brooklyn.util.text.Strings Maven / Gradle / Ivy
Show all versions of brooklyn-utils-common Show documentation
/*
* Copyright (c) 2009-2013 Cloudsoft Corporation Ltd.
*/
package brooklyn.util.text;
import java.text.DecimalFormat;
import java.text.NumberFormat;
import java.util.Map;
import javax.annotation.Nullable;
import brooklyn.util.collections.MutableMap;
import brooklyn.util.time.Time;
import com.google.common.base.CharMatcher;
import com.google.common.base.Functions;
import com.google.common.base.Preconditions;
import com.google.common.base.Predicate;
import com.google.common.base.Supplier;
import com.google.common.base.Suppliers;
public class Strings {
/**
* Checks if the given string is null or is an empty string.
* Useful for pre-String.isEmpty. And useful for StringBuilder etc.
*
* @param s the String to check
* @return true if empty or null, false otherwise.
*
* @see #isNonEmpty(CharSequence)
* @see #isBlank(CharSequence)
* @see #isNonBlank(CharSequence)
*/
public static boolean isEmpty(CharSequence s) {
// Note guava has com.google.common.base.Strings.isNullOrEmpty(String),
// but that is just for String rather than CharSequence
return s == null || s.length()==0;
}
/**
* Checks if the given string is empty or only consists of whitespace.
*
* @param s the String to check
* @return true if blank, empty or null, false otherwise.
*
* @see #isEmpty(CharSequence)
* @see #isNonEmpty(CharSequence)
* @see #isNonBlank(CharSequence)
*/
public static boolean isBlank(CharSequence s) {
return isEmpty(s) || CharMatcher.WHITESPACE.matchesAllOf(s);
}
/**
* The inverse of {@link #isEmpty(CharSequence)}.
*
* @param s the String to check
* @return true if non empty, false otherwise.
*
* @see #isEmpty(CharSequence)
* @see #isBlank(CharSequence)
* @see #isNonBlank(CharSequence)
*/
public static boolean isNonEmpty(CharSequence s) {
return !isEmpty(s);
}
/**
* The inverse of {@link #isBlank(CharSequence)}.
*
* @param s the String to check
* @return true if non blank, false otherwise.
*
* @see #isEmpty(CharSequence)
* @see #isNonEmpty(CharSequence)
* @see #isBlank(CharSequence)
*/
public static boolean isNonBlank(CharSequence s) {
return !isBlank(s);
}
/** throws IllegalArgument if string not empty; cf. guava Preconditions.checkXxxx */
public static void checkNonEmpty(CharSequence s) {
if (s==null) throw new IllegalArgumentException("String must not be null");
if (s.length()==0) throw new IllegalArgumentException("String must not be empty");
}
/** throws IllegalArgument if string not empty; cf. guava Preconditions.checkXxxx */
public static void checkNonEmpty(CharSequence s, String message) {
if (isEmpty(s)) throw new IllegalArgumentException(message);
}
/** removes the first suffix in the list which is present at the end of string
* and returns that string; ignores subsequent suffixes if a matching one is found;
* returns the original string if no suffixes are at the end
*/
public static String removeFromEnd(String string, String ...suffixes) {
if (isEmpty(string)) return string;
for (String suffix : suffixes)
if (string.endsWith(suffix)) return string.substring(0, string.length() - suffix.length());
return string;
}
/** as removeFromEnd, but repeats until all such suffixes are gone */
public static String removeAllFromEnd(String string, String ...suffixes) {
boolean anotherLoopNeeded = true;
while (anotherLoopNeeded) {
if (isEmpty(string)) return string;
anotherLoopNeeded = false;
for (String suffix : suffixes)
if (string.endsWith(suffix)) {
string = string.substring(0, string.length() - suffix.length());
anotherLoopNeeded = true;
break;
}
}
return string;
}
/** removes the first prefix in the list which is present at the start of string
* and returns that string; ignores subsequent prefixes if a matching one is found;
* returns the original string if no prefixes match
*/
public static String removeFromStart(String string, String ...prefixes) {
if (isEmpty(string)) return string;
for (String prefix : prefixes)
if (string.startsWith(prefix)) return string.substring(prefix.length());
return string;
}
/** as removeFromStart, but repeats until all such suffixes are gone */
public static String removeAllFromStart(String string, String ...prefixes) {
boolean anotherLoopNeeded = true;
while (anotherLoopNeeded) {
if (isEmpty(string)) return string;
anotherLoopNeeded = false;
for (String prefix : prefixes)
if (string.startsWith(prefix)) {
string = string.substring(prefix.length());
anotherLoopNeeded = true;
break;
}
}
return string;
}
/** convenience for {@link com.google.common.base.Joiner} */
public static String join(Iterable extends Object> list, String seperator) {
boolean app = false;
StringBuilder out = new StringBuilder();
for (Object s: list) {
if (app) out.append(seperator);
out.append(s);
app = true;
}
return out.toString();
}
/** convenience for {@link com.google.common.base.Joiner} */
public static String join(Object[] list, String seperator) {
boolean app = false;
StringBuilder out = new StringBuilder();
for (Object s: list) {
if (app) out.append(seperator);
out.append(s);
app = true;
}
return out.toString();
}
/** replaces all key->value entries from the replacement map in source (non-regex) */
@SuppressWarnings("rawtypes")
public static String replaceAll(String source, Map replacements) {
for (Object rr: replacements.entrySet()) {
Map.Entry r = (Map.Entry)rr;
source = replaceAllNonRegex(source, ""+r.getKey(), ""+r.getValue());
}
return source;
}
/** NON-REGEX replaceAll -
* replaces all instances in source, of the given pattern, with the given replacement
* (not interpreting any arguments as regular expressions)
*/
public static String replaceAll(String source, String pattern, String replacement) {
if (source==null) return source;
StringBuilder result = new StringBuilder(source.length());
for (int i=0; i() {
@Override
public boolean apply(@Nullable Character input) {
return input != null && Character.isJavaIdentifierPart(input);
}
});
/**
* Returns a valid Java identifier name based on the input.
*
* Removes certain characterss (like apostrophe), replaces one or more invalid
* characterss with {@literal _}, and prepends {@literal _} if the first character
* is only valid as an identifier part (not start).
*
* The result is usually unique to s, though this isn't guaranteed, for example if
* all characters are invalid. For a unique identifier use {@link #makeValidUniqueJavaName(String)}.
*
* @see #makeValidUniqueJavaName(String)
*/
public static String makeValidJavaName(String s) {
if (s==null) return "__null";
if (s.length()==0) return "__empty";
String name = IS_JAVA_IDENTIFIER_PART.negate().collapseFrom(CharMatcher.is('\'').removeFrom(s), '_');
if (!Character.isJavaIdentifierStart(s.charAt(0))) return "_" + name;
return name;
}
/**
* Returns a unique valid java identifier name based on the input.
*
* Translated as per {@link #makeValidJavaName(String)} but with {@link String#hashCode()}
* appended where necessary to guarantee uniqueness.
*
* @see #makeValidJavaName(String)
*/
public static String makeValidUniqueJavaName(String s) {
String name = makeValidJavaName(s);
if (isEmpty(s) || IS_JAVA_IDENTIFIER_PART.matchesAllOf(s) || CharMatcher.is('\'').matchesNoneOf(s)) {
return name;
} else {
return name + "_" + s.hashCode();
}
}
/** @see {@link Identifiers#makeRandomId(int)} */
public static String makeRandomId(int l) {
return Identifiers.makeRandomId(l);
}
/** pads the string with 0's at the left up to len; no padding if i longer than len */
public static String makeZeroPaddedString(int i, int len) {
return makePaddedString(""+i, len, "0", "");
}
/** pads the string with "pad" at the left up to len; no padding if base longer than len */
public static String makePaddedString(String base, int len, String left_pad, String right_pad) {
String s = ""+(base==null ? "" : base);
while (s.length()0 || prec<=1) s="0";
else {
s="0.0";
while (s.length()> 63) == 0) ? 1 : -1;
// int e = (int)((bits >> 52) & 0x7ffL);
// long m = (e == 0) ?
// (bits & 0xfffffffffffffL) << 1 :
// (bits & 0xfffffffffffffL) | 0x10000000000000L;
// //s*m*2^(e-1075);
int log = (int)Math.floor(Math.log10(x));
int numFractionDigits = (log>=prec ? 0 : prec-log-1);
if (numFractionDigits>0) { //need decimal digits
if (skipDecimalThreshhold>0) {
int checkFractionDigits = 0;
double multiplier = 1;
while (checkFractionDigits < numFractionDigits) {
if (Math.abs(x - Math.rint(x*multiplier)/multiplier)0 && s.length()>maxlen) {
//too long:
double signif = x/Math.pow(10,log);
if (s.indexOf('.')>=0) {
//have a decimal point; either we are very small 0.000001
//or prec is larger than maxlen
if (Math.abs(x)<1 && useEForSmallNumbers) {
//very small-- use alternate notation
s = makeRealString(signif, -1, prec, -1) + "E"+log;
} else {
//leave it alone, user error or E not wanted
}
} else {
//no decimal point, integer part is too large, use alt notation
s = makeRealString(signif, -1, prec, -1) + "E"+log;
}
}
}
if (leftPadLen>s.length())
return makePaddedString(s, leftPadLen, " ", "");
else
return s;
}
/** creates a string from a real number, with specified accuracy (more iff it comes for free, ie integer-part);
* switches to E notation if needed to fit within maxlen; can be padded left up too (not useful)
* @param x number to use
* @param maxlen maximum length for the numeric string, if possible (-1 to suppress)
* @param prec number of digits accuracy desired (more kept for integers)
* @param leftPadLen will add spaces at left if necessary to make string this long (-1 to suppress) [probably not usef]
* @return such a string
*/
public static String makeRealStringNearZero(double x, int maxlen, int prec, int leftPadLen) {
if (Math.abs(x)<0.0000000001) x=0;
NumberFormat df = DecimalFormat.getInstance();
//df.setMaximumFractionDigits(maxlen);
df.setMinimumFractionDigits(0);
//df.setMaximumIntegerDigits(prec);
df.setMinimumIntegerDigits(1);
df.setGroupingUsed(false);
String s;
if (x==0) {
if (prec<=1) s="0";
else {
s="0.0";
while (s.length()> 63) == 0) ? 1 : -1;
// int e = (int)((bits >> 52) & 0x7ffL);
// long m = (e == 0) ?
// (bits & 0xfffffffffffffL) << 1 :
// (bits & 0xfffffffffffffL) | 0x10000000000000L;
// //s*m*2^(e-1075);
int log = (int)Math.floor(Math.log10(x));
int scale = (log>=prec ? 0 : prec-log-1);
if (scale>0) { //need decimal digits
double scale10 = Math.pow(10, scale);
x = Math.rint(x*scale10)/scale10;
df.setMinimumFractionDigits(scale);
df.setMaximumFractionDigits(scale);
} else {
//x = Math.rint(x);
df.setMaximumFractionDigits(0);
}
s = df.format(x);
if (maxlen>0 && s.length()>maxlen) {
//too long:
double signif = x/Math.pow(10,log);
if (s.indexOf('.')>=0) {
//have a decimal point; either we are very small 0.000001
//or prec is larger than maxlen
if (Math.abs(x)<1) {
//very small-- use alternate notation
s = makeRealString(signif, -1, prec, -1) + "E"+log;
} else {
//leave it alone, user error
}
} else {
//no decimal point, integer part is too large, use alt notation
s = makeRealString(signif, -1, prec, -1) + "E"+log;
}
}
}
if (leftPadLen>s.length())
return makePaddedString(s, leftPadLen, " ", "");
else
return s;
}
/** returns the first word (whitespace delimited text), or null if there is none (input null or all whitespace) */
public static String getFirstWord(String s) {
if (s==null) return null;
int start = 0;
while (start= s.length())
return null;
while (end= 0) {
if (!Character.isWhitespace(s.charAt(end)))
break;
end--;
}
int start = end;
if (start < 0)
return null;
while (start >= 0) {
if (Character.isWhitespace(s.charAt(start)))
break;
start--;
}
return s.substring(start+1, end+1);
}
/** returns the first word after the given phrase, or null if no such phrase;
* if the character immediately after the phrase is not whitespace, the non-whitespace
* sequence starting with that character will be returned */
public static String getFirstWordAfter(String context, String phrase) {
if (context==null || phrase==null) return null;
int index = context.indexOf(phrase);
if (index<0) return null;
return getFirstWord(context.substring(index + phrase.length()));
}
/** @deprecated use {@link Time#makeTimeStringRounded(long)} */
@Deprecated
public static String makeTimeString(long utcMillis) {
return Time.makeTimeStringRounded(utcMillis);
}
/** returns e.g. { "prefix01", ..., "prefix96" };
* see more functional NumericRangeGlobExpander for "prefix{01-96}"
*/
public static String[] makeArray(String prefix, int count) {
String[] result = new String[count];
int len = (""+count).length();
for (int i=1; i<=count; i++)
result[i-1] = prefix + makePaddedString("", len, "0", ""+i);
return result;
}
public static String[] combineArrays(String[] ...arrays) {
int totalLen = 0;
for (String[] array : arrays) {
if (array!=null) totalLen += array.length;
}
String[] result = new String[totalLen];
int i=0;
for (String[] array : arrays) {
if (array!=null) for (String s : array) {
result[i++] = s;
}
}
return result;
}
public static String toInitialCapOnly(String value) {
if (value==null || value.length()==0) return value;
return value.substring(0, 1).toUpperCase() + value.substring(1).toLowerCase();
}
public static String reverse(String name) {
return new StringBuffer(name).reverse().toString();
}
public static boolean isLowerCase(String s) {
return s.toLowerCase().equals(s);
}
public static String makeRepeated(char c, int length) {
StringBuilder result = new StringBuilder(length);
for (int i = 0; i < length; i++) {
result.append(c);
}
return result.toString();
}
public static String trim(String s) {
if (s==null) return null;
return s.trim();
}
public static String trimEnd(String s) {
if (s==null) return null;
return ("a"+s).trim().substring(1);
}
/** returns up to maxlen characters from the start of s */
public static String maxlen(String s, int maxlen) {
if (s==null) return null;
return s.substring(0, Math.min(s.length(), maxlen));
}
/** returns toString of the object if it is not null, otherwise null */
public static String toString(Object o) {
if (o==null) return null;
return o.toString();
}
public static boolean containsLiteralIgnoreCase(CharSequence input, CharSequence fragment) {
if (input==null) return false;
if (isEmpty(fragment)) return true;
int lastValidStartPos = input.length()-fragment.length();
char f0u = Character.toUpperCase(fragment.charAt(0));
char f0l = Character.toLowerCase(fragment.charAt(0));
i: for (int i=0; i<=lastValidStartPos; i++) {
char ii = input.charAt(i);
if (ii==f0l || ii==f0u) {
for (int j=1; j toStringSupplier(Object src) {
return Suppliers.compose(Functions.toStringFunction(), Suppliers.ofInstance(src));
}
/** wraps a call to {@link String#format(String, Object...)} in a toString, i.e. using %s syntax,
* useful for places where we want deferred evaluation
* (e.g. as message to {@link Preconditions} to skip concatenation when not needed) */
public static FormattedString format(String pattern, Object... args) {
return new FormattedString(pattern, args);
}
/** returns "s" if the argument is not 1, empty string otherwise; useful when constructing plurals */
public static String s(int count) {
return count==1 ? "" : "s";
}
/** converts a map of any objects to a map of strings, preserving nulls and invoking toString where needed */
public static Map toStringMap(Map,?> map) {
Map result = MutableMap.of();
for (Map.Entry,?> e: map.entrySet()) {
result.put(String.valueOf(e.getKey()), String.valueOf(e.getValue()));
}
return result;
}
/** returns base repeated count times */
public static String repeat(String base, int count) {
if (base==null) return null;
StringBuilder result = new StringBuilder();
for (int i=0; i