org.databene.commons.StringUtil Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of databene-commons Show documentation
Show all versions of databene-commons Show documentation
'databene commons' is an open source Java library by Volker Bergmann.
It provides extensions to the Java core library by utility classes, abstract concepts
and concrete implementations.
/*
* Copyright (C) 2004-2015 Volker Bergmann ([email protected]).
* All rights reserved.
*
* 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 org.databene.commons;
import java.util.*;
/**
* Provides String related utility operations.
* Created: 05.07.2006 22:45:12
* @author Volker Bergmann
*/
public final class StringUtil {
/** Bell character (BEL, 0x07) */
public static final String BEL = String.valueOf((char) 7);
/** Backspace character (BS, 0x08) */
public static final String BS = String.valueOf((char) 8);
/** Horizontal Tab character (HT, 0x09) */
public static final String HT = String.valueOf('\t');
/** Line Feed character (LF, 0x0A) */
public static final String LF = String.valueOf('\n');
/** Vertical Tab character (VT, 0x0B) */
public static final String VT = String.valueOf('\u000B');
/** character (FF, 0x0C) */
public static final String FF = String.valueOf('\f');
/** character (CR, 0x0D) */
public static final String CR = String.valueOf('\r');
/**
* Tells if a String is null or isEmpty.
* @param s the string argument to check
* @return true if the String is null or isEmpty, otherwise false.
*/
public static boolean isEmpty(CharSequence s) {
return s == null || s.length() == 0;
}
/**
* Returns the suffix of a String. If the last character is a
* separator, or if no separator was found, the string is assumed
* to have no suffix.
* @param name the String to check
* @param separator the character that separates name from suffix
* @return a suffix if one exists, else null.
*/
public static String suffix(String name, char separator) {
if (name == null)
return null;
int separatorIndex = name.lastIndexOf(separator);
if (separatorIndex < 0 || separatorIndex == name.length() - 1)
return null;
else
return name.substring(separatorIndex + 1);
}
/**
* Returns the last token of a list in string representation.
* If the string does not contain the separator, the string itself
* is the token. If the string ends with a separator, the token is
* null.
* @param name the name to parse
* @param separator the character that separates the tokens
* @return the last token
*/
public static String lastToken(String name, char separator) {
if (name == null)
return null;
int separatorIndex = name.lastIndexOf(separator);
if (separatorIndex < 0)
return name;
else if (separatorIndex == name.length() - 1)
return null;
else
return name.substring(separatorIndex + 1);
}
/**
* Splits a list's String representation into tokens.
* @param text the String representation of a list.
* @param separator the character used to separate tokens
* @return an array of tokens.
*/
public static String[] tokenize(String text, char separator) {
if (text == null)
return null;
if (text.length() == 0)
return new String[] { "" };
int i = 0;
int sep;
List list = new ArrayList();
while ((sep = text.indexOf(separator, i)) >= 0) {
if (sep == i) {
list.add("");
i++;
} else {
list.add(text.substring(i, sep));
i = sep + 1;
}
}
if (i < text.length())
list.add(text.substring(i));
else if (text.endsWith("" + separator))
list.add("");
String[] tokens = new String[list.size()];
for (i = 0; i < tokens.length; i++)
tokens[i] = list.get(i);
return tokens;
}
public static String[] splitAndTrim(String list, char separator) {
return StringUtil.trimAll(list.split(String.valueOf(separator)));
}
public static String normalize(String s) {
char[] srcBuffer = new char[s.length()];
s.getChars(0, s.length(), srcBuffer, 0);
char[] dstBuffer = new char[s.length()];
int dstIndex = 0;
for (char c : srcBuffer)
if (c >= 16)
dstBuffer[dstIndex++] = c;
return new String(dstBuffer, 0, dstIndex);
}
public static StringBuilder appendLeftAligned(StringBuilder builder, String text, int columns) {
builder.append(text);
int columnsToInsert = columns - text.length();
for (int i = 0; i < columnsToInsert; i++)
builder.append(' ');
return builder;
}
public static String increment(String text) {
if (text == null || text.length() == 0)
return "0";
char[] chars = new char[text.length()];
text.getChars(0, text.length(), chars, 0);
chars = increment(chars, chars.length - 1);
return String.valueOf(chars);
}
public static char[] increment(char[] chars, int index) {
char c = chars[index];
switch (c) {
case '9':
chars[index] = 'a';
break;
case 'z': case 'Z':
if (index > 0) {
chars[index] = '0';
chars = increment(chars, index - 1);
} else {
char[] result = new char[chars.length + 1];
Arrays.fill(result, '0');
result[0] = '1';
return result;
}
break;
default:
chars[index]++;
}
return chars;
}
/** interprets an nbsp as space character
* @param c the character to check
* @return true if the character is a whitespace, otherwise false. */
public static boolean isWhitespace(char c) {
return Character.isWhitespace(c) || c == 160;
}
/** Trims a String by removing white spaces (including nbsp) from left and right.
* @param s the String to trim
* @return the trimmed string */
public static String trim(String s) {
if (s == null || s.length() == 0)
return s;
int beginIndex;
for (beginIndex = 0; beginIndex < s.length() && isWhitespace(s.charAt(beginIndex)); beginIndex++) {
}
int endIndex;
for (endIndex = s.length() - 1; endIndex > 0 && isWhitespace(s.charAt(endIndex)); endIndex--) {
}
if (beginIndex > endIndex)
return "";
return s.substring(beginIndex, endIndex + 1);
}
/**
* Trims all String in the array.
* @param array an array of the Strings to trim
* @return the same array but with its elements trimmed
* @since 0.2.05
*/
public static String[] trimAll(String[] array) {
for (int i = 0; i < array.length; i++)
array[i] = trim(array[i]);
return array;
}
/** Returns the platform dependent line separator
* @return the system's line separator */
public static String lineSeparator() {
return System.getProperty("line.separator");
}
public static boolean contains(String s, char c) {
return s.indexOf(c) >= 0;
}
public static String remove(String s, String chars) {
if (s == null)
return s;
StringBuilder result = new StringBuilder(s.length());
for (int i = 0; i < s.length(); i++) {
if (!(contains(chars, s.charAt(i))))
result.append(s.charAt(i));
}
return result.toString();
}
public static String nullToEmpty(String text) {
return (text != null ? text : "");
}
public static String substituteNull(String value, String substitution) {
return (value != null ? value : substitution);
}
public static String normalizeSpace(String s) {
if (s == null || s.length() == 0)
return s;
s = trim(s);
if (s.length() == 0)
return s;
char lastChar = s.charAt(0);
StringBuilder result = new StringBuilder().append(lastChar);
for (int i = 1; i < s.length(); i++) {
char c = s.charAt(i);
if (!isWhitespace(c))
result.append(c);
else if (!(isWhitespace(lastChar)))
result.append(' ');
lastChar = c;
}
return result.toString();
}
public static String trimEnd(String s) {
if (s == null || s.length() == 0)
return s;
int lastIndex = s.length() - 1;
while (lastIndex >= 0 && isWhitespace(s.charAt(lastIndex)))
lastIndex--;
if (lastIndex < 0)
return "";
return s.substring(0, lastIndex + 1);
}
public static int countChars(String s, char c) {
int count = 0;
for (int i = 0; i < s.length(); i++)
if (s.charAt(i) == c)
count++;
return count;
}
public static String[] toArray(List list) {
String[] result = new String[list.size()];
for (int i = 0; i < list.size(); i++)
result[i] = list.get(i);
return result;
}
public static String[][] toArrayArray(List> list) {
String[][] result = new String[list.size()][];
for (int i = 0; i < list.size(); i++)
result[i] = toArray(list.get(i));
return result;
}
public static String padLeft(String text, int length, char c) {
if (text == null)
text = "";
int textLength = text.length();
if (textLength > length)
throw new IllegalArgumentException("Text is too long (" + textLength + ") to be padded to " + length + " columns");
char[] chars = new char[length];
int offset = length - textLength;
fill(chars, 0, offset, c);
getChars(0, textLength, text, chars, offset);
return new String(chars);
}
public static String padRight(String text, int length, char c) {
if (text == null)
text = "";
int textLength = text.length();
if (textLength > length)
throw new IllegalArgumentException("Text is too long (" + textLength + ") to be padded to " + length + " columns");
char[] chars = new char[length];
fill(chars, textLength, length, c);
getChars(0, textLength, text, chars, 0);
return new String(chars);
/*
int textLength = text.length();
if (textLength > length)
throw new IllegalArgumentException("Text is too long (" + textLength + ") to be padded to " + length + " columns");
int padLength = length - textLength;
char[] padChars = new char[padLength];
if (padLength < 20) {
for (int i = 0; i < padLength; i++)
padChars[i] = c;
} else if (c == ' ' && padLength <= WHITESPACE_BUFFER.length) {
System.arraycopy(WHITESPACE_BUFFER, 0, padChars, 0, padLength);
} else {
Arrays.fill(padChars, 0, padLength, c);
}
return text.concat(new String(padChars));
*/
}
private static final int BUFFER_SIZE = 1024;
private static char[] WHITESPACE_BUFFER;
static {
WHITESPACE_BUFFER = new char[BUFFER_SIZE];
for (int i = 0; i < WHITESPACE_BUFFER.length; i++)
WHITESPACE_BUFFER[i] = ' ';
}
public static char[] fill(char[] chars, int fromIndex, int toIndex, char c) {
int length = toIndex - fromIndex;
if (length < 20) {
for (int i = fromIndex; i < toIndex; i++)
chars[i] = c;
} else if (c != ' ' || length > WHITESPACE_BUFFER.length) {
Arrays.fill(chars, fromIndex, toIndex, c);
} else {
System.arraycopy(WHITESPACE_BUFFER, 0, chars, fromIndex, length);
}
return chars;
}
public static char[] getChars(int srcBegin, int srcEnd, String text, char[] chars, int dstBegin) {
int textLength = text.length();
if (textLength >= 6)
text.getChars(srcBegin, srcEnd, chars, dstBegin);
else
for (int i = 0; i < textLength; i++)
chars[i + dstBegin] = text.charAt(i);
return chars;
}
public static char[] getChars(String s) {
char[] chars = new char[s.length()];
return getChars(0, chars.length, s, chars, 0);
}
public static char[] getChars(StringBuilder builder) {
char[] chars = new char[builder.length()];
builder.getChars(0, builder.length(), chars, 0);
return chars;
}
public static String padString(char c, int length) {
if (length < 0)
throw new IllegalArgumentException("Negative pad length: " + length);
char[] chars = new char[length];
Arrays.fill(chars, 0, length, c);
return new String(chars);
}
public static String trimRight(String source) {
if (source == null)
return null;
int i = source.length() - 1;
while (i >= 0 && isWhitespace(source.charAt(i)))
i--;
return source.substring(0, i + 1);
}
public static String trimRight(String source, char padChar) {
if (source == null)
return null;
int i = source.length() - 1;
while (i >= 0 && source.charAt(i) == padChar)
i--;
return source.substring(0, i + 1);
}
public static String trimLeft(String source, char padChar) {
if (source == null)
return null;
int i = 0;
while (i < source.length() && source.charAt(i) == padChar)
i++;
return source.substring(i, source.length());
}
public static String trim(String source, char padChar) {
if (source == null)
return null;
int i0 = 0;
while (i0 < source.length() && source.charAt(i0) == padChar)
i0++;
if (i0 == source.length())
return "";
int i1 = source.length() - 1;
while (i1 > i0 && source.charAt(i1) == padChar)
i1--;
return source.substring(i0, i1 + 1);
}
public static String[] toLowerCase(String[] src) {
String[] dst = new String[src.length];
for (int i = 0; i < src.length; i++)
dst[i] = src[i].toLowerCase();
return dst;
}
public static int indexOfIgnoreCase(String searched, String ... candidates) {
for (int i = 0; i < candidates.length; i++)
if (searched.equalsIgnoreCase(candidates[i]))
return i;
return -1;
}
public static char lastChar(String word) {
return word.charAt(word.length() - 1);
}
/**
* Makes the first character of a String uppercase.
* @param text the text to convert
* @return a text that starts with a uppercase letter
*/
public static String capitalize(String text) {
return text.substring(0, 1).toUpperCase() + text.substring(1);
}
/**
* Makes the first character of a String lowercase.
* @param text the text to convert
* @return a text that starts with a lowercase letter
* @since 0.2.04
*/
public static String uncapitalize(String text) {
return text.substring(0, 1).toLowerCase() + text.substring(1);
}
public static String[] splitOnFirstSeparator(String path, char separator) {
if (path == null)
return new String[] { null, null };
if (path.indexOf(separator) < 0)
return new String[] { path, null };
int sepIndex = path.indexOf(separator);
return splitAroundSeparator(path, sepIndex);
}
public static String[] splitOnLastSeparator(String path, char separator) {
if (path == null)
return new String[] { null, null };
int sepIndex = path.lastIndexOf(separator);
return splitAroundSeparator(path, sepIndex);
}
public static String[] splitAroundSeparator(String path, int sepIndex) {
if (sepIndex < 0)
return new String[] { null, path };
else if (sepIndex == 0)
return new String[] { "", path.substring(1) };
else if (sepIndex == path.length() - 1)
return new String[] { path.substring(0, path.length() - 1), "" };
else
return new String[] { path.substring(0, sepIndex), path.substring(sepIndex + 1) };
}
public static String concat(Character separator, String... parts) {
if (parts == null)
return "";
StringBuilder builder = new StringBuilder();
for (String part : parts) {
if (!isEmpty(part)) {
if (builder.length() > 0 && separator != null)
builder.append(separator);
builder.append(part);
}
}
return builder.toString();
}
public static boolean equalsIgnoreCase(String s1, String s2) {
return (s1 != null ? s1.equalsIgnoreCase(s2) : s2 == null);
}
public static boolean equalsIgnoreCase(String[] a1, String[] a2) {
if (a1.length != a2.length)
return false;
for (int i = 0; i < a1.length; i++)
if (!equalsIgnoreCase(a1[i], a2[i]))
return false;
return true;
}
public static boolean equalsIgnoreCase(Set set1, Set set2) {
if (set1.size() != set2.size())
return false;
for (String s1 : set1) {
boolean found = false;
for (String s2 : set2)
if (equalsIgnoreCase(s1, s2))
found = true;
if (!found)
return false;
}
return true;
}
public static boolean containsIgnoreCase(String searchedWord, Collection words) {
for (String name : words)
if (name.equalsIgnoreCase(searchedWord))
return true;
return false;
}
public static boolean containsIgnoreCase(String searchedWord, String[] words) {
for (String name : words)
if (name.equalsIgnoreCase(searchedWord))
return true;
return false;
}
public static boolean startsWithIgnoreCase(String text, String prefix) {
if (text == null)
return (prefix == null);
if (prefix == null)
return false;
return text.toLowerCase().startsWith(prefix.toLowerCase());
}
public static boolean endsWithIgnoreCase(String text, String suffix) {
if (text == null)
return (suffix == null);
if (suffix == null)
return false;
return text.toLowerCase().endsWith(suffix.toLowerCase());
}
public static String normalizeName(final String name) {
if (StringUtil.isEmpty(name))
return name;
final int NONE = -1;
final int WS = 0;
final int SPECIAL = 1;
final int INITIAL = 2;
final int SUBSEQUENT = 3;
StringBuilder builder = new StringBuilder(name.length());
StringCharacterIterator iterator = new StringCharacterIterator(name);
iterator.skipWhitespace();
int prevType = NONE;
while (iterator.hasNext()) {
char c = iterator.next();
int type;
if (Character.isWhitespace(c))
type = WS;
else if (!Character.isLetter(c))
type = SPECIAL;
else if (prevType == INITIAL)
type = SUBSEQUENT;
else if (prevType == NONE || prevType == WS || prevType == SPECIAL)
type = INITIAL;
else
type = prevType;
if (prevType == WS && (type == INITIAL || type == SPECIAL))
builder.append(' ');
switch (type) {
case INITIAL: builder.append(Character.toUpperCase(c)); break;
case SUBSEQUENT:builder.append(Character.toLowerCase(c)); break;
case SPECIAL: builder.append(c); break;
case WS: break;
default: throw new RuntimeException("Internal error");
}
prevType = type;
}
return builder.toString();
}
public static String escape(String text) {
return escape(text, false, false);
}
/**
* Escapes a string in C/C++/Java style.
* @param text the text to escape
* @param escapeSingleQuotes true if single quotes shall be escaped, otherwise false
* @param escapeDoubleQuotes true if double quotes shall be escaped, otherwise false
* @return the escaped string
* @see "http://en.wikipedia.org/wiki/ASCII"
*/
public static String escape(String text, boolean escapeSingleQuotes, boolean escapeDoubleQuotes) {
if (text == null)
return null;
text = text.replace("\\", "\\\\"); // keep this first, otherwise all other escapes will be doubled
text = text.replace(BEL, "\\u0007");
text = text.replace(BS, "\\u0008");
text = text.replace(CR, "\\r");
text = text.replace(LF, "\\n");
text = text.replace(HT, "\\t");
text = text.replace(VT, "\\u000B");
text = text.replace(FF, "\\f");
if (escapeSingleQuotes)
text = text.replace("'", "\\'");
if (escapeDoubleQuotes)
text = text.replace("\"", "\\\"");
return text;
}
/**
* Unescapes a string in C/C++/Java style.
* @param text the text to unescape
* @return the unescaped text
* @see "http://en.wikipedia.org/wiki/ASCII"
*/
public static String unescape(String text) {
if (text == null)
return null;
StringBuilder builder = new StringBuilder(text.length());
for (int i = 0; i < text.length(); i++) {
char c = text.charAt(i);
if (c != '\\')
builder.append(c);
else if (i < text.length() - 1) {
c = text.charAt(++i);
switch (c) {
case '\'' : builder.append('\''); break;
case '"' : builder.append('"'); break;
case '{' : builder.append('{'); break;
case '}' : builder.append('}'); break;
case 'a' : builder.append(BEL); break;
case 'b' : builder.append(BS); break;
case 'r' : builder.append(CR); break;
case 'n' : builder.append(LF); break;
case 't' : builder.append(HT); break;
case 'f' : builder.append(FF); break;
case 'u' :
long n = Long.parseLong(text.substring(i + 1, i + 5), 16);
builder.append((char) n);
i += 4;
break;
default : builder.append(c); break;
}
} else
builder.append('\\');
}
return builder.toString();
}
public static String replaceTokens(String src, String token, String ... values) {
StringBuilder builder = new StringBuilder();
int paramIndex = 0;
int srcIndex = 0;
while (srcIndex < src.length()) {
int i = src.indexOf(token, srcIndex);
if (i >= 0) {
builder.append(src.substring(srcIndex, i));
builder.append(values[paramIndex++]);
srcIndex = i + token.length();
} else {
builder.append(src.substring(srcIndex));
break;
}
}
return builder.toString();
}
public static String emptyToNull(String s) {
if (s == null || s.length() == 0)
return null;
String trimmed = trim(s);
return (trimmed.length() != 0 ? s : null);
}
public static String trimmedEmptyToNull(String s) {
if (s == null || s.length() == 0)
return null;
String trimmed = trim(s);
return (trimmed.length() != 0 ? trimmed : null);
}
public static String removeSection(String text, String beginMark, String endMark) {
if (StringUtil.isEmpty(text))
return text;
int beginIndex = text.indexOf(beginMark);
int endIndex = text.indexOf(endMark);
if (beginIndex < 0 || endIndex < 0 || beginIndex + beginMark.length() > endIndex + endMark.length())
return text;
return text.substring(0, beginIndex) + text.substring(endIndex + endMark.length());
}
public static String normalizeLineSeparators(String text, String lineSeparator) {
if (StringUtil.isEmpty(text))
return text;
StringBuilder builder = new StringBuilder();
StringCharacterIterator iterator = new StringCharacterIterator(text);
while (iterator.hasNext()) {
char c = iterator.next();
if (c != '\r' && c != '\n')
builder.append(c);
else {
// swallow the \n part of of \r\n
if (c == '\r' && iterator.hasNext()) {
char c2 = iterator.next();
if (c2 != '\n') // oops, it was only a \r
iterator.pushBack();
}
builder.append(lineSeparator);
}
}
return builder.toString();
}
public static String extract(String text, String beginMark, String endMark) {
int beginIndex = (beginMark != null ? text.indexOf(beginMark) + beginMark.length() : 0);
if (endMark != null) {
int endIndex = text.indexOf(endMark, beginIndex + 1);
return (beginIndex >= 0 && endIndex >= 0 && endIndex > beginIndex ? text.substring(beginIndex, endIndex) : null);
} else {
return (beginIndex >= 0 ? text.substring(beginIndex) : null);
}
}
public static String buildPhrase(String... parts) {
if (parts == null)
return "";
StringBuilder builder = new StringBuilder();
for (String part : parts) {
if (!StringUtil.isEmpty(part)) {
if (builder.length() > 0)
builder.append(' ');
builder.append(part);
}
}
return builder.toString();
}
public static String trimLineSeparators(String text) {
if (text == null)
return null;
int start = 0;
while (start < text.length() && isLineSeparatorChar(text.charAt(start)))
start++;
int end = text.length();
while (end > 0 && (text.charAt(end - 1) == '\r' || text.charAt(end - 1) == '\n'))
end--;
return text.substring(start, end);
}
public static List splitLines(String text) {
if (text == null)
return null;
List lines = new ArrayList();
int TEXT = 0;
int CR = 1;
int LF = 2;
int mode = TEXT;
StringBuilder builder = new StringBuilder();
for (int i = 0; i < text.length(); i++) {
char c = text.charAt(i);
if (c == '\r') {
lines.add(builder.toString());
builder.delete(0, builder.length());
mode = CR;
} else if (c == '\n') {
if (mode != CR) {
// only add a line if it is a standalone LF (no CRLF)
lines.add(builder.toString());
builder.delete(0, builder.length());
}
mode = LF;
} else {
mode = TEXT;
builder.append(c);
}
}
if (text.length() == 0 || mode != TEXT || builder.length() > 0)
lines.add(builder.toString());
return lines;
}
public static String removeEmptyLines(String text) {
if (text == null)
return null;
List lines = splitLines(text);
Iterator iterator = lines.iterator();
while (iterator.hasNext())
if (iterator.next().trim().length() == 0)
iterator.remove();
String sep = lineSeparatorUsedIn(text);
return ArrayFormat.format(sep, lines.toArray());
}
private static String lineSeparatorUsedIn(String text) {
int i = 0;
while (i < text.length() && !isLineSeparatorChar(text.charAt(i))) // search the first line separator char
i++;
if (i == text.length()) // if no line sep was found, then return null
return null;
char c1 = text.charAt(i);
if (i == text.length() - 1)
return String.valueOf(c1); // if the found char is the last one in the string, return it, ...
char c2 = text.charAt(i + 1); // ... otherwise check the char that follows
if (isLineSeparatorChar(c2) && c1 != c2)
return "" + c1 + c2; // the line separator consists of two characters
return String.valueOf(c1);
}
public static boolean isLineSeparatorChar(char c) {
return c == '\r' || c == '\n';
}
public static String quoteIfNotNull(String text) {
return (text != null ? "'" + text + "'" : text);
}
public static String getLeadingWhitespace(String line) {
int i;
for (i = 0; i < line.length() && Character.isWhitespace(line.charAt(i)); i++) {
// empty iteration for index counting
}
return line.substring(0, i);
}
public static String limitLength(String text, int maxLength) {
if (text == null)
return null;
else if (text.length() <= maxLength)
return text;
else
return text.substring(0, maxLength);
}
public static String limitLengthWithEllipsis(String text, int maxLength) {
if (text == null)
return null;
else if (text.length() <= maxLength)
return text;
else if (maxLength > 3)
return text.substring(0, maxLength - 3) + "...";
else if (maxLength == 3)
return text.substring(0, 1) + "..";
else if (maxLength > 0)
return text.substring(0, maxLength - 1) + ".";
else
return "";
}
public static String replaceOptionalSuffix(String text, String suffix, String replacement) {
if (text.endsWith(suffix))
return text.substring(0, text.length() - suffix.length()) + replacement;
else
return text;
}
public static String removeSuffixIfPresent(String suffix, String name) {
if (name != null && name.endsWith(suffix))
return name.substring(0, name.length() - suffix.length());
return name;
}
}