oom on FastTaxonomyFacetsCounts

[Sheng <sh...@gmail.com>]

This is probably not the fault of Lucene, as oom happened on the loc :

values = new int[taxoReader.getSize()];

So taxoReader.getSize() probably is too big. My question is is there a more
memory friendly way (also without significant performance penality) to get
FacetResult for a particular dimension from the facets cache to replace the
call to FastTaxonomyCounts.getTopChildren(...)? Otherwise I have to
increase the available heap memory to the process, which I'd consider as
last resort.

Thanks,