/crates/voicevox_core/src/__internal/interp.rs
@@ -0,0 +1,46 @@
+use easy_ext::ext;
+use ndarray::{Array1, ArrayView1, ArrayView2};
+
+use crate::{StyleId, Synthesizer};
+
+#[ext(PerformInference)]
+impl Synthesizer<()> {
+    pub fn predict_duration(
+        &self,
+        phoneme_list: Array1<i64>,
+        style_id: StyleId,
+    ) -> crate::Result<Vec<f32>> {
+        self.predict_duration(phoneme_list, style_id)
+    }
+
+    #[allow(clippy::too_many_arguments)]
+    pub fn predict_intonation(
+        &self,
+        vowel_phoneme_list: Array1<i64>,
+        consonant_phoneme_list: Array1<i64>,
+        start_accent_list: Array1<i64>,
+        end_accent_list: Array1<i64>,
+        start_accent_phrase_list: Array1<i64>,
+        end_accent_phrase_list: Array1<i64>,
+        style_id: StyleId,
+    ) -> crate::Result<Vec<f32>> {
+        self.predict_intonation(
+            vowel_phoneme_list,
+            consonant_phoneme_list,
+            start_accent_list,
+            end_accent_list,
+            start_accent_phrase_list,
+            end_accent_phrase_list,
+            style_id,
+        )
+    }
+
+    pub fn decode(
+        &self,
+        f0: ArrayView1<'_, f32>,
+        phoneme: ArrayView2<'_, f32>,
+        style_id: StyleId,
+    ) -> crate::Result<Vec<f32>> {
+        self.decode(f0, phoneme, style_id)
+    }
+}
diff --git a/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs b/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs
@@ -4,7 +4,7 @@ use once_cell::sync::Lazy;
 use std::collections::HashMap;
 
 #[rustfmt::skip]
-const PHONEME_LIST: &[&str] = &[
+const PHONEME_LIST: [&str; 45] = [
     "pau",
     "A",
     "E",
@@ -70,9 +70,7 @@ pub struct OjtPhoneme {
 }
 
 impl OjtPhoneme {
-    pub fn num_phoneme() -> usize {
-        PHONEME_MAP.len()
-    }
+    pub(crate) const NUM_PHONEME: usize = PHONEME_LIST.len();
 
     pub fn space_phoneme() -> String {
         "pau".into()
@@ -134,8 +132,8 @@ mod tests {
     }
 
     #[rstest]
-    fn test_num_phoneme_works() {
-        assert_eq!(OjtPhoneme::num_phoneme(), 45);
+    fn test_phoneme_map_has_enough_elements() {
+        assert_eq!(OjtPhoneme::NUM_PHONEME, PHONEME_MAP.len());
     }
 
     #[rstest]

diff --git a/crates/voicevox_core/src/engine/mod.rs b/crates/voicevox_core/src/engine/mod.rs
@@ -4,13 +4,12 @@ mod kana_parser;
 mod model;
 mod mora_list;
 mod open_jtalk;
-mod synthesis_engine;
 
 use super::*;
 
 pub use self::acoustic_feature_extractor::*;
 pub use self::full_context_label::*;
 pub use self::kana_parser::*;
 pub use self::model::*;
+pub(crate) use self::mora_list::mora2text;
 pub use self::open_jtalk::OpenJtalk;
-pub use self::synthesis_engine::*;
diff --git a/crates/voicevox_core/src/engine/open_jtalk.rs b/crates/voicevox_core/src/engine/open_jtalk.rs
@@ -1,8 +1,5 @@
 use std::io::Write;
-use std::{
-    path::{Path, PathBuf},
-    sync::Mutex,
-};
+use std::{path::Path, sync::Mutex};
 
 use anyhow::anyhow;
 use tempfile::NamedTempFile;
@@ -22,7 +19,7 @@ pub(crate) struct OpenjtalkFunctionError {
 /// テキスト解析器としてのOpen JTalk。
 pub struct OpenJtalk {
     resources: Mutex<Resources>,
-    dict_dir: Option<PathBuf>,
+    dict_dir: String,
 }
 
 struct Resources {
@@ -35,37 +32,34 @@ struct Resources {
 unsafe impl Send for Resources {}
 
 impl OpenJtalk {
-    // FIXME: この関数は廃止し、`Synthesizer`は`Option<OpenJtalk>`という形でこの構造体を持つ
-    pub fn new_without_dic() -> Self {
-        Self {
-            resources: Mutex::new(Resources {
-                mecab: ManagedResource::initialize(),
-                njd: ManagedResource::initialize(),
-                jpcommon: ManagedResource::initialize(),
-            }),
-            dict_dir: None,
-        }
-    }
     pub fn new(open_jtalk_dict_dir: impl AsRef<Path>) -> crate::result::Result<Self> {
-        let mut s = Self::new_without_dic();
-        s.load(open_jtalk_dict_dir).map_err(|()| {
-            // FIXME: 「システム辞書を読もうとしたけど読めなかった」というエラーをちゃんと用意する
-            ErrorRepr::NotLoadedOpenjtalkDict
-        })?;
-        Ok(s)
+        let mut resources = Resources {
+            mecab: ManagedResource::initialize(),
+            njd: ManagedResource::initialize(),
+            jpcommon: ManagedResource::initialize(),
+        };
+        let dict_dir = open_jtalk_dict_dir
+            .as_ref()
+            .to_str()
+            .unwrap_or_else(|| todo!("Rust APIでは`Utf8Path`で受けるようにする"))
+            .to_owned();
+
+        let result = resources.mecab.load(&dict_dir);
+        if !result {
+            return Err(ErrorRepr::LoadOpenjtalkSystemDic(dict_dir).into());
+        }
+
+        Ok(Self {
+            resources: resources.into(),
+            dict_dir,
+        })
     }
 
     // 先に`load`を呼ぶ必要がある。
     /// ユーザー辞書を設定する。
     ///
     /// この関数を呼び出した後にユーザー辞書を変更した場合は、再度この関数を呼ぶ必要がある。
     pub fn use_user_dict(&self, user_dict: &UserDict) -> crate::result::Result<()> {
-        let dict_dir = self
-            .dict_dir
-            .as_ref()
-            .and_then(|dict_dir| dict_dir.to_str())
-            .ok_or(ErrorRepr::NotLoadedOpenjtalkDict)?;
-
         // ユーザー辞書用のcsvを作成
         let mut temp_csv = NamedTempFile::new().map_err(|e| ErrorRepr::UseUserDict(e.into()))?;
         temp_csv
@@ -80,7 +74,7 @@ impl OpenJtalk {
         mecab_dict_index(&[
             "mecab-dict-index",
             "-d",
-            dict_dir,
+            &self.dict_dir,
             "-u",
             temp_dict_path.to_str().unwrap(),
             "-f",
@@ -93,7 +87,8 @@ impl OpenJtalk {
 
         let Resources { mecab, .. } = &mut *self.resources.lock().unwrap();
 
-        let result = mecab.load_with_userdic(Path::new(dict_dir), Some(Path::new(&temp_dict_path)));
+        let result =
+            mecab.load_with_userdic(self.dict_dir.as_ref(), Some(Path::new(&temp_dict_path)));
 
         if !result {
             return Err(ErrorRepr::UseUserDict(anyhow!("辞書のコンパイルに失敗しました")).into());
@@ -150,26 +145,6 @@ impl OpenJtalk {
             })
         }
     }
-
-    fn load(&mut self, open_jtalk_dict_dir: impl AsRef<Path>) -> std::result::Result<(), ()> {
-        let result = self
-            .resources
-            .lock()
-            .unwrap()
-            .mecab
-            .load(open_jtalk_dict_dir.as_ref());
-        if result {
-            self.dict_dir = Some(open_jtalk_dict_dir.as_ref().into());
-            Ok(())
-        } else {
-            self.dict_dir = None;
-            Err(())
-        }
-    }
-
-    pub fn dict_loaded(&self) -> bool {
-        self.dict_dir.is_some()
-    }
 }
 
 #[cfg(test)]