aboutsummaryrefslogtreecommitdiffstats
path: root/java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java
diff options
context:
space:
mode:
authorSatoshi Kataoka <satok@google.com>2012-07-10 15:33:18 +0900
committerSatoshi Kataoka <satok@google.com>2012-07-10 15:34:40 +0900
commit84ed0966417d93b07c4da2b295244b160d223ce9 (patch)
tree7eca078a3ca8e217c0edd6f96e7284bba080fc63 /java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java
parent7389c601e34dc96807aa1c3c0fef30b77198ca58 (diff)
downloadlatinime-84ed0966417d93b07c4da2b295244b160d223ce9.tar.gz
latinime-84ed0966417d93b07c4da2b295244b160d223ce9.tar.xz
latinime-84ed0966417d93b07c4da2b295244b160d223ce9.zip
Separate SpellCheckerSession from SpellCheckerService
Bug: 6789576 Change-Id: I7c55d36afad7ef6046353b3c9e849a54a6dc83ae
Diffstat (limited to 'java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java')
-rw-r--r--java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java409
1 files changed, 16 insertions, 393 deletions
diff --git a/java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java b/java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java
index 5332c066a..42b05592c 100644
--- a/java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java
+++ b/java/src/com/android/inputmethod/latin/spellcheck/AndroidSpellCheckerService.java
@@ -20,14 +20,9 @@ import android.content.Intent;
import android.content.SharedPreferences;
import android.preference.PreferenceManager;
import android.service.textservice.SpellCheckerService;
-import android.text.TextUtils;
import android.util.Log;
-import android.util.LruCache;
-import android.view.textservice.SentenceSuggestionsInfo;
import android.view.textservice.SuggestionsInfo;
-import android.view.textservice.TextInfo;
-import com.android.inputmethod.compat.SuggestionsInfoCompatUtils;
import com.android.inputmethod.keyboard.ProximityInfo;
import com.android.inputmethod.latin.BinaryDictionary;
import com.android.inputmethod.latin.ContactsBinaryDictionary;
@@ -37,12 +32,10 @@ import com.android.inputmethod.latin.DictionaryFactory;
import com.android.inputmethod.latin.LocaleUtils;
import com.android.inputmethod.latin.R;
import com.android.inputmethod.latin.StringUtils;
-import com.android.inputmethod.latin.SuggestedWords.SuggestedWordInfo;
import com.android.inputmethod.latin.SynchronouslyLoadedContactsBinaryDictionary;
import com.android.inputmethod.latin.SynchronouslyLoadedUserBinaryDictionary;
import com.android.inputmethod.latin.UserBinaryDictionary;
import com.android.inputmethod.latin.WhitelistDictionary;
-import com.android.inputmethod.latin.WordComposer;
import java.lang.ref.WeakReference;
import java.util.ArrayList;
@@ -65,9 +58,9 @@ public class AndroidSpellCheckerService extends SpellCheckerService
public static final String PREF_USE_CONTACTS_KEY = "pref_spellcheck_use_contacts";
- private static final int CAPITALIZE_NONE = 0; // No caps, or mixed case
- private static final int CAPITALIZE_FIRST = 1; // First only
- private static final int CAPITALIZE_ALL = 2; // All caps
+ public static final int CAPITALIZE_NONE = 0; // No caps, or mixed case
+ public static final int CAPITALIZE_FIRST = 1; // First only
+ public static final int CAPITALIZE_ALL = 2; // All caps
private final static String[] EMPTY_STRING_ARRAY = new String[0];
private Map<String, DictionaryPool> mDictionaryPools =
@@ -91,8 +84,8 @@ public class AndroidSpellCheckerService extends SpellCheckerService
public static final int SCRIPT_LATIN = 0;
public static final int SCRIPT_CYRILLIC = 1;
- private static final String SINGLE_QUOTE = "\u0027";
- private static final String APOSTROPHE = "\u2019";
+ public static final String SINGLE_QUOTE = "\u0027";
+ public static final String APOSTROPHE = "\u2019";
private static final TreeMap<String, Integer> mLanguageToScript;
static {
// List of the supported languages and their associated script. We won't check
@@ -129,7 +122,7 @@ public class AndroidSpellCheckerService extends SpellCheckerService
onSharedPreferenceChanged(prefs, PREF_USE_CONTACTS_KEY);
}
- private static int getScriptFromLocale(final Locale locale) {
+ public static int getScriptFromLocale(final Locale locale) {
final Integer script = mLanguageToScript.get(locale.getLanguage());
if (null == script) {
throw new RuntimeException("We have been called with an unsupported language: \""
@@ -194,17 +187,22 @@ public class AndroidSpellCheckerService extends SpellCheckerService
return new AndroidSpellCheckerSession(this);
}
- private static SuggestionsInfo getNotInDictEmptySuggestions() {
+ public static SuggestionsInfo getNotInDictEmptySuggestions() {
return new SuggestionsInfo(0, EMPTY_STRING_ARRAY);
}
- private static SuggestionsInfo getInDictEmptySuggestions() {
+ public static SuggestionsInfo getInDictEmptySuggestions() {
return new SuggestionsInfo(SuggestionsInfo.RESULT_ATTR_IN_THE_DICTIONARY,
EMPTY_STRING_ARRAY);
}
+ public SuggestionsGatherer newSuggestionsGatherer(final String text, int maxLength) {
+ return new SuggestionsGatherer(
+ text, mSuggestionThreshold, mRecommendedThreshold, maxLength);
+ }
+
// TODO: remove this class and replace it by storage local to the session.
- private static class SuggestionsGatherer {
+ public static class SuggestionsGatherer {
public static class Result {
public final String[] mSuggestions;
public final boolean mHasRecommendedSuggestions;
@@ -396,7 +394,7 @@ public class AndroidSpellCheckerService extends SpellCheckerService
}.start();
}
- private DictionaryPool getDictionaryPool(final String locale) {
+ public DictionaryPool getDictionaryPool(final String locale) {
DictionaryPool pool = mDictionaryPools.get(locale);
if (null == pool) {
final Locale localeObject = LocaleUtils.constructLocaleFromString(locale);
@@ -447,7 +445,7 @@ public class AndroidSpellCheckerService extends SpellCheckerService
}
// This method assumes the text is not empty or null.
- private static int getCapitalizationType(String text) {
+ public static int getCapitalizationType(String text) {
// If the first char is not uppercase, then the word is either all lower case,
// and in either case we return CAPITALIZE_NONE.
if (!Character.isUpperCase(text.codePointAt(0))) return CAPITALIZE_NONE;
@@ -464,379 +462,4 @@ public class AndroidSpellCheckerService extends SpellCheckerService
if (1 == capsCount) return CAPITALIZE_FIRST;
return (len == capsCount ? CAPITALIZE_ALL : CAPITALIZE_NONE);
}
-
- private static class AndroidSpellCheckerSession extends Session {
- // Immutable, but need the locale which is not available in the constructor yet
- private DictionaryPool mDictionaryPool;
- // Likewise
- private Locale mLocale;
- // Cache this for performance
- private int mScript; // One of SCRIPT_LATIN or SCRIPT_CYRILLIC for now.
-
- private final AndroidSpellCheckerService mService;
-
- private final SuggestionsCache mSuggestionsCache = new SuggestionsCache();
-
- private static class SuggestionsParams {
- public final String[] mSuggestions;
- public final int mFlags;
- public SuggestionsParams(String[] suggestions, int flags) {
- mSuggestions = suggestions;
- mFlags = flags;
- }
- }
-
- private static class SuggestionsCache {
- private static final char CHAR_DELIMITER = '\uFFFC';
- private static final int MAX_CACHE_SIZE = 50;
- private final LruCache<String, SuggestionsParams> mUnigramSuggestionsInfoCache =
- new LruCache<String, SuggestionsParams>(MAX_CACHE_SIZE);
-
- // TODO: Support n-gram input
- private static String generateKey(String query, String prevWord) {
- if (TextUtils.isEmpty(query) || TextUtils.isEmpty(prevWord)) {
- return query;
- }
- return query + CHAR_DELIMITER + prevWord;
- }
-
- // TODO: Support n-gram input
- public SuggestionsParams getSuggestionsFromCache(String query, String prevWord) {
- return mUnigramSuggestionsInfoCache.get(generateKey(query, prevWord));
- }
-
- // TODO: Support n-gram input
- public void putSuggestionsToCache(
- String query, String prevWord, String[] suggestions, int flags) {
- if (suggestions == null || TextUtils.isEmpty(query)) {
- return;
- }
- mUnigramSuggestionsInfoCache.put(
- generateKey(query, prevWord), new SuggestionsParams(suggestions, flags));
- }
- }
-
- AndroidSpellCheckerSession(final AndroidSpellCheckerService service) {
- mService = service;
- }
-
- @Override
- public void onCreate() {
- final String localeString = getLocale();
- mDictionaryPool = mService.getDictionaryPool(localeString);
- mLocale = LocaleUtils.constructLocaleFromString(localeString);
- mScript = getScriptFromLocale(mLocale);
- }
-
- /*
- * Returns whether the code point is a letter that makes sense for the specified
- * locale for this spell checker.
- * The dictionaries supported by Latin IME are described in res/xml/spellchecker.xml
- * and is limited to EFIGS languages and Russian.
- * Hence at the moment this explicitly tests for Cyrillic characters or Latin characters
- * as appropriate, and explicitly excludes CJK, Arabic and Hebrew characters.
- */
- private static boolean isLetterCheckableByLanguage(final int codePoint,
- final int script) {
- switch (script) {
- case SCRIPT_LATIN:
- // Our supported latin script dictionaries (EFIGS) at the moment only include
- // characters in the C0, C1, Latin Extended A and B, IPA extensions unicode
- // blocks. As it happens, those are back-to-back in the code range 0x40 to 0x2AF,
- // so the below is a very efficient way to test for it. As for the 0-0x3F, it's
- // excluded from isLetter anyway.
- return codePoint <= 0x2AF && Character.isLetter(codePoint);
- case SCRIPT_CYRILLIC:
- // All Cyrillic characters are in the 400~52F block. There are some in the upper
- // Unicode range, but they are archaic characters that are not used in modern
- // russian and are not used by our dictionary.
- return codePoint >= 0x400 && codePoint <= 0x52F && Character.isLetter(codePoint);
- default:
- // Should never come here
- throw new RuntimeException("Impossible value of script: " + script);
- }
- }
-
- /**
- * Finds out whether a particular string should be filtered out of spell checking.
- *
- * This will loosely match URLs, numbers, symbols. To avoid always underlining words that
- * we know we will never recognize, this accepts a script identifier that should be one
- * of the SCRIPT_* constants defined above, to rule out quickly characters from very
- * different languages.
- *
- * @param text the string to evaluate.
- * @param script the identifier for the script this spell checker recognizes
- * @return true if we should filter this text out, false otherwise
- */
- private static boolean shouldFilterOut(final String text, final int script) {
- if (TextUtils.isEmpty(text) || text.length() <= 1) return true;
-
- // TODO: check if an equivalent processing can't be done more quickly with a
- // compiled regexp.
- // Filter by first letter
- final int firstCodePoint = text.codePointAt(0);
- // Filter out words that don't start with a letter or an apostrophe
- if (!isLetterCheckableByLanguage(firstCodePoint, script)
- && '\'' != firstCodePoint) return true;
-
- // Filter contents
- final int length = text.length();
- int letterCount = 0;
- for (int i = 0; i < length; i = text.offsetByCodePoints(i, 1)) {
- final int codePoint = text.codePointAt(i);
- // Any word containing a '@' is probably an e-mail address
- // Any word containing a '/' is probably either an ad-hoc combination of two
- // words or a URI - in either case we don't want to spell check that
- if ('@' == codePoint || '/' == codePoint) return true;
- if (isLetterCheckableByLanguage(codePoint, script)) ++letterCount;
- }
- // Guestimate heuristic: perform spell checking if at least 3/4 of the characters
- // in this word are letters
- return (letterCount * 4 < length * 3);
- }
-
- private SentenceSuggestionsInfo fixWronglyInvalidatedWordWithSingleQuote(
- TextInfo ti, SentenceSuggestionsInfo ssi) {
- final String typedText = ti.getText();
- if (!typedText.contains(SINGLE_QUOTE)) {
- return null;
- }
- final int N = ssi.getSuggestionsCount();
- final ArrayList<Integer> additionalOffsets = new ArrayList<Integer>();
- final ArrayList<Integer> additionalLengths = new ArrayList<Integer>();
- final ArrayList<SuggestionsInfo> additionalSuggestionsInfos =
- new ArrayList<SuggestionsInfo>();
- String currentWord = null;
- for (int i = 0; i < N; ++i) {
- final SuggestionsInfo si = ssi.getSuggestionsInfoAt(i);
- final int flags = si.getSuggestionsAttributes();
- if ((flags & SuggestionsInfo.RESULT_ATTR_IN_THE_DICTIONARY) == 0) {
- continue;
- }
- final int offset = ssi.getOffsetAt(i);
- final int length = ssi.getLengthAt(i);
- final String subText = typedText.substring(offset, offset + length);
- final String prevWord = currentWord;
- currentWord = subText;
- if (!subText.contains(SINGLE_QUOTE)) {
- continue;
- }
- final String[] splitTexts = subText.split(SINGLE_QUOTE, -1);
- if (splitTexts == null || splitTexts.length <= 1) {
- continue;
- }
- final int splitNum = splitTexts.length;
- for (int j = 0; j < splitNum; ++j) {
- final String splitText = splitTexts[j];
- if (TextUtils.isEmpty(splitText)) {
- continue;
- }
- if (mSuggestionsCache.getSuggestionsFromCache(
- splitText, prevWord) == null) {
- continue;
- }
- final int newLength = splitText.length();
- // Neither RESULT_ATTR_IN_THE_DICTIONARY nor RESULT_ATTR_LOOKS_LIKE_TYPO
- final int newFlags = 0;
- final SuggestionsInfo newSi = new SuggestionsInfo(newFlags, EMPTY_STRING_ARRAY);
- newSi.setCookieAndSequence(si.getCookie(), si.getSequence());
- if (DBG) {
- Log.d(TAG, "Override and remove old span over: "
- + splitText + ", " + offset + "," + newLength);
- }
- additionalOffsets.add(offset);
- additionalLengths.add(newLength);
- additionalSuggestionsInfos.add(newSi);
- }
- }
- final int additionalSize = additionalOffsets.size();
- if (additionalSize <= 0) {
- return null;
- }
- final int suggestionsSize = N + additionalSize;
- final int[] newOffsets = new int[suggestionsSize];
- final int[] newLengths = new int[suggestionsSize];
- final SuggestionsInfo[] newSuggestionsInfos = new SuggestionsInfo[suggestionsSize];
- int i;
- for (i = 0; i < N; ++i) {
- newOffsets[i] = ssi.getOffsetAt(i);
- newLengths[i] = ssi.getLengthAt(i);
- newSuggestionsInfos[i] = ssi.getSuggestionsInfoAt(i);
- }
- for (; i < suggestionsSize; ++i) {
- newOffsets[i] = additionalOffsets.get(i - N);
- newLengths[i] = additionalLengths.get(i - N);
- newSuggestionsInfos[i] = additionalSuggestionsInfos.get(i - N);
- }
- return new SentenceSuggestionsInfo(newSuggestionsInfos, newOffsets, newLengths);
- }
-
- @Override
- public SentenceSuggestionsInfo[] onGetSentenceSuggestionsMultiple(
- TextInfo[] textInfos, int suggestionsLimit) {
- final SentenceSuggestionsInfo[] retval = super.onGetSentenceSuggestionsMultiple(
- textInfos, suggestionsLimit);
- if (retval == null || retval.length != textInfos.length) {
- return retval;
- }
- for (int i = 0; i < retval.length; ++i) {
- final SentenceSuggestionsInfo tempSsi =
- fixWronglyInvalidatedWordWithSingleQuote(textInfos[i], retval[i]);
- if (tempSsi != null) {
- retval[i] = tempSsi;
- }
- }
- return retval;
- }
-
- @Override
- public SuggestionsInfo[] onGetSuggestionsMultiple(TextInfo[] textInfos,
- int suggestionsLimit, boolean sequentialWords) {
- final int length = textInfos.length;
- final SuggestionsInfo[] retval = new SuggestionsInfo[length];
- for (int i = 0; i < length; ++i) {
- final String prevWord;
- if (sequentialWords && i > 0) {
- final String prevWordCandidate = textInfos[i - 1].getText();
- // Note that an empty string would be used to indicate the initial word
- // in the future.
- prevWord = TextUtils.isEmpty(prevWordCandidate) ? null : prevWordCandidate;
- } else {
- prevWord = null;
- }
- retval[i] = onGetSuggestions(textInfos[i], prevWord, suggestionsLimit);
- retval[i].setCookieAndSequence(
- textInfos[i].getCookie(), textInfos[i].getSequence());
- }
- return retval;
- }
-
- // Note : this must be reentrant
- /**
- * Gets a list of suggestions for a specific string. This returns a list of possible
- * corrections for the text passed as an argument. It may split or group words, and
- * even perform grammatical analysis.
- */
- @Override
- public SuggestionsInfo onGetSuggestions(final TextInfo textInfo,
- final int suggestionsLimit) {
- return onGetSuggestions(textInfo, null, suggestionsLimit);
- }
-
- private SuggestionsInfo onGetSuggestions(
- final TextInfo textInfo, final String prevWord, final int suggestionsLimit) {
- try {
- final String inText = textInfo.getText();
- final SuggestionsParams cachedSuggestionsParams =
- mSuggestionsCache.getSuggestionsFromCache(inText, prevWord);
- if (cachedSuggestionsParams != null) {
- if (DBG) {
- Log.d(TAG, "Cache hit: " + inText + ", " + cachedSuggestionsParams.mFlags);
- }
- return new SuggestionsInfo(
- cachedSuggestionsParams.mFlags, cachedSuggestionsParams.mSuggestions);
- }
-
- if (shouldFilterOut(inText, mScript)) {
- DictAndProximity dictInfo = null;
- try {
- dictInfo = mDictionaryPool.takeOrGetNull();
- if (null == dictInfo) return getNotInDictEmptySuggestions();
- return dictInfo.mDictionary.isValidWord(inText) ?
- getInDictEmptySuggestions() : getNotInDictEmptySuggestions();
- } finally {
- if (null != dictInfo) {
- if (!mDictionaryPool.offer(dictInfo)) {
- Log.e(TAG, "Can't re-insert a dictionary into its pool");
- }
- }
- }
- }
- final String text = inText.replaceAll(APOSTROPHE, SINGLE_QUOTE);
-
- // TODO: Don't gather suggestions if the limit is <= 0 unless necessary
- final SuggestionsGatherer suggestionsGatherer = new SuggestionsGatherer(text,
- mService.mSuggestionThreshold, mService.mRecommendedThreshold,
- suggestionsLimit);
- final WordComposer composer = new WordComposer();
- final int length = text.length();
- for (int i = 0; i < length; i = text.offsetByCodePoints(i, 1)) {
- final int codePoint = text.codePointAt(i);
- // The getXYForCodePointAndScript method returns (Y << 16) + X
- final int xy = SpellCheckerProximityInfo.getXYForCodePointAndScript(
- codePoint, mScript);
- if (SpellCheckerProximityInfo.NOT_A_COORDINATE_PAIR == xy) {
- composer.add(codePoint, WordComposer.NOT_A_COORDINATE,
- WordComposer.NOT_A_COORDINATE);
- } else {
- composer.add(codePoint, xy & 0xFFFF, xy >> 16);
- }
- }
-
- final int capitalizeType = getCapitalizationType(text);
- boolean isInDict = true;
- DictAndProximity dictInfo = null;
- try {
- dictInfo = mDictionaryPool.takeOrGetNull();
- if (null == dictInfo) return getNotInDictEmptySuggestions();
- final ArrayList<SuggestedWordInfo> suggestions = dictInfo.mDictionary.getWords(
- composer, prevWord, dictInfo.mProximityInfo);
- for (final SuggestedWordInfo suggestion : suggestions) {
- final String suggestionStr = suggestion.mWord.toString();
- suggestionsGatherer.addWord(suggestionStr.toCharArray(), null, 0,
- suggestionStr.length(), suggestion.mScore);
- }
- isInDict = dictInfo.mDictionary.isValidWord(text);
- if (!isInDict && CAPITALIZE_NONE != capitalizeType) {
- // We want to test the word again if it's all caps or first caps only.
- // If it's fully down, we already tested it, if it's mixed case, we don't
- // want to test a lowercase version of it.
- isInDict = dictInfo.mDictionary.isValidWord(text.toLowerCase(mLocale));
- }
- } finally {
- if (null != dictInfo) {
- if (!mDictionaryPool.offer(dictInfo)) {
- Log.e(TAG, "Can't re-insert a dictionary into its pool");
- }
- }
- }
-
- final SuggestionsGatherer.Result result = suggestionsGatherer.getResults(
- capitalizeType, mLocale);
-
- if (DBG) {
- Log.i(TAG, "Spell checking results for " + text + " with suggestion limit "
- + suggestionsLimit);
- Log.i(TAG, "IsInDict = " + isInDict);
- Log.i(TAG, "LooksLikeTypo = " + (!isInDict));
- Log.i(TAG, "HasRecommendedSuggestions = " + result.mHasRecommendedSuggestions);
- if (null != result.mSuggestions) {
- for (String suggestion : result.mSuggestions) {
- Log.i(TAG, suggestion);
- }
- }
- }
-
- final int flags =
- (isInDict ? SuggestionsInfo.RESULT_ATTR_IN_THE_DICTIONARY
- : SuggestionsInfo.RESULT_ATTR_LOOKS_LIKE_TYPO)
- | (result.mHasRecommendedSuggestions
- ? SuggestionsInfoCompatUtils
- .getValueOf_RESULT_ATTR_HAS_RECOMMENDED_SUGGESTIONS()
- : 0);
- final SuggestionsInfo retval = new SuggestionsInfo(flags, result.mSuggestions);
- mSuggestionsCache.putSuggestionsToCache(text, prevWord, result.mSuggestions, flags);
- return retval;
- } catch (RuntimeException e) {
- // Don't kill the keyboard if there is a bug in the spell checker
- if (DBG) {
- throw e;
- } else {
- Log.e(TAG, "Exception while spellcheking: " + e);
- return getNotInDictEmptySuggestions();
- }
- }
- }
- }
}