In the documentation it says that ALL parent IDs must be resident in
memory, the question is why.
Why can't the has_child run the query per shard, load into memory all of
the parent IDs that returned from the query and then use these to filter
the parent docs.
For some cases, the recall on the children would be much lower than the
parents, hence, no need for caching the parent-child relation at all.
Has this been tried? What are/were the considerations for implementing this
as it is?
In the documentation it says that ALL parent IDs must be resident in memory, the question is why.
Why can't the has_child run the query per shard, load into memory all of the parent IDs that returned from the query and then use these to filter the parent docs.
For some cases, the recall on the children would be much lower than the parents, hence, no need for caching the parent-child relation at all.
Has this been tried? What are/were the considerations for implementing this as it is?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.