yes, it's not printing errors but the speed of that is because it's actually skipping a shit-ton of things
i'm gonna just remove it
so, it took all morning, but i changed one of those stupid string fields to bytes now, the ID
cpu: AMD Ryzen 5 PRO 4650G with Radeon Graphics
BenchmarkBinaryEncoding/event2.MarshalJSON-12 14 73695929 ns/op
BenchmarkBinaryEncoding/event2.EventToBinary-12 157 7280781 ns/op
BenchmarkBinaryEncoding/easyjson.Marshal-12 64 19049795 ns/op
BenchmarkBinaryEncoding/gob.Encode-12 18 62296062 ns/op
BenchmarkBinaryEncoding/binary.Marshal-12 10 110174020 ns/op
BenchmarkBinaryDecoding/event2.BinaryToEvent-12 100 13698489 ns/op
BenchmarkBinaryDecoding/easyjson.Unmarshal-12 56 19677698 ns/op
BenchmarkBinaryDecoding/gob.Decode-12 5 226144867 ns/op
BenchmarkBinaryDecoding/binary.Unmarshal-12 122 9011042 ns/op
BenchmarkBinaryDecoding/binary.UnmarshalBinary-12 277 4849297 ns/op
BenchmarkBinaryDecoding/easyjson.Unmarshal+sig-12 1 1624027475 ns/op
BenchmarkBinaryDecoding/binary.Unmarshal+sig-12 1 1596453002 ns/op
just one of the fields, and from 139 to 157 - makes sense because it's not decoding hex anymore
i'm surprised the decode didn't get faster, since it now literally only makes a new slice header (subslicing)
small result for a lot of work but the pubkey and signature are still to come
nostr:npub180cvv07tjdrrgpa0j7j7tmnyl2yr6yr7l8j4s3evf6u64th6gkwsyjh6w6 btw the UnmarshalBinary function you wrote is extremely fast, but the reason why its counterpart MarshalBinary isn't in there is because it can't deal with events over 64kb in size, otherwise i'd re-enable it
Discussion
cpu: AMD Ryzen 5 PRO 4650G with Radeon Graphics
BenchmarkBinaryEncoding/event2.MarshalJSON-12 15 74245789 ns/op
BenchmarkBinaryEncoding/event2.EventToBinary-12 162 7429203 ns/op
BenchmarkBinaryEncoding/easyjson.Marshal-12 60 21221235 ns/op
BenchmarkBinaryEncoding/gob.Encode-12 18 62426577 ns/op
BenchmarkBinaryEncoding/binary.Marshal-12 9 112783137 ns/op
BenchmarkBinaryDecoding/event2.UnmarshalJSON-12 14 90464147 ns/op
BenchmarkBinaryDecoding/event2.BinaryToEvent-12 100 12832888 ns/op
BenchmarkBinaryDecoding/easyjson.Unmarshal-12 57 23280946 ns/op
BenchmarkBinaryDecoding/gob.Decode-12 5 226555916 ns/op
BenchmarkBinaryDecoding/binary.Unmarshal-12 128 8563056 ns/op
the output from https://mleku.dev/nostrbench
now with the pubkey encoding as binary... not as dramatic an improvement it seems
oh well, anyway, it's done now, both id and pubkey fields now do not need any hex encode/decode so that's still a good thing
last is the signature... this will be another 5-10 ops more i figure, then i'm gonna get out the profiler
goos: linux
goarch: amd64
pkg: mleku.net/nostrbench
cpu: AMD Ryzen 5 PRO 4650G with Radeon Graphics
BenchmarkBinaryEncoding/event2.MarshalJSON-12 14 72883244 ns/op
BenchmarkBinaryEncoding/event2.EventToBinary-12 166 7090886 ns/op
BenchmarkBinaryEncoding/easyjson.Marshal-12 67 18620409 ns/op
BenchmarkBinaryEncoding/gob.Encode-12 19 62998941 ns/op
BenchmarkBinaryEncoding/binary.Marshal-12 10 110464204 ns/op
BenchmarkBinaryDecoding/event2.UnmarshalJSON-12 13 87762713 ns/op
BenchmarkBinaryDecoding/event2.BinaryToEvent-12 100 12388974 ns/op
BenchmarkBinaryDecoding/easyjson.Unmarshal-12 57 23463646 ns/op
BenchmarkBinaryDecoding/gob.Decode-12 5 226411371 ns/op
BenchmarkBinaryDecoding/binary.Unmarshal-12 123 8851251 ns/op
#devstr #benchmark #nostr
since switching up the binary encoder for sure the relay is running way way faster... also because i fixed the shit slow logging library, that was a big part of it, none of these other libraries even have logging! if mine has an error it prints a log! so it's still faster and dev friendly at the same time
bah humbug of course i has bugs now, very weirdness
it looks like i broke the filters, that would be why it returns zero counts all the time and zero results get sent, and then idk why not sending eose either, but hey gotta expect gotchas like this... i'll fix em tomorrow most likely, and become aware of how they got broken so when i do the signatures i don't get surprised like this
it does send out the messages to subscribers when i post it but filter searches are b0rked
ok, bug found... funny enough the first change i made was bugged... the event IDs... was all wrong totally... only took a few minutes to get it generating the correct output and plugging that into the filters and suddenly all these results are coming back from the filters coming in from clients
also, the bug was slowing down the encoder... this is now typical results
goos: linux
goarch: amd64
pkg: mleku.net/nostrbench
cpu: AMD Ryzen 5 PRO 4650G with Radeon Graphics
BenchmarkBinaryEncoding/event2.MarshalJSON-12 14 74456910 ns/op
BenchmarkBinaryEncoding/event2.EventToBinary-12 172 6911613 ns/op
BenchmarkBinaryEncoding/easyjson.Marshal-12 63 18776462 ns/op
BenchmarkBinaryEncoding/gob.Encode-12 19 62966594 ns/op
BenchmarkBinaryEncoding/binary.Marshal-12 9 118801192 ns/op
BenchmarkBinaryDecoding/event2.UnmarshalJSON-12 14 88550098 ns/op
BenchmarkBinaryDecoding/event2.BinaryToEvent-12 100 12272410 ns/op
BenchmarkBinaryDecoding/easyjson.Unmarshal-12 58 23138193 ns/op
BenchmarkBinaryDecoding/gob.Decode-12 5 227175445 ns/op
BenchmarkBinaryDecoding/binary.Unmarshal-12 126 8782942 ns/op
i'm still puzzled why the BinaryToEvent is unaffected though... gonna get to it yet, for now, all is well with the world, my relay is actually finding events and sending them back
actually, no, i only fixed one case, looks like the pubkeys are broken