Closed
Description
With the eventual removal of multiple types in one index, the handling of parent-child type data has changed a bit in Elasticsearch. There is a new join
mapping type that allows users to model their data in a parent-child relationship without the use of different types. A few factors that ES-Hadoop must consider for this:
- We must continue to support "parent" headings for legacy Elasticsearch versions.
- Users should be able to specify that a field is considered a join field, and the connector should extract the routing information from the join field.
- If a user specifies both a join field and a routing field, the routing field should be used if it can be found, and if not, the extraction should fall back to extracting from the join field.
- When reading data, we must accurately parse both forms of the join field (String for parents / Object for children). This is doubly sensitive for the schema based integrations.