I am parsing XML returned from a website but sadly it is slightly malformed. I am getting XML like:
<tag attrib="Buy two for £1" />
Which, I am informed, is invalid because £ is an HTML character, not an XML character and definitely cannot appear in an attribute.
What can I do to fix this, assuming I cannot tell the website to obey the rules? I am considering using a FilterInputStream to filter the arriving data before it gets to the SAX parser but this seems over the top.
In the end I failed to do this with the parser. My solution was to write a FilterInputStream that converted all &xxxx; references into their &#nnnn; form.
/* Cleans up often very bad xml.
*
* 1. Strips leading white space.
* 2. Recodes £ etc to &#...;.
* 3. Recodes lone & as &.
*
*/
public class XMLInputStream extends FilterInputStream {
private static final int MIN_LENGTH = 2;
// Everything we've read.
StringBuilder red = new StringBuilder();
// Data I have pushed back.
StringBuilder pushBack = new StringBuilder();
// How much we've given them.
int given = 0;
// How much we've read.
int pulled = 0;
public XMLInputStream(InputStream in) {
super(in);
}
public int length() {
// NB: This is a Troll length (i.e. it goes 1, 2, many) so 2 actually means "at least 2"
try {
StringBuilder s = read(MIN_LENGTH);
pushBack.append(s);
return s.length();
} catch (IOException ex) {
log.warning("Oops ", ex);
}
return 0;
}
private StringBuilder read(int n) throws IOException {
// Input stream finished?
boolean eof = false;
// Read that many.
StringBuilder s = new StringBuilder(n);
while (s.length() < n && !eof) {
// Always get from the pushBack buffer.
if (pushBack.length() == 0) {
// Read something from the stream into pushBack.
eof = readIntoPushBack();
}
// Pushback only contains deliverable codes.
if (pushBack.length() > 0) {
// Grab one character
s.append(pushBack.charAt(0));
// Remove it from pushBack
pushBack.deleteCharAt(0);
}
}
return s;
}
// Returns true at eof.
// Might not actually push back anything but usually will.
private boolean readIntoPushBack() throws IOException {
// File finished?
boolean eof = false;
// Next char.
int ch = in.read();
if (ch >= 0) {
// Discard whitespace at start?
if (!(pulled == 0 && isWhiteSpace(ch))) {
// Good code.
pulled += 1;
// Parse out the &stuff;
if (ch == '&') {
// Process the &
readAmpersand();
} else {
// Not an '&', just append.
pushBack.append((char) ch);
}
}
} else {
// Hit end of file.
eof = true;
}
return eof;
}
// Deal with an ampersand in the stream.
private void readAmpersand() throws IOException {
// Read the whole word, up to and including the ;
StringBuilder reference = new StringBuilder();
int ch;
// Should end in a ';'
for (ch = in.read(); isAlphaNumeric(ch); ch = in.read()) {
reference.append((char) ch);
}
// Did we tidily finish?
if (ch == ';') {
// Yes! Translate it into a &#nnn; code.
String code = XML.hash(reference);
if (code != null) {
// Keep it.
pushBack.append(code);
} else {
throw new IOException("Invalid/Unknown reference '&" + reference + ";'");
}
} else {
// Did not terminate properly!
// Perhaps an & on its own or a malformed reference.
// Either way, escape the &
pushBack.append("&").append(reference).append((char) ch);
}
}
private void given(CharSequence s, int wanted, int got) {
// Keep track of what we've given them.
red.append(s);
given += got;
log.finer("Given: [" + wanted + "," + got + "]-" + s);
}
@Override
public int read() throws IOException {
StringBuilder s = read(1);
given(s, 1, 1);
return s.length() > 0 ? s.charAt(0) : -1;
}
@Override
public int read(byte[] data, int offset, int length) throws IOException {
int n = 0;
StringBuilder s = read(length);
for (int i = 0; i < Math.min(length, s.length()); i++) {
data[offset + i] = (byte) s.charAt(i);
n += 1;
}
given(s, length, n);
return n > 0 ? n : -1;
}
@Override
public String toString() {
String s = red.toString();
String h = "";
// Hex dump the small ones.
if (s.length() < 8) {
// Separator just inserts the string between each.
Separator sep = new Separator(" ");
for (int i = 0; i < s.length(); i++) {
h += sep.sep() + Integer.toHexString(s.charAt(i));
}
}
return "[" + given + "]-\"" + s + "\"" + (h.length() > 0 ? " (" + h + ")" : "");
}
private boolean isWhiteSpace(int ch) {
switch (ch) {
case ' ':
case '\r':
case '\n':
case '\t':
return true;
}
return false;
}
private boolean isAlphaNumeric(int ch) {
return ('a' <= ch && ch <= 'z')
|| ('A' <= ch && ch <= 'Z')
|| ('0' <= ch && ch <= '9');
}
}
XML.java - For completeness. Please confirm the completeness of the list.
public static String hash(CharSequence s) {
final Integer code = SPECIALS.get(s.toString());
if (code != null) {
return "&#" + code + ";";
}
return null;
}
private static final Map<String, Integer> SPECIALS;
static {
// Derived from Wikipedia http://en.wikipedia.org/wiki/List_of_XML_and_HTML_character_entity_references
final Map<String, Integer> map = new HashMap<>();
map.put("quot", 34);
map.put("amp", 38);
map.put("apos", 39);
map.put("lt", 60);
map.put("gt", 62);
map.put("nbsp", 160);
map.put("iexcl", 161);
map.put("cent", 162);
map.put("pound", 163);
map.put("curren", 164);
map.put("yen", 165);
map.put("brvbar", 166);
map.put("sect", 167);
map.put("uml", 168);
map.put("copy", 169);
map.put("ordf", 170);
map.put("laquo", 171);
map.put("not", 172);
map.put("shy", 173);
map.put("reg", 174);
map.put("macr", 175);
map.put("deg", 176);
map.put("plusmn", 177);
map.put("sup2", 178);
map.put("sup3", 179);
map.put("acute", 180);
map.put("micro", 181);
map.put("para", 182);
map.put("middot", 183);
map.put("cedil", 184);
map.put("sup1", 185);
map.put("ordm", 186);
map.put("raquo", 187);
map.put("frac14", 188);
map.put("frac12", 189);
map.put("frac34", 190);
map.put("iquest", 191);
map.put("Agrave", 192);
map.put("Aacute", 193);
map.put("Acirc", 194);
map.put("Atilde", 195);
map.put("Auml", 196);
map.put("Aring", 197);
map.put("AElig", 198);
map.put("Ccedil", 199);
map.put("Egrave", 200);
map.put("Eacute", 201);
map.put("Ecirc", 202);
map.put("Euml", 203);
map.put("Igrave", 204);
map.put("Iacute", 205);
map.put("Icirc", 206);
map.put("Iuml", 207);
map.put("ETH", 208);
map.put("Ntilde", 209);
map.put("Ograve", 210);
map.put("Oacute", 211);
map.put("Ocirc", 212);
map.put("Otilde", 213);
map.put("Ouml", 214);
map.put("times", 215);
map.put("Oslash", 216);
map.put("Ugrave", 217);
map.put("Uacute", 218);
map.put("Ucirc", 219);
map.put("Uuml", 220);
map.put("Yacute", 221);
map.put("THORN", 222);
map.put("szlig", 223);
map.put("agrave", 224);
map.put("aacute", 225);
map.put("acirc", 226);
map.put("atilde", 227);
map.put("auml", 228);
map.put("aring", 229);
map.put("aelig", 230);
map.put("ccedil", 231);
map.put("egrave", 232);
map.put("eacute", 233);
map.put("ecirc", 234);
map.put("euml", 235);
map.put("igrave", 236);
map.put("iacute", 237);
map.put("icirc", 238);
map.put("iuml", 239);
map.put("eth", 240);
map.put("ntilde", 241);
map.put("ograve", 242);
map.put("oacute", 243);
map.put("ocirc", 244);
map.put("otilde", 245);
map.put("ouml", 246);
map.put("divide", 247);
map.put("oslash", 248);
map.put("ugrave", 249);
map.put("uacute", 250);
map.put("ucirc", 251);
map.put("uuml", 252);
map.put("yacute", 253);
map.put("thorn", 254);
map.put("yuml", 255);
map.put("OElig", 338);
map.put("oelig", 339);
map.put("Scaron", 352);
map.put("scaron", 353);
map.put("Yuml", 376);
map.put("fnof", 402);
map.put("circ", 710);
map.put("tilde", 732);
map.put("Alpha", 913);
map.put("Beta", 914);
map.put("Gamma", 915);
map.put("Delta", 916);
map.put("Epsilon", 917);
map.put("Zeta", 918);
map.put("Eta", 919);
map.put("Theta", 920);
map.put("Iota", 921);
map.put("Kappa", 922);
map.put("Lambda", 923);
map.put("Mu", 924);
map.put("Nu", 925);
map.put("Xi", 926);
map.put("Omicron", 927);
map.put("Pi", 928);
map.put("Rho", 929);
map.put("Sigma", 931);
map.put("Tau", 932);
map.put("Upsilon", 933);
map.put("Phi", 934);
map.put("Chi", 935);
map.put("Psi", 936);
map.put("Omega", 937);
map.put("alpha", 945);
map.put("beta", 946);
map.put("gamma", 947);
map.put("delta", 948);
map.put("epsilon", 949);
map.put("zeta", 950);
map.put("eta", 951);
map.put("theta", 952);
map.put("iota", 953);
map.put("kappa", 954);
map.put("lambda", 955);
map.put("mu", 956);
map.put("nu", 957);
map.put("xi", 958);
map.put("omicron", 959);
map.put("pi", 960);
map.put("rho", 961);
map.put("sigmaf", 962);
map.put("sigma", 963);
map.put("tau", 964);
map.put("upsilon", 965);
map.put("phi", 966);
map.put("chi", 967);
map.put("psi", 968);
map.put("omega", 969);
map.put("thetasym", 977);
map.put("upsih", 978);
map.put("piv", 982);
map.put("ensp", 8194);
map.put("emsp", 8195);
map.put("thinsp", 8201);
map.put("zwnj", 8204);
map.put("zwj", 8205);
map.put("lrm", 8206);
map.put("rlm", 8207);
map.put("ndash", 8211);
map.put("mdash", 8212);
map.put("lsquo", 8216);
map.put("rsquo", 8217);
map.put("sbquo", 8218);
map.put("ldquo", 8220);
map.put("rdquo", 8221);
map.put("bdquo", 8222);
map.put("dagger", 8224);
map.put("Dagger", 8225);
map.put("bull", 8226);
map.put("hellip", 8230);
map.put("permil", 8240);
map.put("prime", 8242);
map.put("Prime", 8243);
map.put("lsaquo", 8249);
map.put("rsaquo", 8250);
map.put("oline", 8254);
map.put("frasl", 8260);
map.put("euro", 8364);
map.put("image", 8465);
map.put("weierp", 8472);
map.put("real", 8476);
map.put("trade", 8482);
map.put("alefsym", 8501);
map.put("larr", 8592);
map.put("uarr", 8593);
map.put("rarr", 8594);
map.put("darr", 8595);
map.put("harr", 8596);
map.put("crarr", 8629);
map.put("lArr", 8656);
map.put("uArr", 8657);
map.put("rArr", 8658);
map.put("dArr", 8659);
map.put("hArr", 8660);
map.put("forall", 8704);
map.put("part", 8706);
map.put("exist", 8707);
map.put("empty", 8709);
map.put("nabla", 8711);
map.put("isin", 8712);
map.put("notin", 8713);
map.put("ni", 8715);
map.put("prod", 8719);
map.put("sum", 8721);
map.put("minus", 8722);
map.put("lowast", 8727);
map.put("radic", 8730);
map.put("prop", 8733);
map.put("infin", 8734);
map.put("ang", 8736);
map.put("and", 8743);
map.put("or", 8744);
map.put("cap", 8745);
map.put("cup", 8746);
map.put("int", 8747);
map.put("there4", 8756);
map.put("sim", 8764);
map.put("cong", 8773);
map.put("asymp", 8776);
map.put("ne", 8800);
map.put("equiv", 8801);
map.put("le", 8804);
map.put("ge", 8805);
map.put("sub", 8834);
map.put("sup", 8835);
map.put("nsub", 8836);
map.put("sube", 8838);
map.put("supe", 8839);
map.put("oplus", 8853);
map.put("otimes", 8855);
map.put("perp", 8869);
map.put("sdot", 8901);
map.put("lceil", 8968);
map.put("rceil", 8969);
map.put("lfloor", 8970);
map.put("rfloor", 8971);
map.put("lang", 10216);
map.put("rang", 10217);
map.put("loz", 9674);
map.put("spades", 9824);
map.put("clubs", 9827);
map.put("hearts", 9829);
map.put("diams", 9830);
SPECIALS = Collections.unmodifiableMap(map);
}
You could handle this by providing a custom org.xml.sax.EntityResolver to convert the entity £ to a valid character.
EDIT: I did some further research and found that entity references (such as £) are handled directly as events. Configure your parser (through XMLInputFactory) with the feature javax.xml.stream.isReplacingEntityReferences set to FALSE, which prevents it from trying to resolve the entity references. Then, when you parse the input you will get input events for each entity reference as a call to your handler's startEntity(String name) method. The handler must implement org.xml.sax.ext.DefaultHandler2.
One option to resolve your issue is, as Jim Garrison suggested, providing custom EntityResolver. However, it will fix only the concrete issue you described. If your XML will be malformed by e.g. not closed tags, EntityResolver would not fix it. In such case I'd recommend to use one of available HTML "purifiers" in order to fix HTML syntax into XML-valid form. In my opinion the best available is this one: http://nekohtml.sourceforge.net/
