javax.mail.internet.InternetAddress Maven / Gradle / Ivy
/*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS HEADER.
*
* Copyright (c) 1997-2016 Oracle and/or its affiliates. All rights reserved.
*
* The contents of this file are subject to the terms of either the GNU
* General Public License Version 2 only ("GPL") or the Common Development
* and Distribution License("CDDL") (collectively, the "License"). You
* may not use this file except in compliance with the License. You can
* obtain a copy of the License at
* https://glassfish.dev.java.net/public/CDDL+GPL_1_1.html
* or packager/legal/LICENSE.txt. See the License for the specific
* language governing permissions and limitations under the License.
*
* When distributing the software, include this License Header Notice in each
* file and include the License file at packager/legal/LICENSE.txt.
*
* GPL Classpath Exception:
* Oracle designates this particular file as subject to the "Classpath"
* exception as provided by Oracle in the GPL Version 2 section of the License
* file that accompanied this code.
*
* Modifications:
* If applicable, add the following below the License Header, with the fields
* enclosed by brackets [] replaced by your own identifying information:
* "Portions Copyright [year] [name of copyright owner]"
*
* Contributor(s):
* If you wish your version of this file to be governed by only the CDDL or
* only the GPL Version 2, indicate your decision by adding "[Contributor]
* elects to include this software in this distribution under the [CDDL or GPL
* Version 2] license." If you don't indicate a single choice of license, a
* recipient has the option to distribute your version of this file under
* either the CDDL, the GPL Version 2 or to extend the choice of license to
* its licensees as provided above. However, if you add GPL Version 2 code
* and therefore, elected the GPL Version 2 license, then the option applies
* only if the new code is made subject to such option by the copyright
* holder.
*/
package javax.mail.internet;
import java.io.UnsupportedEncodingException;
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.List;
import java.util.ArrayList;
import java.util.StringTokenizer;
import java.util.Locale;
import java.nio.charset.StandardCharsets;
import javax.mail.*;
import com.sun.mail.util.PropUtil;
/**
* This class represents an Internet email address using the syntax
* of RFC822.
* Typical address syntax is of the form "[email protected]" or
* "Personal Name <[email protected]>".
*
* @author Bill Shannon
* @author John Mani
*/
public class InternetAddress extends Address implements Cloneable {
protected String address; // email address
/**
* The personal name.
*/
protected String personal;
/**
* The RFC 2047 encoded version of the personal name.
*
* This field and the personal
field track each
* other, so if a subclass sets one of these fields directly, it
* should set the other to null
, so that it is
* suitably recomputed.
*/
protected String encodedPersonal;
private static final long serialVersionUID = -7507595530758302903L;
private static final boolean ignoreBogusGroupName =
PropUtil.getBooleanSystemProperty(
"mail.mime.address.ignorebogusgroupname", true);
private static final boolean useCanonicalHostName =
PropUtil.getBooleanSystemProperty(
"mail.mime.address.usecanonicalhostname", true);
private static final boolean allowUtf8 =
PropUtil.getBooleanSystemProperty("mail.mime.allowutf8", false);
/**
* Default constructor.
*/
public InternetAddress() { }
/**
* Constructor.
*
* Parse the given string and create an InternetAddress.
* See the parse
method for details of the parsing.
* The address is parsed using "strict" parsing.
* This constructor does not perform the additional
* syntax checks that the
* InternetAddress(String address, boolean strict)
* constructor does when strict
is true
.
* This constructor is equivalent to
* InternetAddress(address, false)
.
*
* @param address the address in RFC822 format
* @exception AddressException if the parse failed
*/
public InternetAddress(String address) throws AddressException {
// use our address parsing utility routine to parse the string
InternetAddress a[] = parse(address, true);
// if we got back anything other than a single address, it's an error
if (a.length != 1)
throw new AddressException("Illegal address", address);
/*
* Now copy the contents of the single address we parsed
* into the current object, which will be returned from the
* constructor.
* XXX - this sure is a round-about way of getting this done.
*/
this.address = a[0].address;
this.personal = a[0].personal;
this.encodedPersonal = a[0].encodedPersonal;
}
/**
* Parse the given string and create an InternetAddress.
* If strict
is false, the detailed syntax of the
* address isn't checked.
*
* @param address the address in RFC822 format
* @param strict enforce RFC822 syntax
* @exception AddressException if the parse failed
* @since JavaMail 1.3
*/
public InternetAddress(String address, boolean strict)
throws AddressException {
this(address);
if (strict) {
if (isGroup())
getGroup(true); // throw away the result
else
checkAddress(this.address, true, true);
}
}
/**
* Construct an InternetAddress given the address and personal name.
* The address is assumed to be a syntactically valid RFC822 address.
*
* @param address the address in RFC822 format
* @param personal the personal name
* @exception UnsupportedEncodingException if the personal name
* can't be encoded in the given charset
*/
public InternetAddress(String address, String personal)
throws UnsupportedEncodingException {
this(address, personal, null);
}
/**
* Construct an InternetAddress given the address and personal name.
* The address is assumed to be a syntactically valid RFC822 address.
*
* @param address the address in RFC822 format
* @param personal the personal name
* @param charset the MIME charset for the name
* @exception UnsupportedEncodingException if the personal name
* can't be encoded in the given charset
*/
public InternetAddress(String address, String personal, String charset)
throws UnsupportedEncodingException {
this.address = address;
setPersonal(personal, charset);
}
/**
* Return a copy of this InternetAddress object.
* @since JavaMail 1.2
*/
@Override
public Object clone() {
InternetAddress a = null;
try {
a = (InternetAddress)super.clone();
} catch (CloneNotSupportedException e) {} // Won't happen
return a;
}
/**
* Return the type of this address. The type of an InternetAddress
* is "rfc822".
*/
@Override
public String getType() {
return "rfc822";
}
/**
* Set the email address.
*
* @param address email address
*/
public void setAddress(String address) {
this.address = address;
}
/**
* Set the personal name. If the name contains non US-ASCII
* characters, then the name will be encoded using the specified
* charset as per RFC 2047. If the name contains only US-ASCII
* characters, no encoding is done and the name is used as is.
*
* @param name personal name
* @param charset MIME charset to be used to encode the name as
* per RFC 2047
* @see #setPersonal(String)
* @exception UnsupportedEncodingException if the charset encoding
* fails.
*/
public void setPersonal(String name, String charset)
throws UnsupportedEncodingException {
personal = name;
if (name != null)
encodedPersonal = MimeUtility.encodeWord(name, charset, null);
else
encodedPersonal = null;
}
/**
* Set the personal name. If the name contains non US-ASCII
* characters, then the name will be encoded using the platform's
* default charset. If the name contains only US-ASCII characters,
* no encoding is done and the name is used as is.
*
* @param name personal name
* @see #setPersonal(String name, String charset)
* @exception UnsupportedEncodingException if the charset encoding
* fails.
*/
public void setPersonal(String name)
throws UnsupportedEncodingException {
personal = name;
if (name != null)
encodedPersonal = MimeUtility.encodeWord(name);
else
encodedPersonal = null;
}
/**
* Get the email address.
* @return email address
*/
public String getAddress() {
return address;
}
/**
* Get the personal name. If the name is encoded as per RFC 2047,
* it is decoded and converted into Unicode. If the decoding or
* conversion fails, the raw data is returned as is.
*
* @return personal name
*/
public String getPersonal() {
if (personal != null)
return personal;
if (encodedPersonal != null) {
try {
personal = MimeUtility.decodeText(encodedPersonal);
return personal;
} catch (Exception ex) {
// 1. ParseException: either its an unencoded string or
// it can't be parsed
// 2. UnsupportedEncodingException: can't decode it.
return encodedPersonal;
}
}
// No personal or encodedPersonal, return null
return null;
}
/**
* Convert this address into a RFC 822 / RFC 2047 encoded address.
* The resulting string contains only US-ASCII characters, and
* hence is mail-safe.
*
* @return possibly encoded address string
*/
@Override
public String toString() {
String a = address == null ? "" : address;
if (encodedPersonal == null && personal != null)
try {
encodedPersonal = MimeUtility.encodeWord(personal);
} catch (UnsupportedEncodingException ex) { }
if (encodedPersonal != null)
return quotePhrase(encodedPersonal) + " <" + a + ">";
else if (isGroup() || isSimple())
return a;
else
return "<" + a + ">";
}
/**
* Returns a properly formatted address (RFC 822 syntax) of
* Unicode characters.
*
* @return Unicode address string
* @since JavaMail 1.2
*/
public String toUnicodeString() {
String p = getPersonal();
if (p != null)
return quotePhrase(p) + " <" + address + ">";
else if (isGroup() || isSimple())
return address;
else
return "<" + address + ">";
}
/*
* quotePhrase() quotes the words within a RFC822 phrase.
*
* This is tricky, since a phrase is defined as 1 or more
* RFC822 words, separated by LWSP. Now, a word that contains
* LWSP is supposed to be quoted, and this is exactly what the
* MimeUtility.quote() method does. However, when dealing with
* a phrase, any LWSP encountered can be construed to be the
* separator between words, and not part of the words themselves.
* To deal with this funkiness, we have the below variant of
* MimeUtility.quote(), which essentially ignores LWSP when
* deciding whether to quote a word.
*
* It aint pretty, but it gets the job done :)
*/
private static final String rfc822phrase =
HeaderTokenizer.RFC822.replace(' ', '\0').replace('\t', '\0');
private static String quotePhrase(String phrase) {
int len = phrase.length();
boolean needQuoting = false;
for (int i = 0; i < len; i++) {
char c = phrase.charAt(i);
if (c == '"' || c == '\\') {
// need to escape them and then quote the whole string
StringBuffer sb = new StringBuffer(len + 3);
sb.append('"');
for (int j = 0; j < len; j++) {
char cc = phrase.charAt(j);
if (cc == '"' || cc == '\\')
// Escape the character
sb.append('\\');
sb.append(cc);
}
sb.append('"');
return sb.toString();
} else if ((c < 040 && c != '\r' && c != '\n' && c != '\t') ||
(c >= 0177 && !allowUtf8) || rfc822phrase.indexOf(c) >= 0)
// These characters cause the string to be quoted
needQuoting = true;
}
if (needQuoting) {
StringBuffer sb = new StringBuffer(len + 2);
sb.append('"').append(phrase).append('"');
return sb.toString();
} else
return phrase;
}
private static String unquote(String s) {
if (s.startsWith("\"") && s.endsWith("\"") && s.length() > 1) {
s = s.substring(1, s.length() - 1);
// check for any escaped characters
if (s.indexOf('\\') >= 0) {
StringBuffer sb = new StringBuffer(s.length()); // approx
for (int i = 0; i < s.length(); i++) {
char c = s.charAt(i);
if (c == '\\' && i < s.length() - 1)
c = s.charAt(++i);
sb.append(c);
}
s = sb.toString();
}
}
return s;
}
/**
* The equality operator.
*/
@Override
public boolean equals(Object a) {
if (!(a instanceof InternetAddress))
return false;
String s = ((InternetAddress)a).getAddress();
if (s == address)
return true;
if (address != null && address.equalsIgnoreCase(s))
return true;
return false;
}
/**
* Compute a hash code for the address.
*/
@Override
public int hashCode() {
if (address == null)
return 0;
else
return address.toLowerCase(Locale.ENGLISH).hashCode();
}
/**
* Convert the given array of InternetAddress objects into
* a comma separated sequence of address strings. The
* resulting string contains only US-ASCII characters, and
* hence is mail-safe.
*
* @param addresses array of InternetAddress objects
* @exception ClassCastException if any address object in the
* given array is not an InternetAddress object. Note
* that this is a RuntimeException.
* @return comma separated string of addresses
*/
public static String toString(Address[] addresses) {
return toString(addresses, 0);
}
/**
* Convert the given array of InternetAddress objects into
* a comma separated sequence of address strings. The
* resulting string contains Unicode characters.
*
* @param addresses array of InternetAddress objects
* @exception ClassCastException if any address object in the
* given array is not an InternetAddress object. Note
* that this is a RuntimeException.
* @return comma separated string of addresses
* @since JavaMail 1.6
*/
public static String toUnicodeString(Address[] addresses) {
return toUnicodeString(addresses, 0);
}
/**
* Convert the given array of InternetAddress objects into
* a comma separated sequence of address strings. The
* resulting string contains only US-ASCII characters, and
* hence is mail-safe.
*
* The 'used' parameter specifies the number of character positions
* already taken up in the field into which the resulting address
* sequence string is to be inserted. It is used to determine the
* line-break positions in the resulting address sequence string.
*
* @param addresses array of InternetAddress objects
* @param used number of character positions already used, in
* the field into which the address string is to
* be inserted.
* @exception ClassCastException if any address object in the
* given array is not an InternetAddress object. Note
* that this is a RuntimeException.
* @return comma separated string of addresses
*/
public static String toString(Address[] addresses, int used) {
if (addresses == null || addresses.length == 0)
return null;
StringBuilder sb = new StringBuilder();
for (int i = 0; i < addresses.length; i++) {
if (i != 0) { // need to append comma
sb.append(", ");
used += 2;
}
// prefer not to split a single address across lines so used=0 below
String s = MimeUtility.fold(0, addresses[i].toString());
int len = lengthOfFirstSegment(s); // length till CRLF
if (used + len > 76) { // overflows ...
// smash trailing space from ", " above
int curlen = sb.length();
if (curlen > 0 && sb.charAt(curlen - 1) == ' ')
sb.setLength(curlen - 1);
sb.append("\r\n\t"); // .. start new continuation line
used = 8; // account for the starting char
}
sb.append(s);
used = lengthOfLastSegment(s, used);
}
return sb.toString();
}
/**
* Convert the given array of InternetAddress objects into
* a comma separated sequence of address strings. The
* resulting string contains Unicode characters.
*
* The 'used' parameter specifies the number of character positions
* already taken up in the field into which the resulting address
* sequence string is to be inserted. It is used to determine the
* line-break positions in the resulting address sequence string.
*
* @param addresses array of InternetAddress objects
* @param used number of character positions already used, in
* the field into which the address string is to
* be inserted.
* @exception ClassCastException if any address object in the
* given array is not an InternetAddress object. Note
* that this is a RuntimeException.
* @return comma separated string of addresses
* @since JavaMail 1.6
*/
/*
* XXX - This is exactly the same as the above, except it uses
* toUnicodeString instead of toString.
* XXX - Since the line length restrictions are in bytes, not characters,
* we convert all non-ASCII addresses to UTF-8 byte strings,
* which we then convert to ISO-8859-1 Strings where every
* character respresents one UTF-8 byte. At the end we reverse
* the conversion to get back to a correct Unicode string.
* This is a hack to allow all the other character-based methods
* to work properly with UTF-8 bytes.
*/
public static String toUnicodeString(Address[] addresses, int used) {
if (addresses == null || addresses.length == 0)
return null;
StringBuilder sb = new StringBuilder();
boolean sawNonAscii = false;
for (int i = 0; i < addresses.length; i++) {
if (i != 0) { // need to append comma
sb.append(", ");
used += 2;
}
// prefer not to split a single address across lines so used=0 below
String as = ((InternetAddress)addresses[i]).toUnicodeString();
if (MimeUtility.checkAscii(as) != MimeUtility.ALL_ASCII) {
sawNonAscii = true;
as = new String(as.getBytes(StandardCharsets.UTF_8),
StandardCharsets.ISO_8859_1);
}
String s = MimeUtility.fold(0, as);
int len = lengthOfFirstSegment(s); // length till CRLF
if (used + len > 76) { // overflows ...
// smash trailing space from ", " above
int curlen = sb.length();
if (curlen > 0 && sb.charAt(curlen - 1) == ' ')
sb.setLength(curlen - 1);
sb.append("\r\n\t"); // .. start new continuation line
used = 8; // account for the starting char
}
sb.append(s);
used = lengthOfLastSegment(s, used);
}
String ret = sb.toString();
if (sawNonAscii)
ret = new String(ret.getBytes(StandardCharsets.ISO_8859_1),
StandardCharsets.UTF_8);
return ret;
}
/*
* Return the length of the first segment within this string.
* If no segments exist, the length of the whole line is returned.
*/
private static int lengthOfFirstSegment(String s) {
int pos;
if ((pos = s.indexOf("\r\n")) != -1)
return pos;
else
return s.length();
}
/*
* Return the length of the last segment within this string.
* If no segments exist, the length of the whole line plus
* used
is returned.
*/
private static int lengthOfLastSegment(String s, int used) {
int pos;
if ((pos = s.lastIndexOf("\r\n")) != -1)
return s.length() - pos - 2;
else
return s.length() + used;
}
/**
* Return an InternetAddress object representing the current user.
* The entire email address may be specified in the "mail.from"
* property. If not set, the "mail.user" and "mail.host" properties
* are tried. If those are not set, the "user.name" property and
* InetAddress.getLocalHost
method are tried.
* Security exceptions that may occur while accessing this information
* are ignored. If it is not possible to determine an email address,
* null is returned.
*
* @param session Session object used for property lookup
* @return current user's email address
*/
public static InternetAddress getLocalAddress(Session session) {
try {
return _getLocalAddress(session);
} catch (SecurityException sex) { // ignore it
} catch (AddressException ex) { // ignore it
} catch (UnknownHostException ex) { } // ignore it
return null;
}
/**
* A package-private version of getLocalAddress that doesn't swallow
* the exception. Used by MimeMessage.setFrom() to report the reason
* for the failure.
*/
// package-private
static InternetAddress _getLocalAddress(Session session)
throws SecurityException, AddressException, UnknownHostException {
String user = null, host = null, address = null;
if (session == null) {
user = System.getProperty("user.name");
host = getLocalHostName();
} else {
address = session.getProperty("mail.from");
if (address == null) {
user = session.getProperty("mail.user");
if (user == null || user.length() == 0)
user = session.getProperty("user.name");
if (user == null || user.length() == 0)
user = System.getProperty("user.name");
host = session.getProperty("mail.host");
if (host == null || host.length() == 0)
host = getLocalHostName();
}
}
if (address == null && user != null && user.length() != 0 &&
host != null && host.length() != 0)
address = MimeUtility.quote(user.trim(), specialsNoDot + "\t ") +
"@" + host;
if (address == null)
return null;
return new InternetAddress(address);
}
/**
* Get the local host name from InetAddress and return it in a form
* suitable for use in an email address.
*/
private static String getLocalHostName() throws UnknownHostException {
String host = null;
InetAddress me = InetAddress.getLocalHost();
if (me != null) {
// try canonical host name first
if (useCanonicalHostName)
host = me.getCanonicalHostName();
if (host == null)
host = me.getHostName();
// if we can't get our name, use local address literal
if (host == null)
host = me.getHostAddress();
if (host != null && host.length() > 0 && isInetAddressLiteral(host))
host = '[' + host + ']';
}
return host;
}
/**
* Is the address an IPv4 or IPv6 address literal, which needs to
* be enclosed in "[]" in an email address? IPv4 literals contain
* decimal digits and dots, IPv6 literals contain hex digits, dots,
* and colons. We're lazy and don't check the exact syntax, just
* the allowed characters; strings that have only the allowed
* characters in a literal but don't meet the syntax requirements
* for a literal definitely can't be a host name and thus will fail
* later when used as an address literal.
*/
private static boolean isInetAddressLiteral(String addr) {
boolean sawHex = false, sawColon = false;
for (int i = 0; i < addr.length(); i++) {
char c = addr.charAt(i);
if (c >= '0' && c <= '9')
; // digits always ok
else if (c == '.')
; // dot always ok
else if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
sawHex = true; // need to see a colon too
else if (c == ':')
sawColon = true;
else
return false; // anything else, definitely not a literal
}
return !sawHex || sawColon;
}
/**
* Parse the given comma separated sequence of addresses into
* InternetAddress objects. Addresses must follow RFC822 syntax.
*
* @param addresslist comma separated address strings
* @return array of InternetAddress objects
* @exception AddressException if the parse failed
*/
public static InternetAddress[] parse(String addresslist)
throws AddressException {
return parse(addresslist, true);
}
/**
* Parse the given sequence of addresses into InternetAddress
* objects. If strict
is false, simple email addresses
* separated by spaces are also allowed. If strict
is
* true, many (but not all) of the RFC822 syntax rules are enforced.
* In particular, even if strict
is true, addresses
* composed of simple names (with no "@domain" part) are allowed.
* Such "illegal" addresses are not uncommon in real messages.
*
* Non-strict parsing is typically used when parsing a list of
* mail addresses entered by a human. Strict parsing is typically
* used when parsing address headers in mail messages.
*
* @param addresslist comma separated address strings
* @param strict enforce RFC822 syntax
* @return array of InternetAddress objects
* @exception AddressException if the parse failed
*/
public static InternetAddress[] parse(String addresslist, boolean strict)
throws AddressException {
return parse(addresslist, strict, false);
}
/**
* Parse the given sequence of addresses into InternetAddress
* objects. If strict
is false, the full syntax rules for
* individual addresses are not enforced. If strict
is
* true, many (but not all) of the RFC822 syntax rules are enforced.
*
* To better support the range of "invalid" addresses seen in real
* messages, this method enforces fewer syntax rules than the
* parse
method when the strict flag is false
* and enforces more rules when the strict flag is true. If the
* strict flag is false and the parse is successful in separating out an
* email address or addresses, the syntax of the addresses themselves
* is not checked.
*
* @param addresslist comma separated address strings
* @param strict enforce RFC822 syntax
* @return array of InternetAddress objects
* @exception AddressException if the parse failed
* @since JavaMail 1.3
*/
public static InternetAddress[] parseHeader(String addresslist,
boolean strict) throws AddressException {
return parse(MimeUtility.unfold(addresslist), strict, true);
}
/*
* RFC822 Address parser.
*
* XXX - This is complex enough that it ought to be a real parser,
* not this ad-hoc mess, and because of that, this is not perfect.
*
* XXX - Deal with encoded Headers too.
*/
@SuppressWarnings("fallthrough")
private static InternetAddress[] parse(String s, boolean strict,
boolean parseHdr) throws AddressException {
int start, end, index, nesting;
int start_personal = -1, end_personal = -1;
int length = s.length();
boolean ignoreErrors = parseHdr && !strict;
boolean in_group = false; // we're processing a group term
boolean route_addr = false; // address came from route-addr term
boolean rfc822 = false; // looks like an RFC822 address
char c;
List v = new ArrayList<>();
InternetAddress ma;
for (start = end = -1, index = 0; index < length; index++) {
c = s.charAt(index);
switch (c) {
case '(': // We are parsing a Comment. Ignore everything inside.
// XXX - comment fields should be parsed as whitespace,
// more than one allowed per address
rfc822 = true;
if (start >= 0 && end == -1)
end = index;
int pindex = index;
for (index++, nesting = 1; index < length && nesting > 0;
index++) {
c = s.charAt(index);
switch (c) {
case '\\':
index++; // skip both '\' and the escaped char
break;
case '(':
nesting++;
break;
case ')':
nesting--;
break;
default:
break;
}
}
if (nesting > 0) {
if (!ignoreErrors)
throw new AddressException("Missing ')'", s, index);
// pretend the first paren was a regular character and
// continue parsing after it
index = pindex + 1;
break;
}
index--; // point to closing paren
if (start_personal == -1)
start_personal = pindex + 1;
if (end_personal == -1)
end_personal = index;
break;
case ')':
if (!ignoreErrors)
throw new AddressException("Missing '('", s, index);
// pretend the left paren was a regular character and
// continue parsing
if (start == -1)
start = index;
break;
case '<':
rfc822 = true;
if (route_addr) {
if (!ignoreErrors)
throw new AddressException(
"Extra route-addr", s, index);
// assume missing comma between addresses
if (start == -1) {
route_addr = false;
rfc822 = false;
start = end = -1;
break; // nope, nothing there
}
if (!in_group) {
// got a token, add this to our InternetAddress list
if (end == -1) // should never happen
end = index;
String addr = s.substring(start, end).trim();
ma = new InternetAddress();
ma.setAddress(addr);
if (start_personal >= 0) {
ma.encodedPersonal = unquote(
s.substring(start_personal, end_personal).
trim());
}
v.add(ma);
route_addr = false;
rfc822 = false;
start = end = -1;
start_personal = end_personal = -1;
// continue processing this new address...
}
}
int rindex = index;
boolean inquote = false;
outf:
for (index++; index < length; index++) {
c = s.charAt(index);
switch (c) {
case '\\': // XXX - is this needed?
index++; // skip both '\' and the escaped char
break;
case '"':
inquote = !inquote;
break;
case '>':
if (inquote)
continue;
break outf; // out of for loop
default:
break;
}
}
// did we find a matching quote?
if (inquote) {
if (!ignoreErrors)
throw new AddressException("Missing '\"'", s, index);
// didn't find matching quote, try again ignoring quotes
// (e.g., ``<"@foo.com>'')
outq:
for (index = rindex + 1; index < length; index++) {
c = s.charAt(index);
if (c == '\\') // XXX - is this needed?
index++; // skip both '\' and the escaped char
else if (c == '>')
break;
}
}
// did we find a terminating '>'?
if (index >= length) {
if (!ignoreErrors)
throw new AddressException("Missing '>'", s, index);
// pretend the "<" was a regular character and
// continue parsing after it (e.g., ``<@foo.com'')
index = rindex + 1;
if (start == -1)
start = rindex; // back up to include "<"
break;
}
if (!in_group) {
if (start >= 0) {
// seen some characters? use them as the personal name
start_personal = start;
end_personal = rindex;
}
start = rindex + 1;
}
route_addr = true;
end = index;
break;
case '>':
if (!ignoreErrors)
throw new AddressException("Missing '<'", s, index);
// pretend the ">" was a regular character and
// continue parsing (e.g., ``>@foo.com'')
if (start == -1)
start = index;
break;
case '"': // parse quoted string
int qindex = index;
rfc822 = true;
if (start == -1)
start = index;
outq:
for (index++; index < length; index++) {
c = s.charAt(index);
switch (c) {
case '\\':
index++; // skip both '\' and the escaped char
break;
case '"':
break outq; // out of for loop
default:
break;
}
}
if (index >= length) {
if (!ignoreErrors)
throw new AddressException("Missing '\"'", s, index);
// pretend the quote was a regular character and
// continue parsing after it (e.g., ``"@foo.com'')
index = qindex + 1;
}
break;
case '[': // a domain-literal, probably
rfc822 = true;
int lindex = index;
outb:
for (index++; index < length; index++) {
c = s.charAt(index);
switch (c) {
case '\\':
index++; // skip both '\' and the escaped char
break;
case ']':
break outb; // out of for loop
default:
break;
}
}
if (index >= length) {
if (!ignoreErrors)
throw new AddressException("Missing ']'", s, index);
// pretend the "[" was a regular character and
// continue parsing after it (e.g., ``[@foo.com'')
index = lindex + 1;
}
break;
case ';':
if (start == -1) {
route_addr = false;
rfc822 = false;
start = end = -1;
break; // nope, nothing there
}
if (in_group) {
in_group = false;
/*
* If parsing headers, but not strictly, peek ahead.
* If next char is "@", treat the group name
* like the local part of the address, e.g.,
* "Undisclosed-Recipient:;@java.sun.com".
*/
if (parseHdr && !strict &&
index + 1 < length && s.charAt(index + 1) == '@')
break;
ma = new InternetAddress();
end = index + 1;
ma.setAddress(s.substring(start, end).trim());
v.add(ma);
route_addr = false;
rfc822 = false;
start = end = -1;
start_personal = end_personal = -1;
break;
}
if (!ignoreErrors)
throw new AddressException(
"Illegal semicolon, not in group", s, index);
// otherwise, parsing a header; treat semicolon like comma
// fall through to comma case...
case ',': // end of an address, probably
if (start == -1) {
route_addr = false;
rfc822 = false;
start = end = -1;
break; // nope, nothing there
}
if (in_group) {
route_addr = false;
break;
}
// got a token, add this to our InternetAddress list
if (end == -1)
end = index;
String addr = s.substring(start, end).trim();
String pers = null;
if (rfc822 && start_personal >= 0) {
pers = unquote(
s.substring(start_personal, end_personal).trim());
if (pers.trim().length() == 0)
pers = null;
}
/*
* If the personal name field has an "@" and the address
* field does not, assume they were reversed, e.g.,
* ``"joe doe" ([email protected])''.
*/
if (parseHdr && !strict && pers != null &&
pers.indexOf('@') >= 0 &&
addr.indexOf('@') < 0 && addr.indexOf('!') < 0) {
String tmp = addr;
addr = pers;
pers = tmp;
}
if (rfc822 || strict || parseHdr) {
if (!ignoreErrors)
checkAddress(addr, route_addr, false);
ma = new InternetAddress();
ma.setAddress(addr);
if (pers != null)
ma.encodedPersonal = pers;
v.add(ma);
} else {
// maybe we passed over more than one space-separated addr
StringTokenizer st = new StringTokenizer(addr);
while (st.hasMoreTokens()) {
String a = st.nextToken();
checkAddress(a, false, false);
ma = new InternetAddress();
ma.setAddress(a);
v.add(ma);
}
}
route_addr = false;
rfc822 = false;
start = end = -1;
start_personal = end_personal = -1;
break;
case ':':
rfc822 = true;
if (in_group)
if (!ignoreErrors)
throw new AddressException("Nested group", s, index);
if (start == -1)
start = index;
if (parseHdr && !strict) {
/*
* If next char is a special character that can't occur at
* the start of a valid address, treat the group name
* as the entire address, e.g., "Date:, Tue", "Re:@foo".
*/
if (index + 1 < length) {
String addressSpecials = ")>[]:@\\,.";
char nc = s.charAt(index + 1);
if (addressSpecials.indexOf(nc) >= 0) {
if (nc != '@')
break; // don't change in_group
/*
* Handle a common error:
* ``Undisclosed-Recipient:@example.com;''
*
* Scan ahead. If we find a semicolon before
* one of these other special characters,
* consider it to be a group after all.
*/
for (int i = index + 2; i < length; i++) {
nc = s.charAt(i);
if (nc == ';')
break;
if (addressSpecials.indexOf(nc) >= 0)
break;
}
if (nc == ';')
break; // don't change in_group
}
}
// ignore bogus "mailto:" prefix in front of an address,
// or bogus mail header name included in the address field
String gname = s.substring(start, index);
if (ignoreBogusGroupName &&
(gname.equalsIgnoreCase("mailto") ||
gname.equalsIgnoreCase("From") ||
gname.equalsIgnoreCase("To") ||
gname.equalsIgnoreCase("Cc") ||
gname.equalsIgnoreCase("Subject") ||
gname.equalsIgnoreCase("Re")))
start = -1; // we're not really in a group
else
in_group = true;
} else
in_group = true;
break;
// Ignore whitespace
case ' ':
case '\t':
case '\r':
case '\n':
break;
default:
if (start == -1)
start = index;
break;
}
}
if (start >= 0) {
/*
* The last token, add this to our InternetAddress list.
* Note that this block of code should be identical to the
* block above for "case ','".
*/
if (end == -1)
end = length;
String addr = s.substring(start, end).trim();
String pers = null;
if (rfc822 && start_personal >= 0) {
pers = unquote(
s.substring(start_personal, end_personal).trim());
if (pers.trim().length() == 0)
pers = null;
}
/*
* If the personal name field has an "@" and the address
* field does not, assume they were reversed, e.g.,
* ``"joe doe" ([email protected])''.
*/
if (parseHdr && !strict &&
pers != null && pers.indexOf('@') >= 0 &&
addr.indexOf('@') < 0 && addr.indexOf('!') < 0) {
String tmp = addr;
addr = pers;
pers = tmp;
}
if (rfc822 || strict || parseHdr) {
if (!ignoreErrors)
checkAddress(addr, route_addr, false);
ma = new InternetAddress();
ma.setAddress(addr);
if (pers != null)
ma.encodedPersonal = pers;
v.add(ma);
} else {
// maybe we passed over more than one space-separated addr
StringTokenizer st = new StringTokenizer(addr);
while (st.hasMoreTokens()) {
String a = st.nextToken();
checkAddress(a, false, false);
ma = new InternetAddress();
ma.setAddress(a);
v.add(ma);
}
}
}
InternetAddress[] a = new InternetAddress[v.size()];
v.toArray(a);
return a;
}
/**
* Validate that this address conforms to the syntax rules of
* RFC 822. The current implementation checks many, but not
* all, syntax rules. Note that even though the syntax of
* the address may be correct, there's no guarantee that a
* mailbox of that name exists.
*
* @exception AddressException if the address isn't valid.
* @since JavaMail 1.3
*/
public void validate() throws AddressException {
if (isGroup())
getGroup(true); // throw away the result
else
checkAddress(getAddress(), true, true);
}
private static final String specialsNoDotNoAt = "()<>,;:\\\"[]";
private static final String specialsNoDot = specialsNoDotNoAt + "@";
/**
* Check that the address is a valid "mailbox" per RFC822.
* (We also allow simple names.)
*
* XXX - much more to check
* XXX - doesn't handle domain-literals properly (but no one uses them)
*/
private static void checkAddress(String addr,
boolean routeAddr, boolean validate)
throws AddressException {
int i, start = 0;
if (addr == null)
throw new AddressException("Address is null");
int len = addr.length();
if (len == 0)
throw new AddressException("Empty address", addr);
/*
* routeAddr indicates that the address is allowed
* to have an RFC 822 "route".
*/
if (routeAddr && addr.charAt(0) == '@') {
/*
* Check for a legal "route-addr":
* [@domain[,@domain ...]:]local@domain
*/
for (start = 0; (i = indexOfAny(addr, ",:", start)) >= 0;
start = i+1) {
if (addr.charAt(start) != '@')
throw new AddressException("Illegal route-addr", addr);
if (addr.charAt(i) == ':') {
// end of route-addr
start = i + 1;
break;
}
}
}
/*
* The rest should be "local@domain", but we allow simply "local"
* unless called from validate.
*
* local-part must follow RFC 822 - no specials except '.'
* unless quoted.
*/
char c = (char)-1;
char lastc = (char)-1;
boolean inquote = false;
for (i = start; i < len; i++) {
lastc = c;
c = addr.charAt(i);
// a quoted-pair is only supposed to occur inside a quoted string,
// but some people use it outside so we're more lenient
if (c == '\\' || lastc == '\\')
continue;
if (c == '"') {
if (inquote) {
// peek ahead, next char must be "@"
if (validate && i + 1 < len && addr.charAt(i + 1) != '@')
throw new AddressException(
"Quote not at end of local address", addr);
inquote = false;
} else {
if (validate && i != 0)
throw new AddressException(
"Quote not at start of local address", addr);
inquote = true;
}
continue;
} else if (c == '\r') {
// peek ahead, next char must be LF
if (i + 1 < len && addr.charAt(i + 1) != '\n')
throw new AddressException(
"Quoted local address contains CR without LF", addr);
} else if (c == '\n') {
/*
* CRLF followed by whitespace is allowed in a quoted string.
* We allowed naked LF, but ensure LF is always followed by
* whitespace to prevent spoofing the end of the header.
*/
if (i + 1 < len && addr.charAt(i + 1) != ' ' &&
addr.charAt(i + 1) != '\t')
throw new AddressException(
"Quoted local address contains newline without whitespace",
addr);
}
if (inquote)
continue;
if (c == '@') {
if (i == 0)
throw new AddressException("Missing local name", addr);
break; // done with local part
}
if (c <= 040 || c == 0177)
throw new AddressException(
"Local address contains control or whitespace", addr);
if (specialsNoDot.indexOf(c) >= 0)
throw new AddressException(
"Local address contains illegal character", addr);
}
if (inquote)
throw new AddressException("Unterminated quote", addr);
/*
* Done with local part, now check domain.
*
* Note that the MimeMessage class doesn't remember addresses
* as separate objects; it writes them out as headers and then
* parses the headers when the addresses are requested.
* In order to support the case where a "simple" address is used,
* but the address also has a personal name and thus looks like
* it should be a valid RFC822 address when parsed, we only check
* this if we're explicitly called from the validate method.
*/
if (c != '@') {
if (validate)
throw new AddressException("Missing final '@domain'", addr);
return;
}
// check for illegal chars in the domain, but ignore domain literals
start = i + 1;
if (start >= len)
throw new AddressException("Missing domain", addr);
if (addr.charAt(start) == '.')
throw new AddressException("Domain starts with dot", addr);
boolean inliteral = false;
for (i = start; i < len; i++) {
c = addr.charAt(i);
if (c == '[') {
if (i != start)
throw new AddressException(
"Domain literal not at start of domain", addr);
inliteral = true; // domain literal, don't validate
} else if (c == ']') {
if (i != len - 1)
throw new AddressException(
"Domain literal end not at end of domain", addr);
inliteral = false;
} else if (c <= 040 || c == 0177) {
throw new AddressException(
"Domain contains control or whitespace", addr);
} else {
// RFC 2822 rule
//if (specialsNoDot.indexOf(c) >= 0)
/*
* RFC 1034 rule is more strict
* the full rule is:
*
* ::= | " "
* ::=