improved the quick delete feature not to erase whole paragraphs of text in languages without spaces

app/src/main/java/io/github/sspanak/tt9/ime/TypingHandler.java

@@ -133,18 +133,7 @@ public abstract class TypingHandler extends KeyPadHandler {
 		} else {
 			suggestionOps.commitCurrent(false);
 			mInputMode.reset();
-
+			deleteText(settings.getBackspaceAcceleration() && repeat > 0);
-			int charsToDelete;
-
-			if (settings.getBackspaceAcceleration() && repeat > 0) {
-				charsToDelete = Math.max(textField.getPaddedWordBeforeCursorLength(), 1);
-			} else if (!textSelection.isEmpty()) {
-				charsToDelete = textSelection.length();
-				textSelection.clear(false);
-			} else {
-				charsToDelete = 1;
-			}
-			textField.deleteChars(charsToDelete);
 		}
 
 		if (settings.getBackspaceRecomposing() && repeat == 0 && suggestionOps.isEmpty() && !DictionaryLoader.getInstance(this).isRunning()) {
@@ -251,6 +240,18 @@ public abstract class TypingHandler extends KeyPadHandler {
 	}
 
 
+	private void deleteText(boolean deleteMany) {
+		int charsToDelete = 1;
+		if (deleteMany) {
+			charsToDelete = Math.max(textField.getPaddedWordBeforeCursorLength(), 1);
+		} else if (!textSelection.isEmpty()) {
+			charsToDelete = textSelection.length();
+			textSelection.clear(false);
+		}
+		textField.deleteChars(charsToDelete);
+	}
+
+
 	/**
 	 * determineLanguage
 	 * Restore the last language or auto-select a more appropriate one, if the application hints so.

app/src/main/java/io/github/sspanak/tt9/ime/helpers/TextField.java

@@ -166,7 +166,7 @@ public class TextField extends InputField {
 			return 0;
 		}
 
-		int whitespaceShift = Math.max(before.lastBoundaryIndex(), 0);
+		int whitespaceShift = Math.max(before.lastBoundaryIndex(SettingsStore.BACKSPACE_ACCELERATION_MAX_CHARS_NO_SPACE), 0);
 		return Math.min(before.length() - whitespaceShift, (int) (SettingsStore.BACKSPACE_ACCELERATION_MAX_CHARS * 1.5));
 	}
 

app/src/main/java/io/github/sspanak/tt9/preferences/settings/SettingsStore.java

@@ -7,7 +7,8 @@ public class SettingsStore extends SettingsHotkeys {
 	public SettingsStore(Context context) { super(context); }
 
 	/************* internal settings *************/
-	public static final int BACKSPACE_ACCELERATION_MAX_CHARS = 20;
+	public static final int BACKSPACE_ACCELERATION_MAX_CHARS = 20; // maximum chars to be deleted at once in very long words
+	public static final int BACKSPACE_ACCELERATION_MAX_CHARS_NO_SPACE = 4; // maximum chars to be deleted at once for languages with no spaces
 	public static final int BACKSPACE_ACCELERATION_REPEAT_DEBOUNCE = 5;
 	public final static int CLIPBOARD_PREVIEW_LENGTH = 20;
 	public final static int CUSTOM_WORDS_IMPORT_MAX_LINES = 250;

app/src/main/java/io/github/sspanak/tt9/util/Text.java

@@ -4,6 +4,8 @@ import androidx.annotation.NonNull;
 
 import java.util.List;
 import java.util.Locale;
+import java.util.regex.Matcher;
+import java.util.regex.Pattern;
 
 import io.github.sspanak.tt9.ime.modes.InputMode;
 import io.github.sspanak.tt9.languages.Language;
@@ -14,6 +16,8 @@ public class Text extends TextTools {
 	private final Language language;
 	private final String text;
 
+	private final static Pattern QUICK_DELETE_GROUP = Pattern.compile("(?:([\\s\\u3000]{2,})|([.,、。，،]{2,})|([^、。，\\s\\u3000]*.))$");
+
 
 	public Text(Language language, String text) {
 		this.language = language;
@@ -189,26 +193,38 @@ public class Text extends TextTools {
 	}
 
 
-	public int lastBoundaryIndex() {
+	/**
+	 * Returns the starting index of the last word boundary. This could be start of a word, of a whitespace
+	 * block or of a punctuation block (e.g. "..."). In the case of languages that do not use spaces
+	 * it is not possible to determine where a word starts, so in case the text ends with letters only,
+	 * we assume the last word is at most MAX_WORD_LENGTH_NO_SPACE letters long.
+	 */
+	public int lastBoundaryIndex(final int MAX_WORD_LENGTH_NO_SPACE) {
 		if (text == null || text.length() < 2) {
 			return -1;
 		}
 
-		char lastChar = text.charAt(text.length() - 1);
+		Matcher matcher = QUICK_DELETE_GROUP.matcher(text);
-		char penultimateChar = text.charAt(text.length() - 2);
+		for (int i = matcher.find() ? matcher.groupCount() : 0, nonLetterGroup = 0; i >= 0; i--, nonLetterGroup = 1) {
+			String group = matcher.group(i);
+			if (group == null) {
+				continue;
+			}
 
-		boolean endsWithWhitespaceBlock = Character.isWhitespace(lastChar) && Character.isWhitespace(penultimateChar);
+			if (nonLetterGroup == 1) {
-		boolean endsWithPunctuationBlock = (lastChar == '.' || lastChar == ',') && (penultimateChar == '.' || penultimateChar == ',');
+				return matcher.start();
+			}
 
-		for (int i = text.length() - 1, firstChar = 1; i >= 0; i--, firstChar = 0) {
+			Text gr = new Text(group);
-			char currentChar = text.charAt(i);
+			int codePoints = gr.codePointLength();
 
-			if (
+			// In the writing systems that do not use spaces, the last group is not the last word, but
-				(endsWithPunctuationBlock && currentChar != '.' && currentChar != ',')
+			// it could be an entire sentence or a paragraph. That's why we assume an average word length
-				|| (endsWithWhitespaceBlock && !Character.isWhitespace(currentChar))
+			// of N letters
-				|| (!endsWithWhitespaceBlock && !endsWithPunctuationBlock && firstChar == 0 && Character.isWhitespace(currentChar))
+			if (codePoints > MAX_WORD_LENGTH_NO_SPACE && (TextTools.isChinese(group) || TextTools.isJapanese(group) || TextTools.isThai(group))) {
-			) {
+				return text.length() - gr.substringCodePoints(codePoints - 4, codePoints).length();
-				return i + 1;
+			} else {
+				return matcher.start();
 			}
 		}
 

app/src/main/java/io/github/sspanak/tt9/util/TextTools.java

@@ -17,6 +17,7 @@ public class TextTools {
 	private static final Pattern IS_CHINESE = Pattern.compile("\\p{script=Han}+");
 	private static final Pattern IS_JAPANESE = Pattern.compile("\\p{script=Hiragana}+|\\p{script=Katakana}+|\\p{script=Han}+");
 	private static final Pattern IS_HANGUL = Pattern.compile("[\u1100-\u11FF\u302E-\u302F\u3131-\u318F\u3200-\u321F\u3260-\u327E\uA960-\uA97F\uAC00-\uD7FB\uFFA0-\uFFDF]+");
+	private static final Pattern IS_THAI = Pattern.compile("[\\u0E00-\\u0E7F]+");
 	private static final Pattern NEXT_TO_WORD = Pattern.compile("\\b$");
 	private static final Pattern PREVIOUS_IS_LETTER = Pattern.compile("[\\p{L}\\p{M}](?!\\n)$");
 	private static final Pattern START_OF_SENTENCE = Pattern.compile("(?<!\\.)(^|[.?!؟¿¡])\\s+$");
@@ -62,6 +63,11 @@ public class TextTools {
 	}
 
 
+	public static boolean isThai(String str) {
+		return str != null && IS_THAI.matcher(str).find();
+	}
+
+
 	public static int indexOfIgnoreCase(List<String> list, String str) {
 		for (int i = 0, size = list != null && str != null ? list.size() : 0; i < size; i++) {
 			if (list.get(i).equalsIgnoreCase(str)) {