Hello, I currently have a demand for faceting on the children of a join query.
My index is set up in a way that there are parent and child documents. The child documents do have the facet information in a (precisely: some) multivalue field(s). The parent documents themselves do not have any of it. As the join query support allows me to do a simple search within the child documents and return documents from the parent document space I thought there probably is a way to figure out the available facet values from the child document space and present both in the result set, but this seems more difficult than I thought it would be. The join query support would allow me to filter on specific child-document-space facet fields, for example: but I can not really find a way to present *which faceting options are available* in the result set in first place. Denormalizing my index in a way that the parent documents would contain the faceting information is not an option at the moment, because I wanted to keep the index more generic, so that there's not one field per attribute but two generic attribute fields (multi-value), that keep the Key/Value pairs, like the following table shows. I need this setup because at index setup time I do not know which attributes for the various products/items will be available. If I now would denormalize a bunch of shoe child items into the parent product it would always contain all possible size/color combinations, even if some of the child products do not meet the initial search term's criteria, e.g. searching above for (title:Sneakers AND desc:cool) should return just facets for "size" (2), "color" (2), "red" (1), "blue" (1), "40" (1) and "42" (1), which I do postprocess in my client application, so that I know that "red" and "blue" are "color"s and "40" and "42" are "size"s. I thought that you cracks might have an idea on how to continue from there. Best, Tobias -- View this message in context: http://lucene.472066.n3.nabble.com/Join-and-faceting-by-children-s-attributes-tp3512629p3512629.html Sent from the Solr - User mailing list archive at Nabble.com.