Merge "Find multiple previous word information to support n-gram."

1 files changed, 59 insertions, 46 deletions

diff --git a/java/src/com/android/inputmethod/latin/RichInputConnection.java b/java/src/com/android/inputmethod/latin/RichInputConnection.java
index c027892ba..62b55bca1 100644
--- a/java/src/com/android/inputmethod/latin/RichInputConnection.java
+++ b/java/src/com/android/inputmethod/latin/RichInputConnection.java
@@ -26,6 +26,7 @@ import android.view.inputmethod.ExtractedText;
 import android.view.inputmethod.ExtractedTextRequest;
 import android.view.inputmethod.InputConnection;
 
+import com.android.inputmethod.latin.PrevWordsInfo.WordInfo;
 import com.android.inputmethod.latin.settings.SpacingAndPunctuations;
 import com.android.inputmethod.latin.utils.CapsModeUtils;
 import com.android.inputmethod.latin.utils.DebugLogUtils;
@@ -49,8 +50,10 @@ public final class RichInputConnection {
     private static final boolean DBG = false;
     private static final boolean DEBUG_PREVIOUS_TEXT = false;
     private static final boolean DEBUG_BATCH_NESTING = false;
-    // Provision for a long word pair and a separator
-    private static final int LOOKBACK_CHARACTER_NUM = Constants.DICTIONARY_MAX_WORD_LENGTH * 2 + 1;
+    // Provision for long words and separators between the words.
+    private static final int LOOKBACK_CHARACTER_NUM = Constants.DICTIONARY_MAX_WORD_LENGTH
+            * (Constants.MAX_PREV_WORD_COUNT_FOR_N_GRAM + 1) /* words */
+            + Constants.MAX_PREV_WORD_COUNT_FOR_N_GRAM /* separators */;
     private static final Pattern spaceRegex = Pattern.compile("\\s+");
     private static final int INVALID_CURSOR_POSITION = -1;
 
@@ -544,22 +547,25 @@ public final class RichInputConnection {
         return Arrays.binarySearch(sortedSeparators, code) >= 0;
     }
 
-    // Get information of the nth word before cursor. n = 1 retrieves the word immediately before
-    // the cursor, n = 2 retrieves the word before that, and so on. This splits on whitespace only.
+    // Get context information from nth word before the cursor. n = 1 retrieves the words
+    // immediately before the cursor, n = 2 retrieves the words 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 information representing beginning-of-sentence).
-    // Example :
-    // (n = 1) "abc def|" -> def
-    // (n = 1) "abc def |" -> def
-    // (n = 1) "abc 'def|" -> 'def
+    // Example (when Constants.MAX_PREV_WORD_COUNT_FOR_N_GRAM is 2):
+    // (n = 1) "abc def|" -> abc, def
+    // (n = 1) "abc def |" -> abc, def
+    // (n = 1) "abc 'def|" -> empty, 'def
     // (n = 1) "abc def. |" -> beginning-of-sentence
     // (n = 1) "abc def . |" -> beginning-of-sentence
-    // (n = 2) "abc def|" -> abc
-    // (n = 2) "abc def |" -> abc
+    // (n = 2) "abc def|" -> beginning-of-sentence, abc
+    // (n = 2) "abc def |" -> beginning-of-sentence, abc
     // (n = 2) "abc 'def|" -> empty. The context is different from "abc def", but we cannot
     // represent this situation using PrevWordsInfo. See TODO in the method.
-    // (n = 2) "abc def. |" -> abc
-    // (n = 2) "abc def . |" -> def
+    // TODO: The next example's result should be "abc, def". This have to be fixed before we
+    // retrieve the prior context of Beginning-of-Sentence.
+    // (n = 2) "abc def. |" -> beginning-of-sentence, abc
+    // (n = 2) "abc def . |" -> abc, def
     // (n = 2) "abc|" -> beginning-of-sentence
     // (n = 2) "abc |" -> beginning-of-sentence
     // (n = 2) "abc. def|" -> beginning-of-sentence
@@ -567,43 +573,50 @@ public final class RichInputConnection {
             final SpacingAndPunctuations spacingAndPunctuations, final int n) {
         if (prev == null) return PrevWordsInfo.EMPTY_PREV_WORDS_INFO;
         final String[] w = spaceRegex.split(prev);
-
-        // Referring to the word after the nth word.
-        if ((n - 1) > 0 && (n - 1) <= w.length) {
-            final String wordFollowingTheNthPrevWord = w[w.length - n + 1];
-            if (!wordFollowingTheNthPrevWord.isEmpty()) {
-                final char firstChar = wordFollowingTheNthPrevWord.charAt(0);
-                if (spacingAndPunctuations.isWordConnector(firstChar)) {
-                    // The word following the n-th prev word is starting with a word connector.
-                    // TODO: Return meaningful context for this case.
-                    return PrevWordsInfo.EMPTY_PREV_WORDS_INFO;
+        final WordInfo[] prevWordsInfo = new WordInfo[Constants.MAX_PREV_WORD_COUNT_FOR_N_GRAM];
+        for (int i = 0; i < prevWordsInfo.length; i++) {
+            final int focusedWordIndex = w.length - n - i;
+            // Referring to the word after the focused word.
+            if ((focusedWordIndex + 1) >= 0 && (focusedWordIndex + 1) < w.length) {
+                final String wordFollowingTheNthPrevWord = w[focusedWordIndex + 1];
+                if (!wordFollowingTheNthPrevWord.isEmpty()) {
+                    final char firstChar = wordFollowingTheNthPrevWord.charAt(0);
+                    if (spacingAndPunctuations.isWordConnector(firstChar)) {
+                        // The word following the focused word is starting with a word connector.
+                        // TODO: Return meaningful context for this case.
+                        prevWordsInfo[i] = WordInfo.EMPTY_WORD_INFO;
+                        break;
+                    }
                 }
             }
+            // If we can't find (n + i) words, the context is beginning-of-sentence.
+            if (focusedWordIndex < 0) {
+                prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE;
+                break;
+            }
+            final String focusedWord = w[focusedWordIndex];
+            // If the word is empty, the context is beginning-of-sentence.
+            final int length = focusedWord.length();
+            if (length <= 0) {
+                prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE;
+                break;
+            }
+            // If ends in a sentence separator, the context is beginning-of-sentence.
+            final char lastChar = focusedWord.charAt(length - 1);
+            if (spacingAndPunctuations.isSentenceSeparator(lastChar)) {
+                prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE;
+                break;
+            }
+            // 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)) {
+                prevWordsInfo[i] = WordInfo.EMPTY_WORD_INFO;
+                break;
+            }
+            prevWordsInfo[i] = new WordInfo(focusedWord);
         }
-
-        // If we can't find n words, or we found an empty word, the context is
-        // beginning-of-sentence.
-        if (w.length < n) {
-            return PrevWordsInfo.BEGINNING_OF_SENTENCE;
-        }
-        final String nthPrevWord = w[w.length - n];
-        final int length = nthPrevWord.length();
-        if (length <= 0) {
-            return PrevWordsInfo.BEGINNING_OF_SENTENCE;
-        }
-
-        // If ends in a sentence separator, the context is beginning-of-sentence.
-        final char lastChar = nthPrevWord.charAt(length - 1);
-        if (spacingAndPunctuations.isSentenceSeparator(lastChar)) {
-            return PrevWordsInfo.BEGINNING_OF_SENTENCE;
-        }
-        // 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 PrevWordsInfo.EMPTY_PREV_WORDS_INFO;
-        }
-        return new PrevWordsInfo(new PrevWordsInfo.WordInfo(nthPrevWord));
+        return new PrevWordsInfo(prevWordsInfo);
     }
 
     /**