Join GitHub today
GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.Sign up
default search is divided into exact search and wildcard search #3073
@davidoesch @gjn @danduk82
triggered by geoadmin/mf-geoadmin3#4287
The results of the exact match query (one per origin) will be prepended to the results of the wildcard query.
The results of the exact match search will be on top of the list now.
a) search1 (exact) results on top of search2 (wildcard) results current implementation in this pr
b) search1 (exact) results on top of each search2 (wildcard) results origin
We will also have to do some yandex load tests an monitor the impact on sphinx of this change.
the new multi-query requests are two times slower than the old requests (85ms -> 160ms)
the exact search query already returns a maximum of one object per rank/origin, thus 7 objects (zipcode,gg25,district,kantone,gazetteer,address,parcel)
from 1st run from 5 rps / 60s test
Feb 19, 2019
1 check passed
Almost 2 weeks live with this now (released on 22nd of February). And here's the reality:
Almost no detectable/significant change in average, 50th percentile and 90th percentile response times on the location search. I think that the 2 requests are executed in parallel and sphinx can handle it quite well.
I guess the change could lead to a lower ceiling (number of total concurrent requests) we can handle, so a spike will lead to sphinx to break earlier, but standard traffic, it handles equally well.
We did some load tests with yandex before putting this change to production. the max number of requests/sec stays the same :