/*
* Copyright (c) 1999, 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.jndi.toolkit.dir;
import javax.naming.*;
import javax.naming.directory.*;
import java.util.Enumeration;
import java.util.StringTokenizer;
import java.util.Vector;
import java.util.Locale;
/**
* A class for parsing LDAP search filters (defined in RFC 1960, 2254)
*
* @author Jon Ruiz
* @author Rosanna Lee
*/
public class SearchFilter implements AttrFilter {
interface StringFilter extends AttrFilter {
public void parse() throws InvalidSearchFilterException;
}
// %%% "filter" and "pos" are not declared "private" due to bug 4064984.
String filter;
int pos;
private StringFilter rootFilter;
protected static final boolean debug = false;
protected static final char BEGIN_FILTER_TOKEN = '(';
protected static final char END_FILTER_TOKEN = ')';
protected static final char AND_TOKEN = '&';
protected static final char OR_TOKEN = '|';
protected static final char NOT_TOKEN = '!';
protected static final char EQUAL_TOKEN = '=';
protected static final char APPROX_TOKEN = '~';
protected static final char LESS_TOKEN = '<';
protected static final char GREATER_TOKEN = '>';
protected static final char EXTEND_TOKEN = ':';
protected static final char WILDCARD_TOKEN = '*';
public SearchFilter(String filter) throws InvalidSearchFilterException {
this.filter = filter;
pos = 0;
normalizeFilter();
rootFilter = this.createNextFilter();
}
// Returns true if targetAttrs passes the filter
public boolean check(Attributes targetAttrs) throws NamingException {
if (targetAttrs == null)
return false;
return rootFilter.check(targetAttrs);
}
/*
* Utility routines used by member classes
*/
// does some pre-processing on the string to make it look exactly lik
// what the parser expects. This only needs to be called once.
protected void normalizeFilter() {
skipWhiteSpace(); // get rid of any leading whitespaces
// Sometimes, search filters don't have "(" and ")" - add them
if(getCurrentChar() != BEGIN_FILTER_TOKEN) {
filter = BEGIN_FILTER_TOKEN + filter + END_FILTER_TOKEN;
}
// this would be a good place to strip whitespace if desired
if(debug) {System.out.println("SearchFilter: normalized filter:" +
filter);}
}
private void skipWhiteSpace() {
while (Character.isWhitespace(getCurrentChar())) {
consumeChar();
}
}
protected StringFilter createNextFilter()
throws InvalidSearchFilterException {
StringFilter filter;
skipWhiteSpace();
try {
// make sure every filter starts with "("
if(getCurrentChar() != BEGIN_FILTER_TOKEN) {
throw new InvalidSearchFilterException("expected \"" +
BEGIN_FILTER_TOKEN +
"\" at position " +
pos);
}
// skip past the "("
this.consumeChar();
skipWhiteSpace();
// use the next character to determine the type of filter
switch(getCurrentChar()) {
case AND_TOKEN:
if (debug) {System.out.println("SearchFilter: creating AND");}
filter = new CompoundFilter(true);
filter.parse();
break;
case OR_TOKEN:
if (debug) {System.out.println("SearchFilter: creating OR");}
filter = new CompoundFilter(false);
filter.parse();
break;
case NOT_TOKEN:
if (debug) {System.out.println("SearchFilter: creating OR");}
filter = new NotFilter();
filter.parse();
break;
default:
if (debug) {System.out.println("SearchFilter: creating SIMPLE");}
filter = new AtomicFilter();
filter.parse();
break;
}
skipWhiteSpace();
// make sure every filter ends with ")"
if(getCurrentChar() != END_FILTER_TOKEN) {
throw new InvalidSearchFilterException("expected \"" +
END_FILTER_TOKEN +
"\" at position " +
pos);
}
// skip past the ")"
this.consumeChar();
} catch (InvalidSearchFilterException e) {
if (debug) {System.out.println("rethrowing e");}
throw e; // just rethrow these
// catch all - any uncaught exception while parsing will end up here
} catch (Exception e) {
if(debug) {System.out.println(e.getMessage());e.printStackTrace();}
throw new InvalidSearchFilterException("Unable to parse " +
"character " + pos + " in \""+
this.filter + "\"");
}
return filter;
}
protected char getCurrentChar() {
return filter.charAt(pos);
}
protected char relCharAt(int i) {
return filter.charAt(pos + i);
}
protected void consumeChar() {
pos++;
}
protected void consumeChars(int i) {
pos += i;
}
protected int relIndexOf(int ch) {
return filter.indexOf(ch, pos) - pos;
}
protected String relSubstring(int beginIndex, int endIndex){
if(debug){System.out.println("relSubString: " + beginIndex +
" " + endIndex);}
return filter.substring(beginIndex+pos, endIndex+pos);
}
/**
* A class for dealing with compound filters ("and" & "or" filters).
*/
final class CompoundFilter implements StringFilter {
private Vector<StringFilter> subFilters;
private boolean polarity;
CompoundFilter(boolean polarity) {
subFilters = new Vector<>();
this.polarity = polarity;
}
public void parse() throws InvalidSearchFilterException {
SearchFilter.this.consumeChar(); // consume the "&"
while(SearchFilter.this.getCurrentChar() != END_FILTER_TOKEN) {
if (debug) {System.out.println("CompoundFilter: adding");}
StringFilter filter = SearchFilter.this.createNextFilter();
subFilters.addElement(filter);
skipWhiteSpace();
}
}
public boolean check(Attributes targetAttrs) throws NamingException {
for(int i = 0; i<subFilters.size(); i++) {
StringFilter filter = subFilters.elementAt(i);
if(filter.check(targetAttrs) != this.polarity) {
return !polarity;
}
}
return polarity;
}
} /* CompoundFilter */
/**
* A class for dealing with NOT filters
*/
final class NotFilter implements StringFilter {
private StringFilter filter;
public void parse() throws InvalidSearchFilterException {
SearchFilter.this.consumeChar(); // consume the "!"
filter = SearchFilter.this.createNextFilter();
}
public boolean check(Attributes targetAttrs) throws NamingException {
return !filter.check(targetAttrs);
}
} /* notFilter */
// note: declared here since member classes can't have static variables
static final int EQUAL_MATCH = 1;
static final int APPROX_MATCH = 2;
static final int GREATER_MATCH = 3;
static final int LESS_MATCH = 4;
/**
* A class for dealing wtih atomic filters
*/
final class AtomicFilter implements StringFilter {
private String attrID;
private String value;
private int matchType;
public void parse() throws InvalidSearchFilterException {
skipWhiteSpace();
try {
// find the end
int endPos = SearchFilter.this.relIndexOf(END_FILTER_TOKEN);
//determine the match type
int i = SearchFilter.this.relIndexOf(EQUAL_TOKEN);
if(debug) {System.out.println("AtomicFilter: = at " + i);}
int qualifier = SearchFilter.this.relCharAt(i-1);
switch(qualifier) {
case APPROX_TOKEN:
if (debug) {System.out.println("Atomic: APPROX found");}
matchType = APPROX_MATCH;
attrID = SearchFilter.this.relSubstring(0, i-1);
value = SearchFilter.this.relSubstring(i+1, endPos);
break;
case GREATER_TOKEN:
if (debug) {System.out.println("Atomic: GREATER found");}
matchType = GREATER_MATCH;
attrID = SearchFilter.this.relSubstring(0, i-1);
value = SearchFilter.this.relSubstring(i+1, endPos);
break;
case LESS_TOKEN:
if (debug) {System.out.println("Atomic: LESS found");}
matchType = LESS_MATCH;
attrID = SearchFilter.this.relSubstring(0, i-1);
value = SearchFilter.this.relSubstring(i+1, endPos);
break;
case EXTEND_TOKEN:
if(debug) {System.out.println("Atomic: EXTEND found");}
throw new OperationNotSupportedException("Extensible match not supported");
default:
if (debug) {System.out.println("Atomic: EQUAL found");}
matchType = EQUAL_MATCH;
attrID = SearchFilter.this.relSubstring(0,i);
value = SearchFilter.this.relSubstring(i+1, endPos);
break;
}
attrID = attrID.trim();
value = value.trim();
//update our position
SearchFilter.this.consumeChars(endPos);
} catch (Exception e) {
if (debug) {System.out.println(e.getMessage());
e.printStackTrace();}
InvalidSearchFilterException sfe =
new InvalidSearchFilterException("Unable to parse " +
"character " + SearchFilter.this.pos + " in \""+
SearchFilter.this.filter + "\"");
sfe.setRootCause(e);
throw(sfe);
}
if(debug) {System.out.println("AtomicFilter: " + attrID + "=" +
value);}
}
public boolean check(Attributes targetAttrs) {
Enumeration<?> candidates;
try {
Attribute attr = targetAttrs.get(attrID);
if(attr == null) {
return false;
}
candidates = attr.getAll();
} catch (NamingException ne) {
if (debug) {System.out.println("AtomicFilter: should never " +
"here");}
return false;
}
while(candidates.hasMoreElements()) {
String val = candidates.nextElement().toString();
if (debug) {System.out.println("Atomic: comparing: " + val);}
switch(matchType) {
case APPROX_MATCH:
case EQUAL_MATCH:
if(substringMatch(this.value, val)) {
if (debug) {System.out.println("Atomic: EQUAL match");}
return true;
}
break;
case GREATER_MATCH:
if (debug) {System.out.println("Atomic: GREATER match");}
if(val.compareTo(this.value) >= 0) {
return true;
}
break;
case LESS_MATCH:
if (debug) {System.out.println("Atomic: LESS match");}
if(val.compareTo(this.value) <= 0) {
return true;
}
break;
default:
if (debug) {System.out.println("AtomicFilter: unkown " +
"matchType");}
}
}
return false;
}
// used for substring comparisons (where proto has "*" wildcards
private boolean substringMatch(String proto, String value) {
// simple case 1: "*" means attribute presence is being tested
if(proto.equals(new Character(WILDCARD_TOKEN).toString())) {
if(debug) {System.out.println("simple presence assertion");}
return true;
}
// simple case 2: if there are no wildcards, call String.equals()
if(proto.indexOf(WILDCARD_TOKEN) == -1) {
return proto.equalsIgnoreCase(value);
}
if(debug) {System.out.println("doing substring comparison");}
// do the work: make sure all the substrings are present
int currentPos = 0;
StringTokenizer subStrs = new StringTokenizer(proto, "*", false);
// do we need to begin with the first token?
if(proto.charAt(0) != WILDCARD_TOKEN &&
!value.toLowerCase(Locale.ENGLISH).startsWith(
subStrs.nextToken().toLowerCase(Locale.ENGLISH))) {
if(debug) {
System.out.println("faild initial test");
}
return false;
}
while(subStrs.hasMoreTokens()) {
String currentStr = subStrs.nextToken();
if (debug) {System.out.println("looking for \"" +
currentStr +"\"");}
currentPos = value.toLowerCase(Locale.ENGLISH).indexOf(
currentStr.toLowerCase(Locale.ENGLISH), currentPos);
if(currentPos == -1) {
return false;
}
currentPos += currentStr.length();
}
// do we need to end with the last token?
if(proto.charAt(proto.length() - 1) != WILDCARD_TOKEN &&
currentPos != value.length() ) {
if(debug) {System.out.println("faild final test");}
return false;
}
return true;
}
} /* AtomicFilter */
// ----- static methods for producing string filters given attribute set
// ----- or object array
/**
* Creates an LDAP filter as a conjuction of the attributes supplied.
*/
public static String format(Attributes attrs) throws NamingException {
if (attrs == null || attrs.size() == 0) {
return "objectClass=*";
}
String answer;
answer = "(& ";
Attribute attr;
for (NamingEnumeration<? extends Attribute> e = attrs.getAll();
e.hasMore(); ) {
attr = e.next();
if (attr.size() == 0 || (attr.size() == 1 && attr.get() == null)) {
// only checking presence of attribute
answer += "(" + attr.getID() + "=" + "*)";
} else {
for (NamingEnumeration<?> ve = attr.getAll();
ve.hasMore(); ) {
String val = getEncodedStringRep(ve.next());
if (val != null) {
answer += "(" + attr.getID() + "=" + val + ")";
}
}
}
}
answer += ")";
//System.out.println("filter: " + answer);
return answer;
}
// Writes the hex representation of a byte to a StringBuffer.
private static void hexDigit(StringBuffer buf, byte x) {
char c;
c = (char) ((x >> 4) & 0xf);
if (c > 9)
c = (char) ((c-10) + 'A');
else
c = (char)(c + '0');
buf.append(c);
c = (char) (x & 0xf);
if (c > 9)
c = (char)((c-10) + 'A');
else
c = (char)(c + '0');
buf.append(c);
}
/**
* Returns the string representation of an object (such as an attr value).
* If obj is a byte array, encode each item as \xx, where xx is hex encoding
* of the byte value.
* Else, if obj is not a String, use its string representation (toString()).
* Special characters in obj (or its string representation) are then
* encoded appropriately according to RFC 2254.
* * \2a
* ( \28
* ) \29
* \ \5c
* NUL \00
*/
private static String getEncodedStringRep(Object obj) throws NamingException {
String str;
if (obj == null)
return null;
if (obj instanceof byte[]) {
// binary data must be encoded as \hh where hh is a hex char
byte[] bytes = (byte[])obj;
StringBuffer b1 = new StringBuffer(bytes.length*3);
for (int i = 0; i < bytes.length; i++) {
b1.append('\\');
hexDigit(b1, bytes[i]);
}
return b1.toString();
}
if (!(obj instanceof String)) {
str = obj.toString();
} else {
str = (String)obj;
}
int len = str.length();
StringBuffer buf = new StringBuffer(len);
char ch;
for (int i = 0; i < len; i++) {
switch (ch=str.charAt(i)) {
case '*':
buf.append("\\2a");
break;
case '(':
buf.append("\\28");
break;
case ')':
buf.append("\\29");
break;
case '\\':
buf.append("\\5c");
break;
case 0:
buf.append("\\00");
break;
default:
buf.append(ch);
}
}
return buf.toString();
}
/**
* Finds the first occurrence of <tt>ch</tt> in <tt>val</tt> starting
* from position <tt>start</tt>. It doesn't count if <tt>ch</tt>
* has been escaped by a backslash (\)
*/
public static int findUnescaped(char ch, String val, int start) {
int len = val.length();
while (start < len) {
int where = val.indexOf(ch, start);
// if at start of string, or not there at all, or if not escaped
if (where == start || where == -1 || val.charAt(where-1) != '\\')
return where;
// start search after escaped star
start = where + 1;
}
return -1;
}
/**
* Formats the expression <tt>expr</tt> using arguments from the array
* <tt>args</tt>.
*
* <code>{i}</code> specifies the <code>i</code>'th element from
* the array <code>args</code> is to be substituted for the
* string "<code>{i}</code>".
*
* To escape '{' or '}' (or any other character), use '\'.
*
* Uses getEncodedStringRep() to do encoding.
*/
public static String format(String expr, Object[] args)
throws NamingException {
int param;
int where = 0, start = 0;
StringBuffer answer = new StringBuffer(expr.length());
while ((where = findUnescaped('{', expr, start)) >= 0) {
int pstart = where + 1; // skip '{'
int pend = expr.indexOf('}', pstart);
if (pend < 0) {
throw new InvalidSearchFilterException("unbalanced {: " + expr);
/**代码未完, 请加载全部代码(NowJava.com).**/