Optimization: Startup, Runtime

Lazy initialize Frontend parts to enable faster startup time. This will delay generation of first synthesis, though. Cache creation of compiled regex patterns when the UserDict entries change. This makes execution of User Dictionary replacements as good as unnoticable. Additionally: - Fix replacement pattern for UserDictionary entry terms. - Replace NormDict => UserDict - Disable strict mode policies - Add somoe TODO comments Signed-off-by: Daniel Schnell <[email protected]>
grammatek · Mar 22, 2024 · 131289f · 131289f
1 parent f9b1673
commit 131289f
Show file tree

Hide file tree

Showing 12 changed files with 130 additions and 67 deletions.
diff --git a/app/src/main/java/com/grammatek/simaromur/AppRepository.java b/app/src/main/java/com/grammatek/simaromur/AppRepository.java
@@ -46,12 +46,14 @@
 import java.io.InputStream;
 import java.util.ArrayList;
 import java.util.Date;
+import java.util.HashMap;
 import java.util.List;
 import java.util.Objects;
 import java.util.concurrent.Executors;
 import java.util.concurrent.LinkedBlockingQueue;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.TimeUnit;
+import java.util.regex.Pattern;
 
 /**
  * Abstracted application repository as promoted by the Architecture Guide.
@@ -66,8 +68,10 @@ public class AppRepository {
     private final NormDictEntryDao mNormDictDao;
     private LiveData<AppData> mAppData;
     private LiveData<List<com.grammatek.simaromur.db.Voice>> mAllVoices;
+    private LiveData<List<com.grammatek.simaromur.db.NormDictEntry>> mAllUserDictEntries;
     private AppData mCachedAppData;
     private List<com.grammatek.simaromur.db.Voice> mAllCachedVoices;
+    private HashMap<NormDictEntry, Pattern> mAllCachedUserDictEntries;
     private final VoiceController mNetworkVoiceController;
     private final SpeakController mNetworkSpeakController;
     private final ApiDbUtil mApiDbUtil;
@@ -232,6 +236,9 @@ public AppRepository(Application application) throws IOException {
         });
         mAllVoices = mVoiceDao.getAllVoices();
         mAllVoices.observeForever(voices -> {
+            if (voices == null) {
+                return;
+            }
             Log.v(LOG_TAG, "mAllVoices update: " + voices);
             // Update cached voices
             if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.R) {
@@ -240,11 +247,39 @@ public AppRepository(Application application) throws IOException {
                 mAllCachedVoices = Objects.requireNonNullElse(voices, new ArrayList<>());
             }
         });
-
+        mAllUserDictEntries = mNormDictDao.getSortedEntries();
+        mAllUserDictEntries.observeForever(entries -> {
+            if (entries == null) {
+                return;
+            }
+            Log.v(LOG_TAG, "mAllUserDictEntries update: " + entries.size() + " entries");
+            mAllCachedUserDictEntries = createNormDictRegexMap(entries);
+        });
         mMediaPlayer = new MediaPlayObserver();
         mScheduler = Executors.newSingleThreadScheduledExecutor();
         // only do this once at the beginning
         mScheduler.schedule(assetVoiceRunnable, 0, TimeUnit.SECONDS);
+        Log.v(LOG_TAG, "AppRepository() done");
+    }
+
+    /**
+     * Create a map of regex patterns for all normalization dictionary entries for fast
+     * lookup during normalization.
+     *
+     * @param entries   list of normalization dictionary entries
+     * @return         map of regex patterns for all normalization dictionary entries
+     */
+    private HashMap<NormDictEntry, Pattern> createNormDictRegexMap(List<NormDictEntry> entries) {
+        HashMap<NormDictEntry, Pattern> regexMap = new HashMap<>();
+        for (NormDictEntry entry : entries) {
+            // make for every NormDictEntry.term a regular expression matching on word boundaries and
+            // case insensitive. Note: we cannot use \b for the word boundary's term end,
+            // because of possible trailing punctuation
+            Pattern regex = Pattern.compile("\\b(?i)" +
+                    Pattern.quote(entry.term.strip().toLowerCase()) + "(?!\\S)");
+            regexMap.put(entry, regex);
+        }
+        return regexMap;
     }
 
     /**
@@ -345,33 +380,39 @@ public LiveData<List<com.grammatek.simaromur.db.Voice>> getAllVoices() {
      *
      * @return List of all current normalization dictionary entries as LiveData
      */
-    public LiveData<List<NormDictEntry>> getNormDictEntries() {
-        return mNormDictDao.getSortedEntries();
-    }
-
-    /**
-     * Get a LiveData list of all current normalization dictionary entries.
-     *
-     * @return List of all current normalization dictionary entries as LiveData
-     */
-    public List<NormDictEntry> getNormDictEntriesDirect() {
-        return mNormDictDao.getEntries();
+    public LiveData<List<NormDictEntry>> getUserDictEntries() {
+        Log.v(LOG_TAG, "getUserDictEntries");
+        return mAllUserDictEntries;
     }
 
     /**
      * Creates or updates the given entry inside the Db.
      */
-    public void createOrUpdateNormDictEntry(NormDictEntry entry) {
+    public void createOrUpdateUserDictEntry(NormDictEntry entry) {
         mNormDictDao.insert(entry);
+        mUtteranceCacheManager.clearCache();
     }
 
 
     /**
      * Deletes the given entry from the Db.
      * @param mEntry  the entry to be deleted
      */
-    public void deleteNormDictEntry(NormDictEntry mEntry) {
+    public void deleteUserDictEntry(NormDictEntry mEntry) {
         mNormDictDao.delete(mEntry);
+        mUtteranceCacheManager.clearCache();
+    }
+
+    /**
+     * Returns map of all observed/cached user normalization dictionary entries and the corresponding
+     * compiled term regular expression pattern. If there are any changes in the model, this map
+     * is updated automatically.
+     *
+     * @return map of all cached voices
+     */
+    public final HashMap<NormDictEntry, Pattern>  getCachedUserDictEntries() {
+        Log.v(LOG_TAG, "getCachedUserDictEntries");
+        return mAllCachedUserDictEntries;
     }
 
     /**
@@ -1160,5 +1201,4 @@ protected Void doInBackground(Void... voids) {
             return null;
         }
     }
-
 }
diff --git a/app/src/main/java/com/grammatek/simaromur/NormDictViewModel.java b/app/src/main/java/com/grammatek/simaromur/NormDictViewModel.java
@@ -32,7 +32,7 @@ public NormDictViewModel(Application application) {
     // Return all entries
     public LiveData<List<NormDictEntry>> getEntries() {
         if (mAllEntries == null) {
-            mAllEntries = mRepository.getNormDictEntries();
+            mAllEntries = mRepository.getUserDictEntries();
         }
         return mAllEntries;
     }
@@ -82,11 +82,11 @@ public void stopSpeaking(Voice voice) {
 
     public void createOrUpdate(NormDictEntry mEntry) {
         Log.v(LOG_TAG, "update: " + mEntry.term + " -> " + mEntry.replacement);
-        mRepository.createOrUpdateNormDictEntry(mEntry);
+        mRepository.createOrUpdateUserDictEntry(mEntry);
     }
 
     public void delete(NormDictEntry mEntry) {
         Log.v(LOG_TAG, "delete: " + mEntry.term + " -> " + mEntry.replacement);
-        mRepository.deleteNormDictEntry(mEntry);
+        mRepository.deleteUserDictEntry(mEntry);
     }
 }
diff --git a/app/src/main/java/com/grammatek/simaromur/Simaromur.java b/app/src/main/java/com/grammatek/simaromur/Simaromur.java
@@ -7,16 +7,20 @@
 public class Simaromur extends Activity {
 
   public Simaromur() {
+    /**
     if(BuildConfig.DEBUG)
       StrictMode.enableDefaults();
+    */
   }
 
   @Override
   public void onCreate(Bundle savedInstanceState) {
     super.onCreate(savedInstanceState);
+    /**
     StrictMode.setVmPolicy(new StrictMode.VmPolicy.Builder(StrictMode.getVmPolicy())
             .detectLeakedClosableObjects()
             .build());
+    */
     finish();
   }
 }
diff --git a/app/src/main/java/com/grammatek/simaromur/TTSService.java b/app/src/main/java/com/grammatek/simaromur/TTSService.java
@@ -41,10 +41,11 @@ public class TTSService extends TextToSpeechService {
     @Override
     public void onCreate() {
         Log.i(LOG_TAG, "onCreate()");
-        StrictMode.setVmPolicy(new StrictMode.VmPolicy.Builder(StrictMode.getVmPolicy())
-                .detectLeakedClosableObjects()
-                .build());
-
+        /**
+            StrictMode.setVmPolicy(new StrictMode.VmPolicy.Builder(StrictMode.getVmPolicy())
+                    .detectLeakedClosableObjects()
+                    .build());
+        */
         mRepository = App.getAppRepository();
         // This calls onIsLanguageAvailable() and must run after Initialization
         super.onCreate();

diff --git a/app/src/main/java/com/grammatek/simaromur/VoiceViewModel.java b/app/src/main/java/com/grammatek/simaromur/VoiceViewModel.java
@@ -26,7 +26,6 @@ public class VoiceViewModel extends AndroidViewModel {
 
     // these variables are for data caching
     private AppData mAppData;                       // application data
-    private LiveData<List<Voice>> mAllVoices;       // our current voices model
 
     public VoiceViewModel(Application application) {
         super(application);
@@ -43,18 +42,13 @@ public AppData getAppData() {
 
     // Return all voices
     public LiveData<List<Voice>> getAllVoices() {
-        if (mAllVoices == null) {
-            mAllVoices = mRepository.getAllVoices();
-        }
-        return mAllVoices;
+        return mRepository.getAllVoices();
     }
 
     // Return specific voice
     public Voice getVoiceWithId(long voiceId) {
-        if (mAllVoices == null) {
-            return null;
-        }
-        for(Voice voice: Objects.requireNonNull(mAllVoices.getValue())) {
+        List<Voice> voices = getAllVoices().getValue();
+        for(Voice voice: Objects.requireNonNull(voices)) {
             if (voice.voiceId == voiceId)
                 return voice;
         }

diff --git a/app/src/main/java/com/grammatek/simaromur/cache/UtteranceCacheManager.java b/app/src/main/java/com/grammatek/simaromur/cache/UtteranceCacheManager.java
@@ -635,7 +635,6 @@ public static String buildVoiceKey(String voiceName, String voiceVersion) {
         assert(!voiceName.isEmpty());
         assert(!voiceVersion.isEmpty());
         final String key = voiceName + ":" + voiceVersion;
-        Log.v(LOG_TAG, "buildVoiceKey(): " + key);
         return key;
     }
 

diff --git a/app/src/main/java/com/grammatek/simaromur/frontend/Abbreviations.java b/app/src/main/java/com/grammatek/simaromur/frontend/Abbreviations.java
@@ -37,6 +37,7 @@ public Set<String> getNonEndingAbbr() {
     }
 
     private Set<String> readAbbrFromFile(int resID) {
+        // TODO: move to FileUtils
         Set<String> abbrSet = new HashSet<>();
         Resources res = context.getResources();
         String line;

diff --git a/app/src/main/java/com/grammatek/simaromur/frontend/NormalizationManager.java b/app/src/main/java/com/grammatek/simaromur/frontend/NormalizationManager.java
@@ -29,7 +29,7 @@ public class NormalizationManager {
     private static final String POS_MODEL = "pos/is-pos-reduced-maxent.bin";
 
     private final Context mContext;
-    private final POSTaggerME mPosTagger;
+    private POSTaggerME mPosTagger;
     private final TTSUnicodeNormalizer mUnicodeNormalizer;
     private final Tokenizer mTokenizer;
     private final TTSNormalizer mTTSNormalizer;
@@ -39,11 +39,6 @@ public NormalizationManager(Context context, Map<String, PronDictEntry> pronDict
         mUnicodeNormalizer = new TTSUnicodeNormalizer(context, pronDict);
         mTokenizer = new Tokenizer(context);
         mTTSNormalizer = new TTSNormalizer();
-        mPosTagger = initPOSTagger();
-        if (mPosTagger == null) {
-            Log.e(LOG_TAG, "Failed to initialize POS tagger");
-            throw new RuntimeException("Failed to initialize POS tagger");
-        }
     }
 
     /**
@@ -113,6 +108,15 @@ private String[] tagText(final String text) {
 
             return tags;
         }
+
+        if (mPosTagger == null) {
+            // this takes ~2 seconds
+            mPosTagger = initPOSTagger();
+            if (mPosTagger == null) {
+                Log.e(LOG_TAG, "Failed to initialize POS tagger");
+                throw new RuntimeException("Failed to initialize POS tagger");
+            }
+        }
         tags = mPosTagger.tag(tokens);
         if (DEBUG) {
             printProbabilities(tags, mPosTagger, tokens);
@@ -127,7 +131,6 @@ private POSTaggerME initPOSTagger() {
             InputStream iStream = mContext.getAssets().open(POS_MODEL);
             POSModel posModel = new POSModel(iStream);
             posTagger = new POSTaggerME(posModel);
-
         } catch(IOException e) {
             e.printStackTrace();
         }

diff --git a/app/src/main/java/com/grammatek/simaromur/frontend/Pronunciation.java b/app/src/main/java/com/grammatek/simaromur/frontend/Pronunciation.java
@@ -55,10 +55,13 @@ public class Pronunciation {
     }
 
     public Pronunciation(Context context) {
+        Log.v(LOG_TAG, "Pronunciation() called");
         mContext = context;
-        mPronDict = readPronDict();
-        mIpaPronDict = readIpaPronDict();
         mAlphabets = initializeAlphabets();
+        // the following members are lazily initialized:
+        // - mG2P
+        // - mPronDict
+        // - mIpaPronDict
     }
 
     public String transcribe(String text) {
@@ -133,14 +136,15 @@ private String transcribeString(String text) {
         StringBuilder sb = new StringBuilder();
         for (String tok : tokens) {
             String transcr = "";
-            if (mPronDict.containsKey(tok)) {
-                transcr = mPronDict.get(tok).getTranscript().trim();
+            Map<String, PronDictEntry> pronDict = GetPronDict();
+            if (pronDict.containsKey(tok)) {
+                transcr = pronDict.get(tok).getTranscript().trim();
             }
             else if (tok.equals(SymbolsLvLIs.TagPause)){
                 transcr = SymbolsLvLIs.SymbolShortPause;
             }
             else {
-                transcr = mG2P.process(tok).trim();
+                transcr = GetG2p().process(tok).trim();
             }
 
             // bug in Thrax grammar, catch the error here: insert space before C if missing
@@ -182,10 +186,25 @@ private List<String> getValidAlphabets() {
         return List.copyOf(mAlphabets.keySet());
     }
 
-    public void initializeG2P() {
+    private NativeG2P initializeG2P() {
         if (mG2P == null) {
             mG2P = new NativeG2P(this.mContext);
         }
+        return mG2P;
+    }
+
+    private Map<String, PronDictEntry> initializePronDict() {
+        if (mPronDict == null) {
+            mPronDict = readPronDict();
+        }
+        return mPronDict;
+    }
+
+    private Map<String, PronDictEntry> initializeIpaPronDict() {
+        if (mIpaPronDict == null) {
+            mIpaPronDict = readIpaPronDict();
+        }
+        return mIpaPronDict;
     }
 
 
@@ -229,6 +248,7 @@ private Map<String, Map<String, Map<String, String>>> initializeAlphabets() {
     }
 
     private Map<String, PronDictEntry> readPronDict() {
+        Log.v(LOG_TAG, "readPronDict() called");
         Map<String, PronDictEntry> pronDict = new HashMap<>();
         final List<String> fileContent = FileUtils.readLinesFromResourceFile(this.mContext,
                 R.raw.ice_pron_dict_standard_clear_2201_extended);
@@ -242,6 +262,7 @@ private Map<String, PronDictEntry> readPronDict() {
     }
 
     private Map<String, PronDictEntry> readIpaPronDict() {
+        Log.v(LOG_TAG, "readIpaPronDict() called");
         Map<String, PronDictEntry> pronDict = new HashMap<>();
         final List<String> fileContent = FileUtils.readLinesFromResourceFile(this.mContext,
                 R.raw.ice_pron_dict_standard_clear_2201_extended);
@@ -255,14 +276,14 @@ private Map<String, PronDictEntry> readIpaPronDict() {
     }
 
     public NativeG2P GetG2p() {
-        return mG2P;
+        return initializeG2P();
     }
 
     public Map<String, PronDictEntry> GetPronDict() {
-        return mPronDict;
+        return initializePronDict();
     }
     public Map<String, PronDictEntry> GetIpaPronDict() {
-        return mIpaPronDict;
+        return initializeIpaPronDict();
     }
     public Map<String, Map<String, Map<String, String>>> GetAlphabets() {
         return mAlphabets;

diff --git a/app/src/main/java/com/grammatek/simaromur/frontend/PronunciationVits.java b/app/src/main/java/com/grammatek/simaromur/frontend/PronunciationVits.java
@@ -65,9 +65,6 @@ public String transcribe(String text) {
      *          relevant phonemes were found.
      */
     public String transcribe(String text, final String voiceType, final String voiceVersion) {
-        // lazy initialization of g2p
-        if (mPronounciation.GetG2p() == null)
-            mPronounciation.initializeG2P();
         mG2P = mPronounciation.GetG2p();
         assert(mG2P != null);