Github user dsmiley commented on the issue:
https://github.com/apache/lucene-solr/pull/416
Before I look further I want to mention 2 things:
* There needn't be separation between the non-nested and nested algorithm.
We could simply use the nested algorithm but special-case when the
SortedDocValues we get is null due to non-existent path. In that event, all
docs get added to the root, anonymously (unlabelled). I suspect doing it this
way is less code and will be sufficiently clear but we'll see?
* My pseudocode had a TODO about double-resolving the path in order to get
the label at the time the docs are added to the parent. If we make the pending
data structure a Multimap<Integer,Multimap<String,SolrDocument>> then at the
time we read the document and get the path we could store the relationship at
this time. Or perhaps you have a better idea. I thought of putting the label
into a temporary field of the SolrDocument; though I'm not sure I like that any
better; maybe less.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
