/*
* Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
package com.sun.xml.internal.bind.api.impl;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashSet;
import java.util.List;
import java.util.Locale;
Methods that convert strings into various formats.
What JAX-RPC name binding tells us is that even such basic method
like "isLetter" can be different depending on the situation.
For this reason, a whole lot of methods are made non-static,
even though they look like they should be static.
/**
* Methods that convert strings into various formats.
*
* <p>
* What JAX-RPC name binding tells us is that even such basic method
* like "isLetter" can be different depending on the situation.
*
* For this reason, a whole lot of methods are made non-static,
* even though they look like they should be static.
*/
class NameUtil {
protected boolean isPunct(char c) {
return c == '-' || c == '.' || c == ':' || c == '_' || c == '\u00b7' || c == '\u0387' || c == '\u06dd' || c == '\u06de';
}
protected static boolean isDigit(char c) {
return c >= '0' && c <= '9' || Character.isDigit(c);
}
protected static boolean isUpper(char c) {
return c >= 'A' && c <= 'Z' || Character.isUpperCase(c);
}
protected static boolean isLower(char c) {
return c >= 'a' && c <= 'z' || Character.isLowerCase(c);
}
protected boolean isLetter(char c) {
return c >= 'A' && c <= 'Z' || c >= 'a' && c <= 'z' || Character.isLetter(c);
}
private String toLowerCase(String s)
{
return s.toLowerCase(Locale.ENGLISH);
}
private String toUpperCase(char c)
{
return String.valueOf(c).toUpperCase(Locale.ENGLISH);
}
private String toUpperCase(String s)
{
return s.toUpperCase(Locale.ENGLISH);
}
Capitalizes the first character of the specified string,
and de-capitalize the rest of characters.
/**
* Capitalizes the first character of the specified string,
* and de-capitalize the rest of characters.
*/
public String capitalize(String s) {
if (!isLower(s.charAt(0)))
return s;
StringBuilder sb = new StringBuilder(s.length());
sb.append(toUpperCase(s.charAt(0)));
sb.append(toLowerCase(s.substring(1)));
return sb.toString();
}
// Precondition: s[start] is not punctuation
private int nextBreak(String s, int start) {
int n = s.length();
char c1 = s.charAt(start);
int t1 = classify(c1);
for (int i=start+1; i<n; i++) {
// shift (c1,t1) into (c0,t0)
// char c0 = c1; --- conceptually, but c0 won't be used
int t0 = t1;
c1 = s.charAt(i);
t1 = classify(c1);
switch(actionTable[t0*5+t1]) {
case ACTION_CHECK_PUNCT:
if(isPunct(c1)) return i;
break;
case ACTION_CHECK_C2:
if (i < n-1) {
char c2 = s.charAt(i+1);
if (isLower(c2))
return i;
}
break;
case ACTION_BREAK:
return i;
}
}
return -1;
}
// the 5-category classification that we use in this code
// to find work breaks
static protected final int UPPER_LETTER = 0;
static protected final int LOWER_LETTER = 1;
static protected final int OTHER_LETTER = 2;
static protected final int DIGIT = 3;
static protected final int OTHER = 4;
Look up table for actions.
type0*5+type1 would yield the action to be taken.
/**
* Look up table for actions.
* type0*5+type1 would yield the action to be taken.
*/
private static final byte[] actionTable = new byte[5*5];
// action constants. see nextBreak for the meaning
static private final byte ACTION_CHECK_PUNCT = 0;
static private final byte ACTION_CHECK_C2 = 1;
static private final byte ACTION_BREAK = 2;
static private final byte ACTION_NOBREAK = 3;
Decide the action to be taken given
the classification of the preceding character 't0' and
the classification of the next character 't1'.
/**
* Decide the action to be taken given
* the classification of the preceding character 't0' and
* the classification of the next character 't1'.
*/
private static byte decideAction( int t0, int t1 ) {
if(t0==OTHER && t1==OTHER) return ACTION_CHECK_PUNCT;
if(!xor(t0==DIGIT,t1==DIGIT)) return ACTION_BREAK;
if(t0==LOWER_LETTER && t1!=LOWER_LETTER) return ACTION_BREAK;
if(!xor(t0<=OTHER_LETTER,t1<=OTHER_LETTER)) return ACTION_BREAK;
if(!xor(t0==OTHER_LETTER,t1==OTHER_LETTER)) return ACTION_BREAK;
if(t0==UPPER_LETTER && t1==UPPER_LETTER) return ACTION_CHECK_C2;
return ACTION_NOBREAK;
}
private static boolean xor(boolean x,boolean y) {
return (x&&y) || (!x&&!y);
}
static {
// initialize the action table
for( int t0=0; t0<5; t0++ )
for( int t1=0; t1<5; t1++ )
actionTable[t0*5+t1] = decideAction(t0,t1);
}
Classify a character into 5 categories that determine the word break.
/**
* Classify a character into 5 categories that determine the word break.
*/
protected int classify(char c0) {
switch(Character.getType(c0)) {
case Character.UPPERCASE_LETTER: return UPPER_LETTER;
case Character.LOWERCASE_LETTER: return LOWER_LETTER;
case Character.TITLECASE_LETTER:
case Character.MODIFIER_LETTER:
case Character.OTHER_LETTER: return OTHER_LETTER;
case Character.DECIMAL_DIGIT_NUMBER: return DIGIT;
default: return OTHER;
}
}
Tokenizes a string into words and capitalizes the first
character of each word.
This method uses a change in character type as a splitter
of two words. For example, "abc100ghi" will be splitted into
{"Abc", "100","Ghi"}.
/**
* Tokenizes a string into words and capitalizes the first
* character of each word.
*
* <p>
* This method uses a change in character type as a splitter
* of two words. For example, "abc100ghi" will be splitted into
* {"Abc", "100","Ghi"}.
*/
public List<String> toWordList(String s) {
ArrayList<String> ss = new ArrayList<String>();
int n = s.length();
for (int i = 0; i < n;) {
// Skip punctuation
while (i < n) {
if (!isPunct(s.charAt(i)))
break;
i++;
}
if (i >= n) break;
// Find next break and collect word
int b = nextBreak(s, i);
String w = (b == -1) ? s.substring(i) : s.substring(i, b);
ss.add(escape(capitalize(w)));
if (b == -1) break;
i = b;
}
// we can't guarantee a valid Java identifier anyway,
// so there's not much point in rejecting things in this way.
// if (ss.size() == 0)
// throw new IllegalArgumentException("Zero-length identifier");
return ss;
}
protected String toMixedCaseName(List<String> ss, boolean startUpper) {
StringBuilder sb = new StringBuilder();
if(!ss.isEmpty()) {
sb.append(startUpper ? ss.get(0) : toLowerCase(ss.get(0)));
for (int i = 1; i < ss.size(); i++)
sb.append(ss.get(i));
}
return sb.toString();
}
protected String toMixedCaseVariableName(String[] ss,
boolean startUpper,
boolean cdrUpper) {
if (cdrUpper)
for (int i = 1; i < ss.length; i++)
ss[i] = capitalize(ss[i]);
StringBuilder sb = new StringBuilder();
if( ss.length>0 ) {
sb.append(startUpper ? ss[0] : toLowerCase(ss[0]));
for (int i = 1; i < ss.length; i++)
sb.append(ss[i]);
}
return sb.toString();
}
Formats a string into "THIS_KIND_OF_FORMAT_ABC_DEF".
Returns:
Always return a string but there's no guarantee that
the generated code is a valid Java identifier.
/**
* Formats a string into "THIS_KIND_OF_FORMAT_ABC_DEF".
*
* @return
* Always return a string but there's no guarantee that
* the generated code is a valid Java identifier.
*/
public String toConstantName(String s) {
return toConstantName(toWordList(s));
}
Formats a string into "THIS_KIND_OF_FORMAT_ABC_DEF".
Returns:
Always return a string but there's no guarantee that
the generated code is a valid Java identifier.
/**
* Formats a string into "THIS_KIND_OF_FORMAT_ABC_DEF".
*
* @return
* Always return a string but there's no guarantee that
* the generated code is a valid Java identifier.
*/
public String toConstantName(List<String> ss) {
StringBuilder sb = new StringBuilder();
if( !ss.isEmpty() ) {
sb.append(toUpperCase(ss.get(0)));
for (int i = 1; i < ss.size(); i++) {
sb.append('_');
sb.append(toUpperCase(ss.get(i)));
}
}
return sb.toString();
}
Escapes characters is the given string so that they can be
printed by only using US-ASCII characters.
The escaped characters will be appended to the given
StringBuffer.
Params: - sb –
StringBuffer that receives escaped string.
- s –
String to be escaped.
s.substring(start)
will be escaped and copied to the string buffer.
/**
* Escapes characters is the given string so that they can be
* printed by only using US-ASCII characters.
*
* The escaped characters will be appended to the given
* StringBuffer.
*
* @param sb
* StringBuffer that receives escaped string.
* @param s
* String to be escaped. <code>s.substring(start)</code>
* will be escaped and copied to the string buffer.
*/
public static void escape(StringBuilder sb, String s, int start) {
int n = s.length();
for (int i = start; i < n; i++) {
char c = s.charAt(i);
if (Character.isJavaIdentifierPart(c))
sb.append(c);
else {
sb.append('_');
if (c <= '\u000f') sb.append("000");
else if (c <= '\u00ff') sb.append("00");
else if (c <= '\u0fff') sb.append('0');
sb.append(Integer.toString(c, 16));
}
}
}
Escapes characters that are unusable as Java identifiers
by replacing unsafe characters with safe characters.
/**
* Escapes characters that are unusable as Java identifiers
* by replacing unsafe characters with safe characters.
*/
private static String escape(String s) {
int n = s.length();
for (int i = 0; i < n; i++)
if (!Character.isJavaIdentifierPart(s.charAt(i))) {
StringBuilder sb = new StringBuilder(s.substring(0, i));
escape(sb, s, i);
return sb.toString();
}
return s;
}
}