change: VVMにUUIDを割り振り、それをVoiceModelIdとする (#796)

* change: VVMにUUIDを割り振り、それを`VoiceModelId`とする * 不要となったコメントを削除 * UUIDを分解する部分を修正し、テストも追加以下のためuuidもアップデートする。 <uuid-rs/uuid#741>
VOICEVOX · May 24, 2024 · 484bd62 · 484bd62
1 parent 9c3a94e
commit 484bd62
Show file tree

Hide file tree

Showing 26 changed files with 209 additions and 153 deletions.
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -47,7 +47,6 @@ libc = "0.2.134"
 libloading = "0.7.3"
 libtest-mimic = "0.6.0"
 log = "0.4.17"
-nanoid = "0.4.0"
 ndarray = "0.15.6"
 ndarray-stats = "0.5.1"
 octocrab = { version = "0.19.0", default-features = false }
@@ -81,7 +80,7 @@ tracing = "0.1.37"
 tracing-subscriber = "0.3.16"
 typetag = "0.2.5"
 url = "2.3.0"
-uuid = "1.4.0"
+uuid = "1.6.1"
 voicevox_core = { path = "crates/voicevox_core" }
 windows = "0.43.0"
 zip = "0.6.3"

diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml
@@ -25,7 +25,6 @@ futures.workspace = true
 indexmap = { workspace = true, features = ["serde"] }
 itertools.workspace = true
 jlabel.workspace = true
-nanoid.workspace = true
 ndarray.workspace = true
 once_cell.workspace = true
 open_jtalk.workspace = true

diff --git a/crates/voicevox_core/src/__internal/interop.rs b/crates/voicevox_core/src/__internal/interop.rs
@@ -1 +1,4 @@
-pub use crate::{metas::merge as merge_metas, synthesizer::blocking::PerformInference};
+pub use crate::{
+    metas::merge as merge_metas, synthesizer::blocking::PerformInference,
+    voice_model::blocking::IdRef,
+};
diff --git a/crates/voicevox_core/src/manifest.rs b/crates/voicevox_core/src/manifest.rs
@@ -6,7 +6,7 @@ use derive_new::new;
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DisplayFromStr};
 
-use crate::StyleId;
+use crate::{StyleId, VoiceModelId};
 
 pub type RawManifestVersion = String;
 #[derive(Deserialize, Clone, Debug, PartialEq, new)]
@@ -38,10 +38,9 @@ impl Display for InnerVoiceId {
 
 #[derive(Deserialize, Getters, Clone)]
 pub struct Manifest {
-    // FIXME: UUIDにする
-    // https://github.com/VOICEVOX/voicevox_core/issues/581
     #[allow(dead_code)]
     manifest_version: ManifestVersion,
+    pub(crate) id: VoiceModelId,
     metas_filename: String,
     #[serde(flatten)]
     domains: ManifestDomains,

diff --git a/crates/voicevox_core/src/status.rs b/crates/voicevox_core/src/status.rs
@@ -58,7 +58,7 @@ impl<R: InferenceRuntime> Status<R> {
         Ok(())
     }
 
-    pub(crate) fn unload_model(&self, voice_model_id: &VoiceModelId) -> Result<()> {
+    pub(crate) fn unload_model(&self, voice_model_id: VoiceModelId) -> Result<()> {
         self.loaded_models.lock().unwrap().remove(voice_model_id)
     }
 
@@ -77,7 +77,7 @@ impl<R: InferenceRuntime> Status<R> {
         self.loaded_models.lock().unwrap().ids_for::<D>(style_id)
     }
 
-    pub(crate) fn is_loaded_model(&self, voice_model_id: &VoiceModelId) -> bool {
+    pub(crate) fn is_loaded_model(&self, voice_model_id: VoiceModelId) -> bool {
         self.loaded_models
             .lock()
             .unwrap()
@@ -101,7 +101,7 @@ impl<R: InferenceRuntime> Status<R> {
     /// `self`が`model_id`を含んでいないとき、パニックする。
     pub(crate) fn run_session<I>(
         &self,
-        model_id: &VoiceModelId,
+        model_id: VoiceModelId,
         input: I,
     ) -> Result<<I::Signature as InferenceSignature>::Output>
     where
@@ -159,7 +159,7 @@ impl<R: InferenceRuntime> LoadedModels<R> {
             .and_then(|(inner_voice_ids, _)| inner_voice_ids.get(&style_id).copied())
             .unwrap_or_else(|| InnerVoiceId::new(style_id.raw_id()));
 
-        Ok((model_id.clone(), inner_voice_id))
+        Ok((*model_id, inner_voice_id))
     }
 
     /// # Panics
@@ -168,12 +168,12 @@ impl<R: InferenceRuntime> LoadedModels<R> {
     ///
     /// - `self`が`model_id`を含んでいないとき
     /// - 対応する`InferenceDomain`が欠けているとき
-    fn get<I>(&self, model_id: &VoiceModelId) -> InferenceSessionCell<R, I>
+    fn get<I>(&self, model_id: VoiceModelId) -> InferenceSessionCell<R, I>
     where
         I: InferenceInputSignature,
         <I::Signature as InferenceSignature>::Domain: InferenceDomainExt,
     {
-        let (_, session_set) = self.0[model_id]
+        let (_, session_set) = self.0[&model_id]
             .session_sets_with_inner_ids
             .get::<<I::Signature as InferenceSignature>::Domain>()
             .as_ref()
@@ -190,8 +190,8 @@ impl<R: InferenceRuntime> LoadedModels<R> {
         session_set.get()
     }
 
-    fn contains_voice_model(&self, model_id: &VoiceModelId) -> bool {
-        self.0.contains_key(model_id)
+    fn contains_voice_model(&self, model_id: VoiceModelId) -> bool {
+        self.0.contains_key(&model_id)
     }
 
     fn contains_style(&self, style_id: StyleId) -> bool {
@@ -216,9 +216,9 @@ impl<R: InferenceRuntime> LoadedModels<R> {
             source: None,
         };
 
-        if self.0.contains_key(&model_header.id) {
+        if self.0.contains_key(&model_header.manifest.id) {
             return Err(error(LoadModelErrorKind::ModelAlreadyLoaded {
-                id: model_header.id.clone(),
+                id: model_header.manifest.id,
             }));
         }
 
@@ -255,7 +255,7 @@ impl<R: InferenceRuntime> LoadedModels<R> {
         self.ensure_acceptable(model_header)?;
 
         let prev = self.0.insert(
-            model_header.id.clone(),
+            model_header.manifest.id,
             LoadedModel {
                 metas: model_header.metas.clone(),
                 session_sets_with_inner_ids,
@@ -265,12 +265,9 @@ impl<R: InferenceRuntime> LoadedModels<R> {
         Ok(())
     }
 
-    fn remove(&mut self, model_id: &VoiceModelId) -> Result<()> {
-        if self.0.remove(model_id).is_none() {
-            return Err(ErrorRepr::ModelNotFound {
-                model_id: model_id.clone(),
-            }
-            .into());
+    fn remove(&mut self, model_id: VoiceModelId) -> Result<()> {
+        if self.0.remove(&model_id).is_none() {
+            return Err(ErrorRepr::ModelNotFound { model_id }.into());
         }
         Ok(())
     }
@@ -415,13 +412,13 @@ mod tests {
         let model_header = vvm.header();
         let model_contents = &vvm.read_inference_models().await.unwrap();
         assert!(
-            !status.is_loaded_model(&model_header.id),
+            !status.is_loaded_model(model_header.manifest.id),
             "model should  not be loaded"
         );
         let result = status.insert_model(model_header, model_contents);
         assert_debug_fmt_eq!(Ok(()), result);
         assert!(
-            status.is_loaded_model(&model_header.id),
+            status.is_loaded_model(model_header.manifest.id),
             "model should be loaded",
         );
     }

diff --git a/crates/voicevox_core/src/synthesizer.rs b/crates/voicevox_core/src/synthesizer.rs
@@ -207,12 +207,12 @@ pub(crate) mod blocking {
         }
 
         /// 音声モデルの読み込みを解除する。
-        pub fn unload_voice_model(&self, voice_model_id: &VoiceModelId) -> Result<()> {
+        pub fn unload_voice_model(&self, voice_model_id: VoiceModelId) -> Result<()> {
             self.status.unload_model(voice_model_id)
         }
 
         /// 指定したIDの音声モデルが読み込まれているか判定する。
-        pub fn is_loaded_voice_model(&self, voice_model_id: &VoiceModelId) -> bool {
+        pub fn is_loaded_voice_model(&self, voice_model_id: VoiceModelId) -> bool {
             self.status.is_loaded_model(voice_model_id)
         }
 
@@ -841,7 +841,7 @@ pub(crate) mod blocking {
             let PredictDurationOutput {
                 phoneme_length: output,
             } = self.status.run_session(
-                &model_id,
+                model_id,
                 PredictDurationInput {
                     phoneme_list: ndarray::arr1(phoneme_vector),
                     speaker_id: ndarray::arr1(&[inner_voice_id.raw_id().into()]),
@@ -874,7 +874,7 @@ pub(crate) mod blocking {
             let (model_id, inner_voice_id) = self.status.ids_for::<TalkDomain>(style_id)?;
 
             let PredictIntonationOutput { f0_list: output } = self.status.run_session(
-                &model_id,
+                model_id,
                 PredictIntonationInput {
                     length: ndarray::arr0(length as i64),
                     vowel_phoneme_list: ndarray::arr1(vowel_phoneme_vector),
@@ -917,7 +917,7 @@ pub(crate) mod blocking {
             );
 
             let DecodeOutput { wave: output } = self.status.run_session(
-                &model_id,
+                model_id,
                 DecodeInput {
                     f0: ndarray::arr1(&f0_with_padding)
                         .into_shape([length_with_padding, 1])
@@ -1150,11 +1150,11 @@ pub(crate) mod tokio {
             self.0.status.insert_model(model.header(), model_bytes)
         }
 
-        pub fn unload_voice_model(&self, voice_model_id: &VoiceModelId) -> Result<()> {
+        pub fn unload_voice_model(&self, voice_model_id: VoiceModelId) -> Result<()> {
             self.0.unload_voice_model(voice_model_id)
         }
 
-        pub fn is_loaded_voice_model(&self, voice_model_id: &VoiceModelId) -> bool {
+        pub fn is_loaded_voice_model(&self, voice_model_id: VoiceModelId) -> bool {
             self.0.is_loaded_voice_model(voice_model_id)
         }