Topn fill about featurebase HOT 4 CLOSED

@tgruben Yep, I can take a look.

from featurebase.

@tgruben Did the previous implementation do a refetch? From what I see in 6822dd9 it does a TopN() to fetch N*2 results from the nodes trims that down to N results.

This is the only change required for N*2: master...benbjohnson:top-n-2

I could see doing a min N of 10 or so just to make sure that small Ns always fetch a large enough resultset.

Here are the lines from the previous implementation I was referencing:

https://github.com/umbel/pilosa/blob/6822dd99b5245195dbe45fbe9cb1718fdfb51a2f/core/topn.go#L174
https://github.com/umbel/pilosa/blob/6822dd99b5245195dbe45fbe9cb1718fdfb51a2f/core/topn.go#L190

from featurebase.

Yes,but most of the code that does the fill in was located in
core/query.go if you look at TopNPackage and check_pair around line 252

it is basically the results of the first scatter
That TopN package gets the first batch and then the check pair sends out
the requests for the fill in, so it was a little more complicated that just
doubling the request

Does that code make sense or do I need to expand..that one function is
basically the "combine results from all queries part.
(CatQueryStepHandler)

-Todd

On Mon, Feb 29, 2016 at 5:30 PM, Ben Johnson [email protected]
wrote:

@tgruben https://github.com/tgruben Did the previous implementation do
a refetch? From what I see in 6822dd9
6822dd9
it does a TopN() to fetch N*2 results from the nodes trims that down to N
results.

This is the only change required for N*2: master...benbjohnson:top-n-2
master...benbjohnson:top-n-2

I could see doing a min N of 10 or so just to make sure that small Ns
always fetch a large enough resultset.

Here are the lines from the previous implementation I was referencing:

https://github.com/umbel/pilosa/blob/6822dd99b5245195dbe45fbe9cb1718fdfb51a2f/core/topn.go#L174

https://github.com/umbel/pilosa/blob/6822dd99b5245195dbe45fbe9cb1718fdfb51a2f/core/topn.go#L190

—
Reply to this email directly or view it on GitHub
#57 (comment).

from featurebase.

@tgruben Yep, that make sense. I missed that refetching part the first time around. I'll get that added in.

from featurebase.