更多优秀文章，请扫码关注个人微信公众号或搜索“程序猿小杨”添加。

一、错误展现：

Caused by: io.debezium.connector.oracle.logminer.parser.DmlParserException: Failed to parse insert DML: 'insert into "HIS_DATA". 
	at io.debezium.connector.oracle.logminer.parser.LogMinerDmlParser.parseInsert(LogMinerDmlParser.java:109)
	at io.debezium.connector.oracle.logminer.parser.LogMinerDmlParser.parse(LogMinerDmlParser.java:73)
	at io.debezium.connector.oracle.logminer.processor.AbstractLogMinerEventProcessor.parseDmlStatement(AbstractLogMinerEventProcessor.java:1078)
	... 16 common frames omitted
Caused by: java.lang.ArrayIndexOutOfBoundsException: 74
	at io.debezium.connector.oracle.logminer.parser.LogMinerDmlParser.parseColumnListClause(LogMinerDmlParser.java:239)
	at io.debezium.connector.oracle.logminer.parser.LogMinerDmlParser.parseInsert(LogMinerDmlParser.java:100)
	... 18 common frames omitted

二、问题原因：

通过分析源代码，发现debezium-connector-oracle-1.9.7.Final中的方法存在问题，造成数组越界，一般方式场景在：oracle某张表表结构发生新增字段，然后flink cdc的任务加载的还是旧的表字段信息，当被监控表业务场景写入数据的时候，由于sql里面已经带了新增字段，但是table里面的字段还是旧的，造成解析错误，源代码如下：

 private int parseColumnListClause(String sql, int start, String[] columnNames) {
        int index = start;
        boolean inQuote = false;

        for(int var6 = 0; index < sql.length(); ++index) {
            char c = sql.charAt(index);
            if (c == '(' && !inQuote) {
                start = index + 1;
            } else {
                if (c == ')' && !inQuote) {
                    ++index;
                    break;
                }

                if (c == '"') {
                    if (inQuote) {
                        inQuote = false;
                        columnNames[var6++] = sql.substring(start + 1, index);
                        start = index + 2;
                    } else {
                        inQuote = true;
                    }
                }
            }
        }

        return index;
    }

三、解决方案：

解决思路：重写io.debezium.connector.oracle.logminer.parser下的LogMinerDmlParser类中的方法即可，方案有两种：

方案1：如果新增字段的内容不需要，比如：只需获取主键信息及对应的值，那就忽略掉新增的字段，新增字段内容不包含在解析后的数据中。那只需要调整一下代码即可：

在代码中新建一个包路径为：io.debezium.connector.oracle.logminer.parser，

复制源码路径下的LogMinerDmlParser类中内容：

//
// Source code recreated from a .class file by IntelliJ IDEA
// (powered by Fernflower decompiler)
//

package io.debezium.connector.oracle.logminer.parser;

import io.debezium.DebeziumException;
import io.debezium.connector.oracle.logminer.LogMinerHelper;
import io.debezium.relational.Column;
import io.debezium.relational.Table;

public class LogMinerDmlParser implements DmlParser {
    private static final String NULL_SENTINEL = "${DBZ_NULL}";
    private static final String NULL = "NULL";
    private static final String INSERT_INTO = "insert into ";
    private static final String UPDATE = "update ";
    private static final String DELETE_FROM = "delete from ";
    private static final String AND = "and ";
    private static final String OR = "or ";
    private static final String SET = " set ";
    private static final String WHERE = " where ";
    private static final String VALUES = " values ";
    private static final String IS_NULL = "IS NULL";
    private static final String UNSUPPORTED = "Unsupported";
    private static final String UNSUPPORTED_TYPE = "Unsupported Type";
    private static final int INSERT_INTO_LENGTH = "insert into ".length();
    private static final int UPDATE_LENGTH = "update ".length();
    private static final int DELETE_FROM_LENGTH = "delete from ".length();
    private static final int VALUES_LENGTH = " values ".length();
    private static final int SET_LENGTH = " set ".length();
    private static final int WHERE_LENGTH = " where ".length();

    public LogMinerDmlParser() {
    }

    public LogMinerDmlEntry parse(String sql, Table table) {
        if (table == null) {
            throw new DmlParserException("DML parser requires a non-null table");
        } else {
            if (sql != null && sql.length() > 0) {
                switch(sql.charAt(0)) {
                case 'd':
                    return this.parseDelete(sql, table);
                case 'i':
                    return this.pa