Top-N about featurebase HOT 5 CLOSED

The parser_test test for top-n is not correct.

top-n(frame=x, n=2)

it should be

top-n(all(),frame="b.n", n=2) for the ranked unfiltered list
or
top-n(get(id=10, frame="b.n"), frame="b.n",n=50,filter=[1,2,3]) ==> top-n for bitmaps with common profiles as id=10 in bitmaps that belong to the category 1 or 2 or 3
or
top-n(get(id=10, frame="b.n"), frame="b.n",n=50 )

where the filter can be any of the category id (just a list of id int keys) of the list.

Does that make sense?

from featurebase.

Yep, that makes sense. Thanks, Todd!

from featurebase.

TopN is looking fabulous. One thing that seems to be missing is persistence of the "cache". It appears from my testing that the bitmaps must first be fetched in order to appear in the top list. My hope was that this could be avoided. Have any ideas on how that might work?

from featurebase.

@tgruben Yeah, I can put the cache persistence back in there. Sorry for chopping that. :)

Since the bitmap retrieval should be pretty fast can I just persist the bitmap IDs periodically?

from featurebase.

Yep.

from featurebase.