Perf: Names with acronyms, eg "C Brown" are very slow #14933
Description
Queries without a full first name, eg "C Brown" vs "Christopher Brown", are very slow. These two benchmarks are very dependent on the caches state. If I rerun them, then I get the result back immediately (nice!). So for accurate tests, make sure you wait ~15 minutes between executions. Or if you have other ways to skip the cache so for easier benchmarking, then I'm interested to hear that as well.
Christopher Brown
- requests take 4.6sec with a fresh cache
- Web UI URL: https://www.fec.gov/data/receipts/?data_type=processed&contributor_name=Christopher+brown&contributor_city=LONDON&contributor_state=ZZ
- cURL:
time curl 'https://api.open.fec.gov/v1/schedules/schedule_a/?api_key=U4CGxchr88wIKcoCqfL8mxExQearu8iTo6S9f0wU&contributor_name=Christopher+Brown&contributor_city=LONDON&contributor_state=ZZ'
C Brown
- request take 2.6 minutes with a fresh cache
- Web UI URL: https://www.fec.gov/data/receipts/?data_type=processed&contributor_name=C+brown&contributor_city=LONDON&contributor_state=ZZ
- cURL:
time curl 'https://api.open.fec.gov/v1/schedules/schedule_a/?api_key=U4CGxchr88wIKcoCqfL8mxExQearu8iTo6S9f0wU&contributor_name=C+Brown&contributor_city=LONDON&contributor_state=ZZ'
I'm wondering:
- can we improve this? Eg by adding an index on just the first letter of the first name? I think you are using postgres, but not sure. This seems like a pretty common pattern that would be good to optimize
- If we can't improve it, can we at least pin down the exact conditions that cause this so I know what sorts of queries to avoid? eg is it just when there are any single letter tokens in the name?