futokb/java/src/com/android/inputmethod/latin/StringUtils.java

198 lines
7.5 KiB
Java
Raw Normal View History

/*
* Copyright (C) 2012 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.android.inputmethod.latin;
import android.text.TextUtils;
import java.util.ArrayList;
import java.util.Locale;
public class StringUtils {
private StringUtils() {
// This utility class is not publicly instantiable.
}
public static int codePointCount(String text) {
if (TextUtils.isEmpty(text)) return 0;
return text.codePointCount(0, text.length());
}
public static boolean containsInArray(String key, String[] array) {
for (final String element : array) {
if (key.equals(element)) return true;
}
return false;
}
public static boolean containsInCsv(String key, String csv) {
if (TextUtils.isEmpty(csv)) return false;
return containsInArray(key, csv.split(","));
}
public static String appendToCsvIfNotExists(String key, String csv) {
if (TextUtils.isEmpty(csv)) return key;
if (containsInCsv(key, csv)) return csv;
return csv + "," + key;
}
public static String removeFromCsvIfExists(String key, String csv) {
if (TextUtils.isEmpty(csv)) return "";
final String[] elements = csv.split(",");
if (!containsInArray(key, elements)) return csv;
final ArrayList<String> result = new ArrayList<String>(elements.length - 1);
for (final String element : elements) {
if (!key.equals(element)) result.add(element);
}
return TextUtils.join(",", result);
}
/**
* Returns true if a and b are equal ignoring the case of the character.
* @param a first character to check
* @param b second character to check
* @return {@code true} if a and b are equal, {@code false} otherwise.
*/
public static boolean equalsIgnoreCase(char a, char b) {
// Some language, such as Turkish, need testing both cases.
return a == b
|| Character.toLowerCase(a) == Character.toLowerCase(b)
|| Character.toUpperCase(a) == Character.toUpperCase(b);
}
/**
* Returns true if a and b are equal ignoring the case of the characters, including if they are
* both null.
* @param a first CharSequence to check
* @param b second CharSequence to check
* @return {@code true} if a and b are equal, {@code false} otherwise.
*/
public static boolean equalsIgnoreCase(CharSequence a, CharSequence b) {
if (a == b)
return true; // including both a and b are null.
if (a == null || b == null)
return false;
final int length = a.length();
if (length != b.length())
return false;
for (int i = 0; i < length; i++) {
if (!equalsIgnoreCase(a.charAt(i), b.charAt(i)))
return false;
}
return true;
}
/**
* Returns true if a and b are equal ignoring the case of the characters, including if a is null
* and b is zero length.
* @param a CharSequence to check
* @param b character array to check
* @param offset start offset of array b
* @param length length of characters in array b
* @return {@code true} if a and b are equal, {@code false} otherwise.
* @throws IndexOutOfBoundsException
* if {@code offset < 0 || length < 0 || offset + length > data.length}.
* @throws NullPointerException if {@code b == null}.
*/
public static boolean equalsIgnoreCase(CharSequence a, char[] b, int offset, int length) {
if (offset < 0 || length < 0 || length > b.length - offset)
throw new IndexOutOfBoundsException("array.length=" + b.length + " offset=" + offset
+ " length=" + length);
if (a == null)
return length == 0; // including a is null and b is zero length.
if (a.length() != length)
return false;
for (int i = 0; i < length; i++) {
if (!equalsIgnoreCase(a.charAt(i), b[offset + i]))
return false;
}
return true;
}
/**
* Returns true if cs contains any upper case characters.
*
* @param cs the CharSequence to check
* @return {@code true} if cs contains any upper case characters, {@code false} otherwise.
*/
public static boolean hasUpperCase(final CharSequence cs) {
final int length = cs.length();
for (int i = 0, cp = 0; i < length; i += Character.charCount(cp)) {
cp = Character.codePointAt(cs, i);
if (Character.isUpperCase(cp)) {
return true;
}
}
return false;
}
/**
* Remove duplicates from an array of strings.
*
* This method will always keep the first occurrence of all strings at their position
* in the array, removing the subsequent ones.
*/
public static void removeDupes(final ArrayList<CharSequence> suggestions) {
if (suggestions.size() < 2) return;
int i = 1;
// Don't cache suggestions.size(), since we may be removing items
while (i < suggestions.size()) {
final CharSequence cur = suggestions.get(i);
// Compare each suggestion with each previous suggestion
for (int j = 0; j < i; j++) {
CharSequence previous = suggestions.get(j);
if (TextUtils.equals(cur, previous)) {
suggestions.remove(i);
i--;
break;
}
}
i++;
}
}
public static String toTitleCase(String s, Locale locale) {
if (s.length() <= 1) {
// TODO: is this really correct? Shouldn't this be s.toUpperCase()?
return s;
}
// TODO: fix the bugs below
// - This does not work for Greek, because it returns upper case instead of title case.
// - It does not work for Serbian, because it fails to account for the "lj" character,
// which should be "Lj" in title case and "LJ" in upper case.
// - It does not work for Dutch, because it fails to account for the "ij" digraph, which
// are two different characters but both should be capitalized as "IJ" as if they were
// a single letter.
// - It also does not work with unicode surrogate code points.
return s.toUpperCase(locale).charAt(0) + s.substring(1);
}
public static int[] toCodePointArray(final String string) {
final char[] characters = string.toCharArray();
final int length = characters.length;
final int[] codePoints = new int[Character.codePointCount(characters, 0, length)];
int codePoint = Character.codePointAt(characters, 0);
int dsti = 0;
for (int srci = Character.charCount(codePoint);
srci < length; srci += Character.charCount(codePoint), ++dsti) {
codePoints[dsti] = codePoint;
codePoint = Character.codePointAt(characters, srci);
}
codePoints[dsti] = codePoint;
return codePoints;
}
}