/*
 *  Licensed to the Apache Software Foundation (ASF) under one or more
 *  contributor license agreements.  See the NOTICE file distributed with
 *  this work for additional information regarding copyright ownership.
 *  The ASF licenses this file to You 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.apache.http.conn.ssl;

import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import javax.security.auth.x500.X500Principal;

A distinguished name (DN) parser. This parser only supports extracting a string value from a DN. It doesn't support values in the hex-string style.
@hide
/** * A distinguished name (DN) parser. This parser only supports extracting a * string value from a DN. It doesn't support values in the hex-string style. * * @hide */
@Deprecated final class AndroidDistinguishedNameParser { private final String dn; private final int length; private int pos; private int beg; private int end;
tmp vars to store positions of the currently parsed item
/** tmp vars to store positions of the currently parsed item */
private int cur;
distinguished name chars
/** distinguished name chars */
private char[] chars; public AndroidDistinguishedNameParser(X500Principal principal) { // RFC2253 is used to ensure we get attributes in the reverse // order of the underlying ASN.1 encoding, so that the most // significant values of repeated attributes occur first. this.dn = principal.getName(X500Principal.RFC2253); this.length = this.dn.length(); } // gets next attribute type: (ALPHA 1*keychar) / oid private String nextAT() { // skip preceding space chars, they can present after // comma or semicolon (compatibility with RFC 1779) for (; pos < length && chars[pos] == ' '; pos++) { } if (pos == length) { return null; // reached the end of DN } // mark the beginning of attribute type beg = pos; // attribute type chars pos++; for (; pos < length && chars[pos] != '=' && chars[pos] != ' '; pos++) { // we don't follow exact BNF syntax here: // accept any char except space and '=' } if (pos >= length) { throw new IllegalStateException("Unexpected end of DN: " + dn); } // mark the end of attribute type end = pos; // skip trailing space chars between attribute type and '=' // (compatibility with RFC 1779) if (chars[pos] == ' ') { for (; pos < length && chars[pos] != '=' && chars[pos] == ' '; pos++) { } if (chars[pos] != '=' || pos == length) { throw new IllegalStateException("Unexpected end of DN: " + dn); } } pos++; //skip '=' char // skip space chars between '=' and attribute value // (compatibility with RFC 1779) for (; pos < length && chars[pos] == ' '; pos++) { } // in case of oid attribute type skip its prefix: "oid." or "OID." // (compatibility with RFC 1779) if ((end - beg > 4) && (chars[beg + 3] == '.') && (chars[beg] == 'O' || chars[beg] == 'o') && (chars[beg + 1] == 'I' || chars[beg + 1] == 'i') && (chars[beg + 2] == 'D' || chars[beg + 2] == 'd')) { beg += 4; } return new String(chars, beg, end - beg); } // gets quoted attribute value: QUOTATION *( quotechar / pair ) QUOTATION private String quotedAV() { pos++; beg = pos; end = beg; while (true) { if (pos == length) { throw new IllegalStateException("Unexpected end of DN: " + dn); } if (chars[pos] == '"') { // enclosing quotation was found pos++; break; } else if (chars[pos] == '\\') { chars[end] = getEscaped(); } else { // shift char: required for string with escaped chars chars[end] = chars[pos]; } pos++; end++; } // skip trailing space chars before comma or semicolon. // (compatibility with RFC 1779) for (; pos < length && chars[pos] == ' '; pos++) { } return new String(chars, beg, end - beg); } // gets hex string attribute value: "#" hexstring private String hexAV() { if (pos + 4 >= length) { // encoded byte array must be not less then 4 c throw new IllegalStateException("Unexpected end of DN: " + dn); } beg = pos; // store '#' position pos++; while (true) { // check for end of attribute value // looks for space and component separators if (pos == length || chars[pos] == '+' || chars[pos] == ',' || chars[pos] == ';') { end = pos; break; } if (chars[pos] == ' ') { end = pos; pos++; // skip trailing space chars before comma or semicolon. // (compatibility with RFC 1779) for (; pos < length && chars[pos] == ' '; pos++) { } break; } else if (chars[pos] >= 'A' && chars[pos] <= 'F') { chars[pos] += 32; //to low case } pos++; } // verify length of hex string // encoded byte array must be not less then 4 and must be even number int hexLen = end - beg; // skip first '#' char if (hexLen < 5 || (hexLen & 1) == 0) { throw new IllegalStateException("Unexpected end of DN: " + dn); } // get byte encoding from string representation byte[] encoded = new byte[hexLen / 2]; for (int i = 0, p = beg + 1; i < encoded.length; p += 2, i++) { encoded[i] = (byte) getByte(p); } return new String(chars, beg, hexLen); } // gets string attribute value: *( stringchar / pair ) private String escapedAV() { beg = pos; end = pos; while (true) { if (pos >= length) { // the end of DN has been found return new String(chars, beg, end - beg); } switch (chars[pos]) { case '+': case ',': case ';': // separator char has been found return new String(chars, beg, end - beg); case '\\': // escaped char chars[end++] = getEscaped(); pos++; break; case ' ': // need to figure out whether space defines // the end of attribute value or not cur = end; pos++; chars[end++] = ' '; for (; pos < length && chars[pos] == ' '; pos++) { chars[end++] = ' '; } if (pos == length || chars[pos] == ',' || chars[pos] == '+' || chars[pos] == ';') { // separator char or the end of DN has been found return new String(chars, beg, cur - beg); } break; default: chars[end++] = chars[pos]; pos++; } } } // returns escaped char private char getEscaped() { pos++; if (pos == length) { throw new IllegalStateException("Unexpected end of DN: " + dn); } switch (chars[pos]) { case '"': case '\\': case ',': case '=': case '+': case '<': case '>': case '#': case ';': case ' ': case '*': case '%': case '_': //FIXME: escaping is allowed only for leading or trailing space char return chars[pos]; default: // RFC doesn't explicitly say that escaped hex pair is // interpreted as UTF-8 char. It only contains an example of such DN. return getUTF8(); } } // decodes UTF-8 char // see http://www.unicode.org for UTF-8 bit distribution table private char getUTF8() { int res = getByte(pos); pos++; //FIXME tmp if (res < 128) { // one byte: 0-7F return (char) res; } else if (res >= 192 && res <= 247) { int count; if (res <= 223) { // two bytes: C0-DF count = 1; res = res & 0x1F; } else if (res <= 239) { // three bytes: E0-EF count = 2; res = res & 0x0F; } else { // four bytes: F0-F7 count = 3; res = res & 0x07; } int b; for (int i = 0; i < count; i++) { pos++; if (pos == length || chars[pos] != '\\') { return 0x3F; //FIXME failed to decode UTF-8 char - return '?' } pos++; b = getByte(pos); pos++; //FIXME tmp if ((b & 0xC0) != 0x80) { return 0x3F; //FIXME failed to decode UTF-8 char - return '?' } res = (res << 6) + (b & 0x3F); } return (char) res; } else { return 0x3F; //FIXME failed to decode UTF-8 char - return '?' } } // Returns byte representation of a char pair // The char pair is composed of DN char in // specified 'position' and the next char // According to BNF syntax: // hexchar = DIGIT / "A" / "B" / "C" / "D" / "E" / "F" // / "a" / "b" / "c" / "d" / "e" / "f" private int getByte(int position) { if (position + 1 >= length) { throw new IllegalStateException("Malformed DN: " + dn); } int b1, b2; b1 = chars[position]; if (b1 >= '0' && b1 <= '9') { b1 = b1 - '0'; } else if (b1 >= 'a' && b1 <= 'f') { b1 = b1 - 87; // 87 = 'a' - 10 } else if (b1 >= 'A' && b1 <= 'F') { b1 = b1 - 55; // 55 = 'A' - 10 } else { throw new IllegalStateException("Malformed DN: " + dn); } b2 = chars[position + 1]; if (b2 >= '0' && b2 <= '9') { b2 = b2 - '0'; } else if (b2 >= 'a' && b2 <= 'f') { b2 = b2 - 87; // 87 = 'a' - 10 } else if (b2 >= 'A' && b2 <= 'F') { b2 = b2 - 55; // 55 = 'A' - 10 } else { throw new IllegalStateException("Malformed DN: " + dn); } return (b1 << 4) + b2; }
Parses the DN and returns the most significant attribute value for an attribute type, or null if none found.
Params:
  • attributeType – attribute type to look for (e.g. "ca")
/** * Parses the DN and returns the most significant attribute value * for an attribute type, or null if none found. * * @param attributeType attribute type to look for (e.g. "ca") */
public String findMostSpecific(String attributeType) { // Initialize internal state. pos = 0; beg = 0; end = 0; cur = 0; chars = dn.toCharArray(); String attType = nextAT(); if (attType == null) { return null; } while (true) { String attValue = ""; if (pos == length) { return null; } switch (chars[pos]) { case '"': attValue = quotedAV(); break; case '#': attValue = hexAV(); break; case '+': case ',': case ';': // compatibility with RFC 1779: semicolon can separate RDNs //empty attribute value break; default: attValue = escapedAV(); } // Values are ordered from most specific to least specific // due to the RFC2253 formatting. So take the first match // we see. if (attributeType.equalsIgnoreCase(attType)) { return attValue; } if (pos >= length) { return null; } if (chars[pos] == ',' || chars[pos] == ';') { } else if (chars[pos] != '+') { throw new IllegalStateException("Malformed DN: " + dn); } pos++; attType = nextAT(); if (attType == null) { throw new IllegalStateException("Malformed DN: " + dn); } } }
Parses the DN and returns all values for an attribute type, in the order of decreasing significance (most significant first).
Params:
  • attributeType – attribute type to look for (e.g. "ca")
/** * Parses the DN and returns all values for an attribute type, in * the order of decreasing significance (most significant first). * * @param attributeType attribute type to look for (e.g. "ca") */
public List<String> getAllMostSpecificFirst(String attributeType) { // Initialize internal state. pos = 0; beg = 0; end = 0; cur = 0; chars = dn.toCharArray(); List<String> result = Collections.emptyList(); String attType = nextAT(); if (attType == null) { return result; } while (pos < length) { String attValue = ""; switch (chars[pos]) { case '"': attValue = quotedAV(); break; case '#': attValue = hexAV(); break; case '+': case ',': case ';': // compatibility with RFC 1779: semicolon can separate RDNs //empty attribute value break; default: attValue = escapedAV(); } // Values are ordered from most specific to least specific // due to the RFC2253 formatting. So take the first match // we see. if (attributeType.equalsIgnoreCase(attType)) { if (result.isEmpty()) { result = new ArrayList<String>(); } result.add(attValue); } if (pos >= length) { break; } if (chars[pos] == ',' || chars[pos] == ';') { } else if (chars[pos] != '+') { throw new IllegalStateException("Malformed DN: " + dn); } pos++; attType = nextAT(); if (attType == null) { throw new IllegalStateException("Malformed DN: " + dn); } } return result; } }