Merge branch 'main' into add-morphing

.github/workflows/download_test.yml

@@ -7,14 +7,16 @@ on:
         description: "テスト対象のコアのバージョン。無指定時はprerelease込みの最新release。"
         type: string
         required: false
-  push:
-    branches:
-      - main
-  pull_request:
-    paths:
-      - "Cargo.*"
-      - "crates/downloader/**"
-      - ".github/workflows/download_test.yml"
+  # 新しいビルドができるまで、このworkflowが壊れて動かないことを許容する
+  # https://github.com/VOICEVOX/voicevox_core/issues/741#issuecomment-1935303742
+  #push:
+  #  branches:
+  #    - main
+  #pull_request:
+  #  paths:
+  #    - "Cargo.*"
+  #    - "crates/downloader/**"
+  #    - ".github/workflows/download_test.yml"
 
 env:
   VERSION: ${{ inputs.version || 'prerelease-latest' }}

crates/voicevox_core/src/__internal/interop.rs

@@ -1 +1 @@
-pub use crate::synthesizer::blocking::PerformInference;
+pub use crate::{metas::merge as merge_metas, synthesizer::blocking::PerformInference};

crates/voicevox_core/src/engine/open_jtalk.rs

@@ -128,9 +128,12 @@ pub(crate) mod blocking {
     }
 
     impl Inner {
-        // FIXME: 中断可能にする
+        // TODO: 中断可能にする
         pub(super) fn use_user_dict(&self, words: &str) -> crate::result::Result<()> {
-            let result = {
+            // 空の辞書を読み込もうとするとクラッシュするのでユーザー辞書なしでロード
+            if words.is_empty() {
+                self.load_with_userdic(None)
+            } else {
                 // 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()))?;
@@ -158,17 +161,17 @@ pub(crate) mod blocking {
                     "-q",
                 ]);
 
-                let Resources { mecab, .. } = &mut *self.resources.lock().unwrap();
+                self.load_with_userdic(Some(temp_dict_path.as_ref()))
+            }
+        }
+        fn load_with_userdic(&self, dict_path: Option<&Path>) -> crate::result::Result<()> {
+            let Resources { mecab, .. } = &mut *self.resources.lock().unwrap();
 
-                mecab.load_with_userdic(self.dict_dir.as_ref(), Some(Path::new(&temp_dict_path)))
-            };
+            let result = mecab.load_with_userdic(self.dict_dir.as_ref(), 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(),
-                );
+                return Err(ErrorRepr::UseUserDict(anyhow!("辞書を読み込めませんでした。")).into());
             }
-
             Ok(())
         }
     }

crates/voicevox_core/src/infer/status.rs

@@ -8,13 +8,14 @@ use std::{
 use anyhow::bail;
 use educe::Educe;
 use enum_map::{Enum as _, EnumMap};
+use indexmap::IndexMap;
 use itertools::{iproduct, Itertools as _};
 
 use crate::{
     error::{ErrorRepr, LoadModelError, LoadModelErrorKind, LoadModelResult},
     infer::{InferenceOperation, ParamInfo},
     manifest::ModelInnerId,
-    metas::{SpeakerMeta, StyleId, StyleMeta, VoiceModelMeta},
+    metas::{self, SpeakerMeta, StyleId, StyleMeta, VoiceModelMeta},
     voice_model::{VoiceModelHeader, VoiceModelId},
     Result,
 };
@@ -119,7 +120,7 @@ impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
 #[derive(Educe)]
 #[educe(Default(bound = "R: InferenceRuntime, D: InferenceDomain"))]
 struct LoadedModels<R: InferenceRuntime, D: InferenceDomain>(
-    BTreeMap<VoiceModelId, LoadedModel<R, D>>,
+    IndexMap<VoiceModelId, LoadedModel<R, D>>,
 );
 
 struct LoadedModel<R: InferenceRuntime, D: InferenceDomain> {
@@ -130,11 +131,7 @@ struct LoadedModel<R: InferenceRuntime, D: InferenceDomain> {
 
 impl<R: InferenceRuntime, D: InferenceDomain> LoadedModels<R, D> {
     fn metas(&self) -> VoiceModelMeta {
-        self.0
-            .values()
-            .flat_map(|LoadedModel { metas, .. }| metas)
-            .cloned()
-            .collect()
+        metas::merge(self.0.values().flat_map(|LoadedModel { metas, .. }| metas))
     }
 
     fn ids_for(&self, style_id: StyleId) -> Result<(VoiceModelId, ModelInnerId)> {
@@ -184,20 +181,29 @@ impl<R: InferenceRuntime, D: InferenceDomain> LoadedModels<R, D> {
     ///
     /// # Errors
     ///
-    /// 音声モデルIDかスタイルIDが`model_header`と重複するとき、エラーを返す。
+    /// 次の場合にエラーを返す。
+    ///
+    /// - 音声モデルIDかスタイルIDが`model_header`と重複するとき
     fn ensure_acceptable(&self, model_header: &VoiceModelHeader) -> LoadModelResult<()> {
-        let loaded = self.styles();
-        let external = model_header
-            .metas
-            .iter()
-            .flat_map(|speaker| speaker.styles());
-
         let error = |context| LoadModelError {
             path: model_header.path.clone(),
             context,
             source: None,
         };
 
+        let loaded = self.speakers();
+        let external = model_header.metas.iter();
+        for (loaded, external) in iproduct!(loaded, external) {
+            if loaded.speaker_uuid() == external.speaker_uuid() {
+                loaded.warn_diff_except_styles(external);
+            }
+        }
+
+        let loaded = self.styles();
+        let external = model_header
+            .metas
+            .iter()
+            .flat_map(|speaker| speaker.styles());
         if self.0.contains_key(&model_header.id) {
             return Err(error(LoadModelErrorKind::ModelAlreadyLoaded {
                 id: model_header.id.clone(),
@@ -242,11 +248,12 @@ impl<R: InferenceRuntime, D: InferenceDomain> LoadedModels<R, D> {
         Ok(())
     }
 
+    fn speakers(&self) -> impl Iterator<Item = &SpeakerMeta> + Clone {
+        self.0.values().flat_map(|LoadedModel { metas, .. }| metas)
+    }
+
     fn styles(&self) -> impl Iterator<Item = &StyleMeta> {
-        self.0
-            .values()
-            .flat_map(|LoadedModel { metas, .. }| metas)
-            .flat_map(|speaker| speaker.styles())
+        self.speakers().flat_map(|speaker| speaker.styles())
     }
 }
 

crates/voicevox_core/src/metas.rs

@@ -1,8 +1,42 @@
-use std::fmt::Display;
+use std::fmt::{Debug, Display};
 
 use derive_getters::Getters;
 use derive_new::new;
+use indexmap::IndexMap;
+use itertools::Itertools as _;
 use serde::{Deserialize, Serialize};
+use tracing::warn;
+
+/// [`speaker_uuid`]をキーとして複数の[`SpeakerMeta`]をマージする。
+///
+/// マージする際話者は[`SpeakerMeta::order`]、スタイルは[`StyleMeta::order`]をもとに安定ソートされる。
+/// `order`が無い話者とスタイルは、そうでないものよりも後ろに置かれる。
+///
+/// [`speaker_uuid`]: SpeakerMeta::speaker_uuid
+pub fn merge<'a>(metas: impl IntoIterator<Item = &'a SpeakerMeta>) -> Vec<SpeakerMeta> {
+    return metas
+        .into_iter()
+        .fold(IndexMap::<_, SpeakerMeta>::new(), |mut acc, speaker| {
+            acc.entry(&speaker.speaker_uuid)
+                .and_modify(|acc| acc.styles.extend(speaker.styles.clone()))
+                .or_insert_with(|| speaker.clone());
+            acc
+        })
+        .into_values()
+        .update(|speaker| {
+            speaker
+                .styles
+                .sort_by_key(|&StyleMeta { order, .. }| key(order));
+        })
+        .sorted_by_key(|&SpeakerMeta { order, .. }| key(order))
+        .collect();
+
+    fn key(order: Option<u32>) -> impl Ord {
+        order
+            .map(Into::into)
+            .unwrap_or_else(|| u64::from(u32::MAX) + 1)
+    }
+}
 
 /// [`StyleId`]の実体。
 ///
@@ -15,7 +49,7 @@ pub type RawStyleId = u32;
 ///
 /// [**話者**(_speaker_)]: SpeakerMeta
 /// [**スタイル**(_style_)]: StyleMeta
-#[derive(PartialEq, Eq, Clone, Copy, Ord, PartialOrd, Hash, Deserialize, Serialize, new, Debug)]
+#[derive(PartialEq, Eq, Clone, Copy, Ord, Hash, PartialOrd, Deserialize, Serialize, new, Debug)]
 pub struct StyleId(RawStyleId);
 
 impl StyleId {
@@ -68,6 +102,60 @@ pub struct SpeakerMeta {
     /// 話者の対応機能。
     #[serde(default)]
     supported_features: SpeakerSupportedFeatures,
+    /// 話者の順番。
+    ///
+    /// `SpeakerMeta`の列は、この値に対して昇順に並んでいるべきである。
+    order: Option<u32>,
+}
+
+impl SpeakerMeta {
+    /// # Panics
+    ///
+    /// `speaker_uuid`が異なるときパニックする。
+    pub(crate) fn warn_diff_except_styles(&self, other: &Self) {
+        let Self {
+            name: name1,
+            styles: _,
+            version: version1,
+            speaker_uuid: speaker_uuid1,
+            supported_features: supported_features1,
+            order: order1,
+        } = self;
+
+        let Self {
+            name: name2,
+            styles: _,
+            version: version2,
+            speaker_uuid: speaker_uuid2,
+            supported_features: supported_features2,
+            order: order2,
+        } = other;
+
+        if speaker_uuid1 != speaker_uuid2 {
+            panic!("must be equal: {speaker_uuid1} != {speaker_uuid2:?}");
+        }
+
+        warn_diff(speaker_uuid1, "name", name1, name2);
+        warn_diff(speaker_uuid1, "version", version1, version2);
+        warn_diff(
+            speaker_uuid1,
+            "supported_features",
+            supported_features1,
+            supported_features2,
+        );
+        warn_diff(speaker_uuid1, "order", order1, order2);
+
+        fn warn_diff<T: PartialEq + Debug>(
+            speaker_uuid: &str,
+            field_name: &str,
+            left: &T,
+            right: &T,
+        ) {
+            if left != right {
+                warn!("`{speaker_uuid}`: different `{field_name}` ({left:?} != {right:?})");
+            }
+        }
+    }
 }
 
 /// **スタイル**(_style_)のメタ情報。
@@ -77,9 +165,118 @@ pub struct StyleMeta {
     id: StyleId,
     /// スタイル名。
     name: String,
+    /// スタイルの順番。
+    ///
+    /// [`SpeakerMeta::styles`]は、この値に対して昇順に並んでいるべきである。
+    order: Option<u32>,
+}
+
+#[cfg(test)]
+mod tests {
+    use once_cell::sync::Lazy;
+    use serde_json::json;
+
+    #[test]
+    fn merge_works() -> anyhow::Result<()> {
+        static INPUT: Lazy<serde_json::Value> = Lazy::new(|| {
+            json!([
+                {
+                    "name": "B",
+                    "styles": [
+                        {
+                            "id": 3,
+                            "name": "B_1",
+                            "order": 0
+                        }
+                    ],
+                    "version": "0.0.0",
+                    "speaker_uuid": "f34ab151-c0f5-4e0a-9ad2-51ce30dba24d",
+                    "order": 1
+                },
+                {
+                    "name": "A",
+                    "styles": [
+                        {
+                            "id": 2,
+                            "name": "A_3",
+                            "order": 2
+                        }
+                    ],
+                    "version": "0.0.0",
+                    "speaker_uuid": "d6fd707c-a451-48e9-8f00-fe9ee3bf6264",
+                    "order": 0
+                },
+                {
+                    "name": "A",
+                    "styles": [
+                        {
+                            "id": 1,
+                            "name": "A_1",
+                            "order": 0
+                        },
+                        {
+                            "id": 0,
+                            "name": "A_2",
+                            "order": 1
+                        }
+                    ],
+                    "version": "0.0.0",
+                    "speaker_uuid": "d6fd707c-a451-48e9-8f00-fe9ee3bf6264",
+                    "order": 0
+                }
+            ])
+        });
+
+        static EXPECTED: Lazy<serde_json::Value> = Lazy::new(|| {
+            json!([
+                {
+                    "name": "A",
+                    "styles": [
+                        {
+                            "id": 1,
+                            "name": "A_1",
+                            "order": 0
+                        },
+                        {
+                            "id": 0,
+                            "name": "A_2",
+                            "order": 1
+                        },
+                        {
+                            "id": 2,
+                            "name": "A_3",
+                            "order": 2
+                        }
+                    ],
+                    "version": "0.0.0",
+                    "speaker_uuid": "d6fd707c-a451-48e9-8f00-fe9ee3bf6264",
+                    "order": 0
+                },
+                {
+                    "name": "B",
+                    "styles": [
+                        {
+                            "id": 3,
+                            "name": "B_1",
+                            "order": 0
+                        }
+                    ],
+                    "version": "0.0.0",
+                    "speaker_uuid": "f34ab151-c0f5-4e0a-9ad2-51ce30dba24d",
+                    "order": 1
+                }
+            ])
+        });
+
+        let input = &serde_json::from_value::<Vec<_>>(INPUT.clone())?;
+        let actual = serde_json::to_value(super::merge(input))?;
+
+        pretty_assertions::assert_eq!(*EXPECTED, actual);
+        Ok(())
+    }
 }
 
-#[derive(Default, Deserialize, Serialize, Clone)]
+#[derive(Default, Deserialize, Serialize, Clone, PartialEq, Debug)]
 pub struct SpeakerSupportedFeatures {
     #[serde(default)]
     pub(crate) permitted_synthesis_morphing: PermittedSynthesisMorphing,