test: refactor to allow benchmark Vote (WIP, MVP) #1028
Conversation
altergui
changed the title
| qt.Assert(t, v2.VoteID.String(), qt.Equals, voter.vote.VoteID.String()) | ||
| qt.Assert(t, v2.BlockHeight, qt.Equals, uint32(2)) | ||
| qt.Assert(t, v2.VoterID.String(), qt.Equals, fmt.Sprintf("%x", voter.key.Address().Bytes())) | ||
| } | ||
| } | ||
|
|
||
| func TestAPIaccount(t *testing.T) { |
There was a problem hiding this comment.
there's a lot of code in TestAPIaccount that could be deduplicated now, after the proposed refactoring. i'd include this in the same PR
|
how does it look so far? am i on the right track? |
| @@ -0,0 +1,26 @@ | |||
| package test | |||
There was a problem hiding this comment.
@mvdan can you check the basics of this benchmark and maybe propose improvements? The idea is to have a full benchmark to test performance of state+app+indexer+api
There was a problem hiding this comment.
I think the logic looks okay. Two suggestions come to mind:
-
What kind of election is this - encrypted or not? We should probably benchmark the one that is most common. If both are common, then I would suggest that the benchmark do both side by side. We could similarly think about other kinds of votes, like overwrites.
-
You use a sequential benchmark via
b.N, which is fine, but not very realistic - in reality, people won't vote in a big election in a big line :) For that reason, I would use a parallel benchmark via https://pkg.go.dev/testing#B.RunParallel. You can still useb.Nas the total number of iterations to set up the census and anything else you need, but the looping should use https://pkg.go.dev/testing#PB.Next instead.
There was a problem hiding this comment.
Also, should the benchmark advance votes while iterating? Because again, in reality, you won't have an entire election get all of its votes within a single block. Plus, the indexer does a significant amount of its work in the Commit method, so you should advance at least one block while the benchmark timer is on.
altergui
force-pushed
the
f/apiv2_benchmarks
branch
from
6d00303
to
a841409
Compare
altergui
force-pushed
the
f/apiv2_benchmarks
branch
from
a841409
to
bb0b5e0
Compare
it's failing with: ERR vochain/app.go:287 > checkTx error="voteTx: maxCensusSize reached 0/0" app_benchmark_test.go:131: checkTX failed: checkTx voteTx: maxCensusSize reached 0/0
altergui
force-pushed
the
f/apiv2_benchmarks
branch
from
39f608a
to
a60c948
Compare
it's failing with:
--- FAIL: BenchmarkFetchTx
bench_test.go:157:
error:
got non-nil error
got:
e"transaction not found"
stack:
/home/runner/work/vocdoni-node/vocdoni-node/vochain/indexer/bench_test.go:157
qt.Assert(b, err, qt.IsNil)
altergui
force-pushed
the
f/apiv2_benchmarks
branch
from
a60c948
to
c17ff68
Compare
started working on #965 and opted for a refactor that will allow me reuse most of the code