[Core] Add serialization / deserialization to KnowledgeDatabase

[yuripourre]

My suggestion is to use JSON so humans can also populate the database.

Now that the id is a String we do not should care about updating indexes to keep everything sorted.

Obviously use an external serializer and a helper class. Do not pollute pojos with annotations.

General approach (words dictionary may be different) A concept with id = 1 and name.id = nameId1.

"knowledge base": {
  "typeId0": {"name":"type name word"},
  "nameId1": {"name": "name word", "typeId":"typeId0"},
  "1": {"nameId":"nameId1", "typeId":"typeId1", "customAttr1Id":"attrId1", "customAttr2Id":"attrId2"}
}

Persist as depth first search and all words are leaves.

Ideally, the best order would be;

• Save all types
• Save all names
• Save all concepts

[yuripourre]

Remember to only persist the first level (forget about nested attributes). And all (children) attributes should be persisted as Id only.

Knowledge Base should index all concepts using Id.

Suggestion: words, types, concepts. They can be persisted in knowledge base separately.

[yuripourre]

We can try this one: http://jsoniter.com/

Json libraries comparison: https://github.com/fabienrenaud/java-json-benchmark

[yuripourre]

Do not forget to add a test to serialize and deserialize the same knowledge base.

[yuripourre]

Also, add tests to handling databases of words since it will be a common use case.