SOLVED: XSD validation performance disaster

Question

Is it possible to speed-up XMLType schemaValidate() and isSchemaValid() functions somehow?
The following statement take ~one second of CPU time per table row on Pentium4:
update tbl set is_valid = func(xml);
where func(xml) is begin return xml.isSchemaValid('schema.xsd'); end;
Looks like like full XSD re-parse for every row.
The xml is 600 bytes long on average and schema is 40k formated text. The XML column is non-schema based XMLType. The schema name is constant, but I need to process relatively large number of xml fragments.
Server is 10.2.0.3 so bug 5099703 does not apply.
I see two complimentary options:
- load Xerces2-J and validate with grammar caching, and hope it may be faster;
- split the work between dbms_scheduler one-shot jobs to do processing in parallel, or maybe just parallel update will be enough.
But keeping it simple would be the best option at all.

Answer 1

I hope my solution will be in help for someone struggling with pure performance of PL/SQL XMLType validation functions.
The performance in my case is two orders of magnitute better now: 80 rows/sec instead of 1 row/sec.
I created custom function that could be used as:
update messages set xml_is_valid =
    decode(xml_validator.isSchemaValid(xmlserialize(content xml), 'KRFileR.xsd'), 1, 'Y', 'N') The XML_VALIDATOR package is:
create or replace package xml_validator as
function isSchemaValid(xml clob, schema varchar2) return number;
procedure schemaValidate(xml clob, schema varchar2);
end;
show err
create or replace package body xml_validator as
function isSchemaValid(xml clob, schema varchar2) return number
    as language java
    name 'xmlvalidator.XMLValidator.isSchemaValid(
        oracle.sql.CLOB,
        java.lang.String
        ) return int';
procedure schemaValidate(xml clob, schema varchar2)
    as language java
    name 'xmlvalidator.XMLValidator.schemaValidate(
        oracle.sql.CLOB,
        java.lang.String
end;
show errThe XMLValidator class itself:
package xmlvalidator;
import java.io.IOException;
import java.io.ByteArrayOutputStream;
import java.io.StringReader;
import java.util.HashMap;
import java.sql.SQLException;
import oracle.xml.parser.v2.*;
import oracle.xml.parser.schema.*;
import oracle.sql.CLOB;
import org.xml.sax.EntityResolver;
import org.xml.sax.InputSource;
import org.xml.sax.SAXException;
public class XMLValidator {
  //protected static HashMap<String, DOMParser> cache = new HashMap<String, DOMParser>();
    protected static HashMap cache = new HashMap(); // oracle java is 1.4
    protected static EntityResolver resolver = new OracleSchemaEntityResolver();
    protected static ByteArrayOutputStream errors = new ByteArrayOutputStream();
    public static String getErrors() {
        return errors.toString();
    // mimick XMLType functions semantic
    public static int isSchemaValid(CLOB xml, String schema) {
        try {
            schemaValidate(xml, schema);
        } catch (Exception e) {
            return 0;
        return 1;
    public static void schemaValidate(CLOB xml, String schema)
            throws XMLParseException, SAXException, IOException, XSDException, SQLException {
      //DOMParser parser = cache.get(schema);
        DOMParser parser = (DOMParser)cache.get(schema);
        if (parser == null) {
            XSDBuilder xsdbuild = new XSDBuilder();
            xsdbuild.setEntityResolver(resolver);
            XMLSchema xmlschema = xsdbuild.build(resolver.resolveEntity(null, schema));
            parser = new DOMParser();
            parser.setXMLSchema(xmlschema);
            parser.setValidationMode(XMLParser.SCHEMA_VALIDATION);
            parser.setErrorStream(errors);
            cache.put(schema, parser); // the whole parser is cached to avoid unknown
                                       // hidden costs of creating new parser
        errors.reset();
        parser.reset();
        parser.parse(new InputSource(xml.getCharacterStream()));
        parser.reset();
}The OracleSchemaEntityResolver class to pull schema definition from Oracle registered schemas:
package xmlvalidator;
import oracle.xml.parser.schema.*;
import oracle.xml.parser.v2.*;
import java.net.*;
import java.io.*;
import org.w3c.dom.*;
import org.xml.sax.*;
import java.util.*;
import java.sql.*;
import oracle.jdbc.*;
public class OracleSchemaEntityResolver implements EntityResolver {
    protected static Connection conn;
    protected static PreparedStatement stmt;
    static {
        try {
            conn = DriverManager.getConnection("jdbc:default:connection:");
            stmt = conn.prepareStatement(
                "select xmlserialize(document schema) schema from user_xml_schemas where schema_url = ?"
        } catch(Exception e) { throw new RuntimeException(e); }
    public InputSource resolveEntity(String targetNS, String systemId)
            throws SAXException, IOException {
        Reader r;
      //System.out.println("resolving: " + targetNS + ":" + systemId);
        try {
            stmt.setString(1, systemId);
            ResultSet rs = stmt.executeQuery();
            rs.next();
            r = rs.getCharacterStream(1);
            rs.close(); // CLOB locator should be still valid even after ResultSet is closed
        } catch(SQLException e) { throw new RuntimeException(e); }
        InputSource src = new InputSource(r);
        src.setSystemId(systemId);
        return src;
}Create JAR with makejar.bat:
del XMLValidator.jar
cd XMLValidator/src
jar cf ../../XMLValidator.jar xmlvalidator/XMLValidator.java xmlvalidator/OracleSchemaEntityResolver.java
pauseLoad into database with load.bat:
set u=scott/qwerty@//server.domain.com:1521/orcl
set r="((* SCOTT)(* PUBLIC)(* SYS))"
call loadjava -thin -force -user %u% -resolver %r% -resolve XMLValidator.jar
pauseEnjoy!