Initial revision
From-SVN: r102074
This commit is contained in:
parent
6f4434b39b
commit
f911ba985a
4557 changed files with 1000262 additions and 0 deletions
268
libjava/classpath/java/util/StringTokenizer.java
Normal file
268
libjava/classpath/java/util/StringTokenizer.java
Normal file
|
@ -0,0 +1,268 @@
|
|||
/* StringTokenizer -- breaks a String into tokens
|
||||
Copyright (C) 1998, 1999, 2001, 2002, 2005 Free Software Foundation, Inc.
|
||||
|
||||
This file is part of GNU Classpath.
|
||||
|
||||
GNU Classpath is free software; you can redistribute it and/or modify
|
||||
it under the terms of the GNU General Public License as published by
|
||||
the Free Software Foundation; either version 2, or (at your option)
|
||||
any later version.
|
||||
|
||||
GNU Classpath 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 for more details.
|
||||
|
||||
You should have received a copy of the GNU General Public License
|
||||
along with GNU Classpath; see the file COPYING. If not, write to the
|
||||
Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
||||
02110-1301 USA.
|
||||
|
||||
Linking this library statically or dynamically with other modules is
|
||||
making a combined work based on this library. Thus, the terms and
|
||||
conditions of the GNU General Public License cover the whole
|
||||
combination.
|
||||
|
||||
As a special exception, the copyright holders of this library give you
|
||||
permission to link this library with independent modules to produce an
|
||||
executable, regardless of the license terms of these independent
|
||||
modules, and to copy and distribute the resulting executable under
|
||||
terms of your choice, provided that you also meet, for each linked
|
||||
independent module, the terms and conditions of the license of that
|
||||
module. An independent module is a module which is not derived from
|
||||
or based on this library. If you modify this library, you may extend
|
||||
this exception to your version of the library, but you are not
|
||||
obligated to do so. If you do not wish to do so, delete this
|
||||
exception statement from your version. */
|
||||
|
||||
|
||||
package java.util;
|
||||
|
||||
/**
|
||||
* This class splits a string into tokens. The caller can set on which
|
||||
* delimiters the string should be split and if the delimiters should be
|
||||
* returned. This is much simpler than {@link java.io.StreamTokenizer}.
|
||||
*
|
||||
* <p>You may change the delimiter set on the fly by calling
|
||||
* nextToken(String). But the semantic is quite difficult; it even
|
||||
* depends on calling <code>hasMoreTokens()</code>. You should call
|
||||
* <code>hasMoreTokens()</code> before, otherwise the old delimiters
|
||||
* after the last token are candidates for being returned.
|
||||
*
|
||||
* <p>If you want to get the delimiters, you have to use the three argument
|
||||
* constructor. The delimiters are returned as token consisting of a
|
||||
* single character.
|
||||
*
|
||||
* @author Jochen Hoenicke
|
||||
* @author Warren Levy (warrenl@cygnus.com)
|
||||
* @see java.io.StreamTokenizer
|
||||
* @status updated to 1.4
|
||||
*/
|
||||
public class StringTokenizer implements Enumeration
|
||||
{
|
||||
// WARNING: StringTokenizer is a CORE class in the bootstrap cycle. See the
|
||||
// comments in vm/reference/java/lang/Runtime for implications of this fact.
|
||||
|
||||
/**
|
||||
* The position in the str, where we currently are.
|
||||
*/
|
||||
private int pos;
|
||||
|
||||
/**
|
||||
* The string that should be split into tokens.
|
||||
*/
|
||||
private final String str;
|
||||
|
||||
/**
|
||||
* The length of the string.
|
||||
*/
|
||||
private final int len;
|
||||
|
||||
/**
|
||||
* The string containing the delimiter characters.
|
||||
*/
|
||||
private String delim;
|
||||
|
||||
/**
|
||||
* Tells, if we should return the delimiters.
|
||||
*/
|
||||
private final boolean retDelims;
|
||||
|
||||
/**
|
||||
* Creates a new StringTokenizer for the string <code>str</code>,
|
||||
* that should split on the default delimiter set (space, tab,
|
||||
* newline, return and formfeed), and which doesn't return the
|
||||
* delimiters.
|
||||
*
|
||||
* @param str The string to split
|
||||
* @throws NullPointerException if str is null
|
||||
*/
|
||||
public StringTokenizer(String str)
|
||||
{
|
||||
this(str, " \t\n\r\f", false);
|
||||
}
|
||||
|
||||
/**
|
||||
* Create a new StringTokenizer, that splits the given string on
|
||||
* the given delimiter characters. It doesn't return the delimiter
|
||||
* characters.
|
||||
*
|
||||
* @param str the string to split
|
||||
* @param delim a string containing all delimiter characters
|
||||
* @throws NullPointerException if either argument is null
|
||||
*/
|
||||
public StringTokenizer(String str, String delim)
|
||||
{
|
||||
this(str, delim, false);
|
||||
}
|
||||
|
||||
/**
|
||||
* Create a new StringTokenizer, that splits the given string on
|
||||
* the given delimiter characters. If you set
|
||||
* <code>returnDelims</code> to <code>true</code>, the delimiter
|
||||
* characters are returned as tokens of their own. The delimiter
|
||||
* tokens always consist of a single character.
|
||||
*
|
||||
* @param str the string to split
|
||||
* @param delim a string containing all delimiter characters
|
||||
* @param returnDelims tells, if you want to get the delimiters
|
||||
* @throws NullPointerException if str or delim is null
|
||||
*/
|
||||
public StringTokenizer(String str, String delim, boolean returnDelims)
|
||||
{
|
||||
len = str.length();
|
||||
this.str = str;
|
||||
// The toString() hack causes the NullPointerException.
|
||||
this.delim = delim.toString();
|
||||
this.retDelims = returnDelims;
|
||||
this.pos = 0;
|
||||
}
|
||||
|
||||
/**
|
||||
* Tells if there are more tokens.
|
||||
*
|
||||
* @return true if the next call of nextToken() will succeed
|
||||
*/
|
||||
public boolean hasMoreTokens()
|
||||
{
|
||||
if (! retDelims)
|
||||
{
|
||||
while (pos < len && delim.indexOf(str.charAt(pos)) >= 0)
|
||||
pos++;
|
||||
}
|
||||
return pos < len;
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the nextToken, changing the delimiter set to the given
|
||||
* <code>delim</code>. The change of the delimiter set is
|
||||
* permanent, ie. the next call of nextToken(), uses the same
|
||||
* delimiter set.
|
||||
*
|
||||
* @param delim a string containing the new delimiter characters
|
||||
* @return the next token with respect to the new delimiter characters
|
||||
* @throws NoSuchElementException if there are no more tokens
|
||||
* @throws NullPointerException if delim is null
|
||||
*/
|
||||
public String nextToken(String delim) throws NoSuchElementException
|
||||
{
|
||||
this.delim = delim;
|
||||
return nextToken();
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the nextToken of the string.
|
||||
*
|
||||
* @return the next token with respect to the current delimiter characters
|
||||
* @throws NoSuchElementException if there are no more tokens
|
||||
*/
|
||||
public String nextToken() throws NoSuchElementException
|
||||
{
|
||||
if (pos < len && delim.indexOf(str.charAt(pos)) >= 0)
|
||||
{
|
||||
if (retDelims)
|
||||
return str.substring(pos, ++pos);
|
||||
while (++pos < len && delim.indexOf(str.charAt(pos)) >= 0);
|
||||
}
|
||||
if (pos < len)
|
||||
{
|
||||
int start = pos;
|
||||
while (++pos < len && delim.indexOf(str.charAt(pos)) < 0);
|
||||
|
||||
return str.substring(start, pos);
|
||||
}
|
||||
throw new NoSuchElementException();
|
||||
}
|
||||
|
||||
/**
|
||||
* This does the same as hasMoreTokens. This is the
|
||||
* <code>Enumeration</code> interface method.
|
||||
*
|
||||
* @return true, if the next call of nextElement() will succeed
|
||||
* @see #hasMoreTokens()
|
||||
*/
|
||||
public boolean hasMoreElements()
|
||||
{
|
||||
return hasMoreTokens();
|
||||
}
|
||||
|
||||
/**
|
||||
* This does the same as nextTokens. This is the
|
||||
* <code>Enumeration</code> interface method.
|
||||
*
|
||||
* @return the next token with respect to the current delimiter characters
|
||||
* @throws NoSuchElementException if there are no more tokens
|
||||
* @see #nextToken()
|
||||
*/
|
||||
public Object nextElement() throws NoSuchElementException
|
||||
{
|
||||
return nextToken();
|
||||
}
|
||||
|
||||
/**
|
||||
* This counts the number of remaining tokens in the string, with
|
||||
* respect to the current delimiter set.
|
||||
*
|
||||
* @return the number of times <code>nextTokens()</code> will succeed
|
||||
* @see #nextToken()
|
||||
*/
|
||||
public int countTokens()
|
||||
{
|
||||
int count = 0;
|
||||
int delimiterCount = 0;
|
||||
boolean tokenFound = false; // Set when a non-delimiter is found
|
||||
int tmpPos = pos;
|
||||
|
||||
// Note for efficiency, we count up the delimiters rather than check
|
||||
// retDelims every time we encounter one. That way, we can
|
||||
// just do the conditional once at the end of the method
|
||||
while (tmpPos < len)
|
||||
{
|
||||
if (delim.indexOf(str.charAt(tmpPos++)) >= 0)
|
||||
{
|
||||
if (tokenFound)
|
||||
{
|
||||
// Got to the end of a token
|
||||
count++;
|
||||
tokenFound = false;
|
||||
}
|
||||
delimiterCount++; // Increment for this delimiter
|
||||
}
|
||||
else
|
||||
{
|
||||
tokenFound = true;
|
||||
// Get to the end of the token
|
||||
while (tmpPos < len
|
||||
&& delim.indexOf(str.charAt(tmpPos)) < 0)
|
||||
++tmpPos;
|
||||
}
|
||||
}
|
||||
|
||||
// Make sure to count the last token
|
||||
if (tokenFound)
|
||||
count++;
|
||||
|
||||
// if counting delmiters add them into the token count
|
||||
return retDelims ? count + delimiterCount : count;
|
||||
}
|
||||
} // class StringTokenizer
|
Loading…
Add table
Add a link
Reference in a new issue