I am having an issue with aggregation framework in one particular aspect - 
its not handling _missing values and _other values as they were handled in 
facets (_other was not handled even in any stats facets just in terms)
traditional OLAP would slice/roll-up the same data set by various 
dimensions so roll up counts/totals would come out the same even if we are 
grouping on a field which may be null for some records

With aggregation framework there is no way to do it except for an 
exceedingly convoluted use of "missing" aggregation (just try to have 
missing values aggregated as part of the overall bucket-set when doing 
multi-level aggregation)
you can find lot more details, example and my proposal here 
https://github.com/elasticsearch/elasticsearch/issues/5324 

Unfortunately it did not get any reply from the development team so I can 
only assume they are not convinced or did not read it (would be nice if 
they at least said so may be they missed it all together)

I do not want to replace null values with some fake values representing 
null and have it bleed all over the applications consuming JSON data from 
elastic. So if Elastic is not going to handle missing, what are my options 
(apart from the option described in my proposal on github)?

Could I use "null_value" in my index mapping for nullable fields? Will they 
be used for aggregation rather than _source. Even if they are will it work 
when null value  is one of objects and I am aggregating on that object 
properties?
(i.e. case type is {caseNumber:123, caseType:{id:10, name:'Civil'}} and I 
am aggregating on caseType.id and caseType could be null)



-- 
You received this message because you are subscribed to the Google Groups 
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/elasticsearch/6ad26233-59f4-4464-a380-2832685e528b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to