Add beginning of sentence information in PrevWordsInfo.

Bug: 14119293
Bug: 14425059
Change-Id: I65320920e840082b0b697bb621676716d0933e0c

java/src/com/android/inputmethod/latin/LatinIME.java

@@ -1435,12 +1435,13 @@ public class LatinIME extends InputMethodService implements KeyboardActionListen
                 // We're checking the previous word in the text field against the memorized previous
                 // word. If we are composing a word we should have the second word before the cursor
                 // memorized, otherwise we should have the first.
-                final CharSequence rereadPrevWord = mInputLogic.getNthPreviousWordForSuggestion(
+                final PrevWordsInfo rereadPrevWordsInfo =
-                        currentSettings.mSpacingAndPunctuations,
+                        mInputLogic.getPrevWordsInfoFromNthPreviousWordForSuggestion(
-                        mInputLogic.mWordComposer.isComposingWord() ? 2 : 1);
+                                currentSettings.mSpacingAndPunctuations,
-                if (!TextUtils.equals(prevWordsInfo.mPrevWord, rereadPrevWord)) {
+                                mInputLogic.mWordComposer.isComposingWord() ? 2 : 1);
+                if (!TextUtils.equals(prevWordsInfo.mPrevWord, rereadPrevWordsInfo.mPrevWord)) {
                     throw new RuntimeException("Unexpected previous word: "
-                            + prevWordsInfo.mPrevWord + " <> " + rereadPrevWord);
+                            + prevWordsInfo.mPrevWord + " <> " + rereadPrevWordsInfo.mPrevWord);
                 }
             }
         }

java/src/com/android/inputmethod/latin/PrevWordsInfo.java

@@ -16,6 +16,9 @@
 
 package com.android.inputmethod.latin;
 
+import android.util.Log;
+
+// TODO: Support multiple previous words for n-gram.
 public class PrevWordsInfo {
     // The previous word. May be null after resetting and before starting a new composing word, or
     // when there is no context like at the start of text for example. It can also be set to null
@@ -23,7 +26,18 @@ public class PrevWordsInfo {
     // or a comma.
     public final String mPrevWord;
 
+    // TODO: Have sentence separator.
+    // Whether the current context is beginning of sentence or not.
+    public final boolean mIsBeginningOfSentence;
+
+    // Beginning of sentence.
+    public PrevWordsInfo() {
+        mPrevWord = null;
+        mIsBeginningOfSentence = true;
+    }
+
     public PrevWordsInfo(final String prevWord) {
         mPrevWord = prevWord;
+        mIsBeginningOfSentence = false;
     }
 }

java/src/com/android/inputmethod/latin/RichInputConnection.java

@@ -538,10 +538,12 @@ public final class RichInputConnection {
     }
 
     @SuppressWarnings("unused")
-    public String getNthPreviousWord(final SpacingAndPunctuations spacingAndPunctuations,
+    public PrevWordsInfo getPrevWordsInfoFromNthPreviousWord(
-            final int n) {
+            final SpacingAndPunctuations spacingAndPunctuations, final int n) {
         mIC = mParent.getCurrentInputConnection();
-        if (null == mIC) return null;
+        if (null == mIC) {
+            return new PrevWordsInfo(null);
+        }
         final CharSequence prev = getTextBeforeCursor(LOOKBACK_CHARACTER_NUM, 0);
         if (DEBUG_PREVIOUS_TEXT && null != prev) {
             final int checkLength = LOOKBACK_CHARACTER_NUM - 1;
@@ -561,46 +563,57 @@ public final class RichInputConnection {
                 }
             }
         }
-        return getNthPreviousWord(prev, spacingAndPunctuations, n);
+        return getPrevWordsInfoFromNthPreviousWord(prev, spacingAndPunctuations, n);
     }
 
     private static boolean isSeparator(final int code, final int[] sortedSeparators) {
         return Arrays.binarySearch(sortedSeparators, code) >= 0;
     }
 
-    // Get the nth word before cursor. n = 1 retrieves the word immediately before the cursor,
+    // Get information of the nth word before cursor. n = 1 retrieves the word immediately before
-    // n = 2 retrieves the word before that, and so on. This splits on whitespace only.
+    // the cursor, n = 2 retrieves the word before that, and so on. This splits on whitespace only.
     // Also, it won't return words that end in a separator (if the nth word before the cursor
-    // ends in a separator, it returns null).
+    // ends in a separator, it returns information represents beginning-of-sentence).
     // Example :
     // (n = 1) "abc def|" -> def
     // (n = 1) "abc def |" -> def
-    // (n = 1) "abc def. |" -> null
+    // (n = 1) "abc def. |" -> beginning-of-sentence
-    // (n = 1) "abc def . |" -> null
+    // (n = 1) "abc def . |" -> beginning-of-sentence
     // (n = 2) "abc def|" -> abc
     // (n = 2) "abc def |" -> abc
     // (n = 2) "abc def. |" -> abc
     // (n = 2) "abc def . |" -> def
-    // (n = 2) "abc|" -> null
+    // (n = 2) "abc|" -> beginning-of-sentence
-    // (n = 2) "abc |" -> null
+    // (n = 2) "abc |" -> beginning-of-sentence
-    // (n = 2) "abc. def|" -> null
+    // (n = 2) "abc. def|" -> beginning-of-sentence
-    public static String getNthPreviousWord(final CharSequence prev,
+    public static PrevWordsInfo getPrevWordsInfoFromNthPreviousWord(final CharSequence prev,
             final SpacingAndPunctuations spacingAndPunctuations, final int n) {
-        if (prev == null) return null;
+        if (prev == null) return new PrevWordsInfo(null);
         final String[] w = spaceRegex.split(prev);
 
-        // If we can't find n words, or we found an empty word, return null.
+        // If we can't find n words, or we found an empty word, the context is
-        if (w.length < n) return null;
+        // beginning-of-sentence.
+        if (w.length < n) {
+            return new PrevWordsInfo();
+        }
         final String nthPrevWord = w[w.length - n];
         final int length = nthPrevWord.length();
-        if (length <= 0) return null;
+        if (length <= 0) {
+            return  new PrevWordsInfo();
+        }
 
-        // If ends in a separator, return null
+        // If ends in a sentence separator, the context is beginning-of-sentence.
         final char lastChar = nthPrevWord.charAt(length - 1);
+        if (spacingAndPunctuations.isSentenceSeparator(lastChar)) {
+            new PrevWordsInfo();
+        }
+        // If ends in a word separator or connector, the context is unclear.
+        // TODO: Return meaningful context for this case.
         if (spacingAndPunctuations.isWordSeparator(lastChar)
-                || spacingAndPunctuations.isWordConnector(lastChar)) return null;
+                || spacingAndPunctuations.isWordConnector(lastChar)) {
-
+            return new PrevWordsInfo(null);
-        return nthPrevWord;
+        }
+        return new PrevWordsInfo(nthPrevWord);
     }
 
     /**

java/src/com/android/inputmethod/latin/WordComposer.java

@@ -371,12 +371,12 @@ public final class WordComposer {
      * Also, batch input needs to know about the current caps mode to display correctly
      * capitalized suggestions.
      * @param mode the mode at the time of start
-     * @param previousWord the previous word as context for suggestions. May be null if none.
+     * @param prevWordsInfo the information of previous words
      */
     public void setCapitalizedModeAndPreviousWordAtStartComposingTime(final int mode,
-            final CharSequence previousWord) {
+            final PrevWordsInfo prevWordsInfo) {
         mCapitalizedMode = mode;
-        mPrevWordsInfo = new PrevWordsInfo(null == previousWord ? null : previousWord.toString());
+        mPrevWordsInfo = prevWordsInfo;
     }
 
     /**

java/src/com/android/inputmethod/latin/inputlogic/InputLogic.java

@@ -575,7 +575,7 @@ public final class InputLogic {
         mWordComposer.setCapitalizedModeAndPreviousWordAtStartComposingTime(
                 getActualCapsMode(settingsValues, keyboardSwitcher.getKeyboardShiftMode()),
                 // Prev word is 1st word before cursor
-                getNthPreviousWordForSuggestion(
+                getPrevWordsInfoFromNthPreviousWordForSuggestion(
                         settingsValues.mSpacingAndPunctuations, 1 /* nthPreviousWord */));
     }
 
@@ -614,7 +614,8 @@ public final class InputLogic {
                             getCurrentAutoCapsState(settingsValues), getCurrentRecapitalizeState());
                     mWordComposer.setCapitalizedModeAndPreviousWordAtStartComposingTime(
                             getActualCapsMode(settingsValues,
-                                    keyboardSwitcher.getKeyboardShiftMode()), commitParts[0]);
+                                    keyboardSwitcher.getKeyboardShiftMode()),
+                                            new PrevWordsInfo(commitParts[0]));
                     ++mAutoCommitSequenceNumber;
                 }
             }
@@ -765,7 +766,8 @@ public final class InputLogic {
                 // We pass 1 to getPreviousWordForSuggestion because we were not composing a word
                 // yet, so the word we want is the 1st word before the cursor.
                 mWordComposer.setCapitalizedModeAndPreviousWordAtStartComposingTime(
-                        inputTransaction.mShiftState, getNthPreviousWordForSuggestion(
+                        inputTransaction.mShiftState,
+                        getPrevWordsInfoFromNthPreviousWordForSuggestion(
                                 settingsValues.mSpacingAndPunctuations, 1 /* nthPreviousWord */));
             }
             mConnection.setComposingText(getTextWithUnderline(
@@ -1326,7 +1328,8 @@ public final class InputLogic {
             // Show predictions.
             mWordComposer.setCapitalizedModeAndPreviousWordAtStartComposingTime(
                     WordComposer.CAPS_MODE_OFF,
-                    getNthPreviousWordForSuggestion(settingsValues.mSpacingAndPunctuations, 1));
+                    getPrevWordsInfoFromNthPreviousWordForSuggestion(
+                            settingsValues.mSpacingAndPunctuations, 1));
             mLatinIME.mHandler.postUpdateSuggestionStrip();
             return;
         }
@@ -1374,11 +1377,9 @@ public final class InputLogic {
         // We want the previous word for suggestion. If we have chars in the word
         // before the cursor, then we want the word before that, hence 2; otherwise,
         // we want the word immediately before the cursor, hence 1.
-        final CharSequence prevWord = getNthPreviousWordForSuggestion(
+        final PrevWordsInfo prevWordsInfo = getPrevWordsInfoFromNthPreviousWordForSuggestion(
                 settingsValues.mSpacingAndPunctuations,
                 0 == numberOfCharsInWordBeforeCursor ? 1 : 2);
-        final PrevWordsInfo prevWordsInfo =
-                new PrevWordsInfo(prevWord != null ? prevWord.toString() : null);
         mWordComposer.setComposingWord(codePoints,
                 mLatinIME.getCoordinatesForCurrentKeyboard(codePoints), prevWordsInfo);
         mWordComposer.setCursorPositionWithinWord(
@@ -1590,21 +1591,23 @@ public final class InputLogic {
     }
 
     /**
-     * Get the nth previous word before the cursor as context for the suggestion process.
+     * Get information fo previous words from the nth previous word before the cursor as context
+     * for the suggestion process.
      * @param spacingAndPunctuations the current spacing and punctuations settings.
      * @param nthPreviousWord reverse index of the word to get (1-indexed)
-     * @return the nth previous word before the cursor.
+     * @return the information of previous words
      */
     // TODO: Make this private
-    public CharSequence getNthPreviousWordForSuggestion(
+    public PrevWordsInfo getPrevWordsInfoFromNthPreviousWordForSuggestion(
             final SpacingAndPunctuations spacingAndPunctuations, final int nthPreviousWord) {
         if (spacingAndPunctuations.mCurrentLanguageHasSpaces) {
             // If we are typing in a language with spaces we can just look up the previous
-            // word from textview.
+            // word information from textview.
-            return mConnection.getNthPreviousWord(spacingAndPunctuations, nthPreviousWord);
+            return mConnection.getPrevWordsInfoFromNthPreviousWord(
+                    spacingAndPunctuations, nthPreviousWord);
         } else {
-            return LastComposedWord.NOT_A_COMPOSED_WORD == mLastComposedWord ? null
+            return LastComposedWord.NOT_A_COMPOSED_WORD == mLastComposedWord ? new PrevWordsInfo()
-                    : mLastComposedWord.mCommittedWord;
+                    : new PrevWordsInfo(mLastComposedWord.mCommittedWord.toString());
         }
     }
 
@@ -1972,8 +1975,8 @@ public final class InputLogic {
                         suggestedWords);
         // Use the 2nd previous word as the previous word because the 1st previous word is the word
         // to be committed.
-        final PrevWordsInfo prevWordsInfo = new PrevWordsInfo(mConnection.getNthPreviousWord(
+        final PrevWordsInfo prevWordsInfo = mConnection.getPrevWordsInfoFromNthPreviousWord(
-                settingsValues.mSpacingAndPunctuations, 2));
+                settingsValues.mSpacingAndPunctuations, 2);
         mConnection.commitText(chosenWordWithSuggestions, 1);
         // Add the word to the user history dictionary
         performAdditionToUserHistoryDictionary(settingsValues, chosenWord, prevWordsInfo);

tests/src/com/android/inputmethod/latin/RichInputConnectionAndTextRangeTests.java

@@ -155,13 +155,17 @@ public class RichInputConnectionAndTextRangeTests extends AndroidTestCase {
      */
     public void testGetPreviousWord() {
         // If one of the following cases breaks, the bigram suggestions won't work.
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def", mSpacingAndPunctuations, 2), "abc");
+                "abc def", mSpacingAndPunctuations, 2).mPrevWord, "abc");
-        assertNull(RichInputConnection.getNthPreviousWord(
+        assertNull(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc", mSpacingAndPunctuations, 2));
+                "abc", mSpacingAndPunctuations, 2).mPrevWord);
-        assertNull(RichInputConnection.getNthPreviousWord(
+        assertNull(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc. def", mSpacingAndPunctuations, 2));
+                "abc. def", mSpacingAndPunctuations, 2).mPrevWord);
 
+        assertFalse(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
+                "abc def", mSpacingAndPunctuations, 2).mIsBeginningOfSentence);
+        assertTrue(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
+                "abc", mSpacingAndPunctuations, 2).mIsBeginningOfSentence);
         // The following tests reflect the current behavior of the function
         // RichInputConnection#getNthPreviousWord.
         // TODO: However at this time, the code does never go
@@ -169,23 +173,23 @@ public class RichInputConnectionAndTextRangeTests extends AndroidTestCase {
         // this function if needed - especially since it does not seem very
         // logical. These tests are just there to catch any unintentional
         // changes in the behavior of the RichInputConnection#getPreviousWord method.
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def ", mSpacingAndPunctuations, 2), "abc");
+                "abc def ", mSpacingAndPunctuations, 2).mPrevWord, "abc");
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def.", mSpacingAndPunctuations, 2), "abc");
+                "abc def.", mSpacingAndPunctuations, 2).mPrevWord, "abc");
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def .", mSpacingAndPunctuations, 2), "def");
+                "abc def .", mSpacingAndPunctuations, 2).mPrevWord, "def");
-        assertNull(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc ", mSpacingAndPunctuations, 2));
+                "abc ", mSpacingAndPunctuations, 2).mPrevWord, null);
 
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def", mSpacingAndPunctuations, 1), "def");
+                "abc def", mSpacingAndPunctuations, 1).mPrevWord, "def");
-        assertEquals(RichInputConnection.getNthPreviousWord(
+        assertEquals(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def ", mSpacingAndPunctuations, 1), "def");
+                "abc def ", mSpacingAndPunctuations, 1).mPrevWord, "def");
-        assertNull(RichInputConnection.getNthPreviousWord(
+        assertNull(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def.", mSpacingAndPunctuations, 1));
+                "abc def.", mSpacingAndPunctuations, 1).mPrevWord);
-        assertNull(RichInputConnection.getNthPreviousWord(
+        assertNull(RichInputConnection.getPrevWordsInfoFromNthPreviousWord(
-                "abc def .", mSpacingAndPunctuations, 1));
+                "abc def .", mSpacingAndPunctuations, 1).mPrevWord);
     }
 
     /**