Answers to: Can not persist Business Data to Repo after changing data type

Comment by Reguel Wermelinger on Reguel Wermelinger's answer

Reguel Wermelinger — Tue, 04 Jul 2017 08:50:20 -0400

This demo is just for illustrative purposes. In a real world scenario you should pay special care that the serializer / deserializer implementation is rock solid. For simple demonstration I've just used a simple field separator here...

Answer by Reguel Wermelinger

Reguel Wermelinger — Tue, 04 Jul 2017 08:29:24 -0400

You do not run into this problem if you just introduce a new attribute for the new type instead of re-using the existing attribute that already contains other data. Business Data is explicitly designed to allow simple data migration. So it does not throw any errors for unknown fields.

However, there is a work around

If you just wrapped the String in a complex object you could work around the problem by influencing the JSON serialization. With Jackson you could for instance store the new complex object also as string and read it back to a complex object. And of course this also allows you to handle legacy serialized objects.

The rough concept is described here: http://developer.axonivy.com/doc/latest/DesignerGuideHtml/ivy.datamodeling.html#ivy.businessdata.serialization.module

A sample implementation could do the following. Here we serialize a complex Person object as a raw String.

FIRST: Write your own serializer module:

package com.axonivy;

import java.io.IOException;

import org.apache.commons.lang3.StringUtils;

import ch.ivyteam.ivy.environment.Ivy;

import com.axonivy.sample.Person;
import com.fasterxml.jackson.core.JsonGenerator;
import com.fasterxml.jackson.core.JsonParser;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.DeserializationContext;
import com.fasterxml.jackson.databind.SerializerProvider;
import com.fasterxml.jackson.databind.deser.std.StdScalarDeserializer;
import com.fasterxml.jackson.databind.module.SimpleModule;
import com.fasterxml.jackson.databind.ser.std.StdScalarSerializer;

public class CustomBusinessDataModule extends SimpleModule {

    private static final long serialVersionUID = 1L;
    private static final String SEPARATOR = "|";

    public CustomBusinessDataModule()
    {
        Ivy.log().info("registered custom module!");
        addSerializer(Person.class, new CustomPersonSerializer());
        addDeserializer(Person.class, new CustomPersonDeSerializer());
    }

    @SuppressWarnings("serial")
    private static class CustomPersonSerializer extends StdScalarSerializer<Person>
    {

        public CustomPersonSerializer() {
            super(Person.class);
        }

        @Override
        public void serialize(Person value, JsonGenerator gen,
                SerializerProvider provider) throws IOException {
            gen.writeString(value.getFirstname()+SEPARATOR+value.getLastname());
        }
    }

    @SuppressWarnings("serial")
    private static class CustomPersonDeSerializer extends StdScalarDeserializer<Person>
    {
        public CustomPersonDeSerializer()
        {
            super(Person.class);
        }

        @Override
        public Person deserialize(JsonParser p, DeserializationContext ctxt)
                throws IOException, JsonProcessingException {
            String value = p.readValueAs(String.class);
            if (!StringUtils.contains(value, SEPARATOR))
            {
                // old version.... -> take special care!
                return new Person();
            }

            String[] parts = StringUtils.split(value, SEPARATOR);
            Person person = new Person();
            person.setFirstname(parts[0]);
            person.setLastname(parts[1]);
            return person;
        }
    }

}

SECOND: register it via service SPI:

Here you see the custom serializer in action:

businessDataSerializer_66.iar

Comment by Reguel Wermelinger on trungdv's question

Reguel Wermelinger — Fri, 26 May 2017 05:19:31 -0400

Currently there is no API in the axon.ivy core that would trigger an index-recreation or schema update on the ElasticSearch server. But you could basically flush the whole ElasticSearch index and trigger the axon.ivy engine to synchronize its data again with the ElasticSearch index. The API to do so is here: ch.ivyteam.ivy.business.data.store.restricted.IBusinessDataManager.startIndexRecreation(). There is also a built in mechanism on axon.ivy engine startup that synchronizes the indexes if the ElasticSeach index is not up to date.

Comment by trungdv on trungdv's question

trungdv — Thu, 25 May 2017 22:01:36 -0400

i have found the root cause. it's because on new version we just changed the type of property qualificationMessageResults.message from String to an object type, so it lead to issue that when we deploy this version to server, the old business data (where message is still String type) can not be index anymore. And it throw exception unknow property descriptor.

Problematic is that ElasticSearch already determined the type of this field as string. I heard elasticsearch has way to reindex, does ivy support it?