[WIP] エンジンのモック作成＋それを使ったコンポーネントテスト

package.json

@@ -87,6 +87,7 @@
     "@types/async-lock": "1.4.0",
     "@types/encoding-japanese": "1.0.18",
     "@types/glob": "8.0.0",
+    "@types/kuromoji": "0.1.3",
     "@types/markdown-it": "12.2.0",
     "@types/multistream": "4.1.0",
     "@types/semver": "7.3.9",
@@ -114,6 +115,7 @@
     "eslint-plugin-storybook": "0.8.0",
     "eslint-plugin-vue": "9.26.0",
     "happy-dom": "8.4.2",
+    "kuromoji": "0.1.2",
     "license-checker-rseidelsohn": "4.3.0",
     "markdownlint-cli": "0.37.0",
     "patch-package": "8.0.0",

src/backend/electron/manager/RuntimeInfoManager.ts

@@ -8,6 +8,7 @@ import AsyncLock from "async-lock";
 import log from "electron-log/main";
 import type { AltPortInfos } from "@/store/type";
 import { EngineId, EngineInfo } from "@/type/preload";
+import { convertToUrlString } from "@/domain/url";
 
 /**
  * ランタイム情報書き出しに必要なEngineInfo
@@ -85,13 +86,14 @@ export class RuntimeInfoManager {
           const altPort: string | undefined =
             this.altportInfos[engineInfo.uuid];
           const port = altPort ?? engineInfo.defaultPort;
-          // NOTE: URLを正規化する
-          const url = new URL(`${engineInfo.protocol}//${engineInfo.hostname}`);
-          url.port = port;
           return {
             uuid: engineInfo.uuid,
-            // NOTE: URLインターフェースは"pathname"が空文字でも"/"を付けるので手動で結合する。
-            url: `${url.origin}${engineInfo.pathname}`,
+            url: convertToUrlString({
+              protocol: engineInfo.protocol,
+              hostname: engineInfo.hostname,
+              port,
+              pathname: engineInfo.pathname,
+            }),
             name: engineInfo.name,
           };
         }),

src/domain/japanese/index.ts

@@ -40,3 +40,18 @@ export const convertLongVowel = (text: string): string => {
     .replace(/(?<=[ん]ー*)ー/g, "ん")
     .replace(/(?<=[っ]ー*)ー/g, "っ");
 };
+
+// 参考：https://github.com/VOICEVOX/voicevox_core/blob/0848630d81ae3e917c6ff2038f0b15bbd4270702/crates/voicevox_core/src/user_dict/word.rs#L83-L90
+export const moraPattern = new RegExp(
+  "(?:" +
+    "[イ][ェ]|[ヴ][ャュョ]|[トド][ゥ]|[テデ][ィャュョ]|[デ][ェ]|[クグ][ヮ]|" + // rule_others
+    "[キシチニヒミリギジビピ][ェャュョ]|" + // rule_line_i
+    "[ツフヴ][ァ]|[ウスツフヴズ][ィ]|[ウツフヴ][ェォ]|" + // rule_line_u
+    "[ァ-ヴー]|" + // rule_one_mora
+    "[い][ぇ]|[ゃゅょ]|[とど][ぅ]|[てで][ぃゃゅょ]|[で][ぇ]|[くぐ][ゎ]|" + // rule_others
+    "[きしちにひみりぎじびぴ][ぇゃゅょ]|" + // rule_line_i
+    "[つふゔ][ぁ]|[うすつふゔず][ぃ]|[うつふゔ][ぇぉ]|" + // rule_line_u
+    "[ぁ-ゔー]" + // rule_one_mora
+    ")",
+  "g",
+);

src/domain/url.ts

@@ -0,0 +1,17 @@
+export type EngineUrlParams = {
+  protocol: string; // `http:`など
+  hostname: string; // `example.com`など
+  port: string; // `50021`など。空文字列もありえる。
+  pathname: string; // `/engine`など。空文字列もありえる。
+};
+
+/**
+ * URLを構成するパラメータから、VOICEVOXエディタが初期から想定しているURL文字列を組み立てる。
+ * pathnameが空文字の場合は末尾にスラッシュを付与しない、などがビルトインのURLクラスと異なる。
+ */
+export function convertToUrlString(params: EngineUrlParams) {
+  const { protocol, hostname, port, pathname } = params;
+  const url = new URL(`${protocol}//${hostname}`); // NOTE: URLを正規化する
+  url.port = port;
+  return `${url.origin}${pathname}`; // NOTE: URLインターフェースは"pathname"が空文字でも"/"を付けるので手動で結合する。
+}

src/sing/convertToWavFileData.ts → src/helpers/convertToWavFileData.ts

@@ -1,4 +1,13 @@
-export const convertToWavFileData = (audioBuffer: AudioBuffer) => {
+export const convertToWavFileData = (
+  audioBuffer:
+    | AudioBuffer
+    | {
+        sampleRate: number;
+        length: number;
+        numberOfChannels: number;
+        getChannelData(channel: number): Float32Array;
+      },
+) => {
   const bytesPerSample = 4; // Float32
   const formatCode = 3; // WAVE_FORMAT_IEEE_FLOAT
 

src/mock/engineMock/audioQueryMock.ts

@@ -0,0 +1,195 @@
+/**
+ * AudioQueryとFrameAudioQueryのモック。
+ * VOICEVOX ENGINEリポジトリの処理とほぼ同じ。
+ */
+
+import { AccentPhrase, AudioQuery, FrameAudioQuery, Mora } from "@/openapi";
+
+function generateSilenceMora(length: number): Mora {
+  return {
+    text: "　",
+    vowel: "sil",
+    vowelLength: length,
+    pitch: 0.0,
+  };
+}
+
+function toFlattenMoras(accentPhrases: AccentPhrase[]): Mora[] {
+  let moras: Mora[] = [];
+  accentPhrases.forEach((accentPhrase) => {
+    moras = moras.concat(accentPhrase.moras);
+    if (accentPhrase.pauseMora) {
+      moras.push(accentPhrase.pauseMora);
+    }
+  });
+  return moras;
+}
+
+function toFlattenPhonemes(moras: Mora[]): string[] {
+  const phonemes: string[] = [];
+  for (const mora of moras) {
+    if (mora.consonant) {
+      phonemes.push(mora.consonant);
+    }
+    phonemes.push(mora.vowel);
+  }
+  return phonemes;
+}
+
+/** 前後の無音モーラを追加する */
+function applyPrePostSilence(moras: Mora[], query: AudioQuery): Mora[] {
+  const preSilenceMoras = [generateSilenceMora(query.prePhonemeLength)];
+  const postSilenceMoras = [generateSilenceMora(query.postPhonemeLength)];
+  return preSilenceMoras.concat(moras).concat(postSilenceMoras);
+}
+
+/** 無音時間を置き換える */
+function applyPauseLength(moras: Mora[], query: AudioQuery): Mora[] {
+  if (query.pauseLength != undefined) {
+    for (const mora of moras) {
+      if (mora.vowel == "pau") {
+        mora.vowelLength = query.pauseLength;
+      }
+    }
+  }
+  return moras;
+}
+
+/** 無音時間スケールを適用する */
+function applyPauseLengthScale(moras: Mora[], query: AudioQuery): Mora[] {
+  if (query.pauseLengthScale != undefined) {
+    for (const mora of moras) {
+      if (mora.vowel == "pau") {
+        mora.vowelLength *= query.pauseLengthScale;
+      }
+    }
+  }
+  return moras;
+}
+
+/** 話速スケールを適用する */
+function applySpeedScale(moras: Mora[], query: AudioQuery): Mora[] {
+  for (const mora of moras) {
+    mora.vowelLength /= query.speedScale;
+    if (mora.consonantLength) {
+      mora.consonantLength /= query.speedScale;
+    }
+  }
+  return moras;
+}
+
+/** 音高スケールを適用する */
+function applyPitchScale(moras: Mora[], query: AudioQuery): Mora[] {
+  for (const mora of moras) {
+    mora.pitch *= 2 ** query.pitchScale;
+  }
+  return moras;
+}
+
+/** 抑揚スケールを適用する */
+function applyIntonationScale(moras: Mora[], query: AudioQuery): Mora[] {
+  const voiced = moras.filter((mora) => mora.pitch > 0);
+  if (voiced.length == 0) {
+    return moras;
+  }
+
+  const meanF0 =
+    voiced.reduce((sum, mora) => sum + mora.pitch, 0) / voiced.length;
+  for (const mora of voiced) {
+    mora.pitch = (mora.pitch - meanF0) * query.intonationScale + meanF0;
+  }
+  return moras;
+}
+
+/** 疑問文の最後に音高の高いモーラを追加する */
+function applyInterrogativeUpspeak(accentPhrases: Array<AccentPhrase>) {
+  accentPhrases.forEach((accentPhrase) => {
+    const moras = accentPhrase.moras;
+    if (
+      moras.length > 0 &&
+      accentPhrase.isInterrogative &&
+      moras[moras.length - 1].pitch > 0
+    ) {
+      const lastMora = moras[moras.length - 1];
+      const upspeakMora: Mora = {
+        text: "ー",
+        vowel: lastMora.vowel,
+        vowelLength: 0.15,
+        pitch: lastMora.pitch + 0.3,
+      };
+      accentPhrase.moras.push(upspeakMora);
+    }
+  });
+}
+
+function secondToFrame(second: number): number {
+  const FRAME_RATE = 24000 / 256;
+  return Math.round(second * FRAME_RATE);
+}
+
+/** モーラや音素ごとのフレーム数を数える */
+function countFramePerUnit(moras: Mora[]): {
+  framePerPhoneme: number[];
+  framePerMora: number[];
+} {
+  const framePerPhoneme: number[] = [];
+  const framePerMora: number[] = [];
+
+  for (const mora of moras) {
+    const vowelFrames = secondToFrame(mora.vowelLength);
+    const consonantFrames = mora.consonantLength
+      ? secondToFrame(mora.consonantLength)
+      : 0;
+    const moraFrames = vowelFrames + consonantFrames;
+
+    if (mora.consonant) {
+      framePerPhoneme.push(consonantFrames);
+    }
+    framePerPhoneme.push(vowelFrames);
+    framePerMora.push(moraFrames);
+  }
+
+  return { framePerPhoneme, framePerMora };
+}
+
+/** AudioQueryを適当にFrameAudioQueryに変換する */
+export function audioQueryToFrameAudioQueryMock(
+  audioQuery: AudioQuery,
+  { enableInterrogativeUpspeak }: { enableInterrogativeUpspeak: boolean },
+): FrameAudioQuery {
+  const accentPhrases = audioQuery.accentPhrases;
+
+  if (enableInterrogativeUpspeak) {
+    applyInterrogativeUpspeak(accentPhrases);
+  }
+
+  let moras = toFlattenMoras(accentPhrases);
+  moras = applyPrePostSilence(moras, audioQuery);
+  moras = applyPauseLength(moras, audioQuery);
+  moras = applyPauseLengthScale(moras, audioQuery);
+  moras = applySpeedScale(moras, audioQuery);
+  moras = applyPitchScale(moras, audioQuery);
+  moras = applyIntonationScale(moras, audioQuery);
+
+  const { framePerPhoneme, framePerMora } = countFramePerUnit(moras);
+
+  const f0 = moras.flatMap((mora, i) =>
+    Array<number>(framePerMora[i]).fill(
+      mora.pitch == 0 ? 0 : Math.exp(mora.pitch),
+    ),
+  );
+  const volume = Array<number>(f0.length).fill(1.0);
+  const phonemes = toFlattenPhonemes(moras).map((phoneme, i) => ({
+    phoneme,
+    frameLength: framePerPhoneme[i],
+  }));
+
+  return {
+    f0,
+    volume,
+    phonemes,
+    volumeScale: audioQuery.volumeScale,
+    outputSamplingRate: audioQuery.outputSamplingRate,
+    outputStereo: audioQuery.outputStereo,
+  };
+}

src/mock/engineMock/constants.ts

@@ -0,0 +1,6 @@
+/**
+ * 外部と内部で利用する定数の一覧。
+ */
+
+export const dicPath = "node_modules/kuromoji/dict";
+export const assetsPath = "tests/assets";