Hazenix opened a new pull request, #3842:
URL: https://github.com/apache/dubbo-go/pull/3842
Refs #3728 (Triple non-IDL codec reuse).
## Summary
The built-in Triple MsgPack codec created a default `MsgpackHandle` for
every encode and decode. Reusing the handle keeps the type cache available
across calls.
```diff
- Encode or decode with a fresh default MsgpackHandle
+ Encode or decode with a shared default MsgpackHandle
Keep a separate Encoder or Decoder for each call
```
## Evidence
Ran `BenchmarkWrapperCodecFastPath` and `BenchmarkWrapperCodecSlowPath` on
Apple M4 Pro with Go 1.25.0 and one benchmark CPU. The measurements are medians
from six alternating before/after runs. All 16 MsgPack cases saved four
allocations per operation. Across the 128 B and 1 KiB cases, median time fell
by 7.0%–19.5%.
| FastPath case | ns/op | B/op | allocs/op |
|---|---:|---:|---:|
| 128 B Triple | 511 → 416 | 1,648 → 1,392 | 10 → 6 |
| 128 B Envelope | 520 → 425 | 1,653 → 1,397 | 11 → 7 |
| 1 KiB Triple | 760 → 668 | 3,184 → 2,928 | 11 → 7 |
| 1 KiB Envelope | 765 → 681 | 3,189 → 2,933 | 12 → 8 |
In a ten-pair follow-up, 16 KiB Envelope SlowPath was 9.1% faster at the
median. The 1 MiB Envelope FastPath median moved by +0.6%, with five faster and
five slower pairs. These benchmarks measure wrapper encoding and framing on the
send path.
- The new regression compares request and response wrapper bytes against a
fresh-handle encoder and decodes the old bytes on both sides. Sixteen parallel
round trips passed twenty repeated runs under `-race`.
- `go test -race ./protocol/triple/... -count=1` and the external codec
plugin RPC regression tests passed.
## Merge Danger
Reversible with a single revert. The change is scoped to the built-in Triple
MsgPack codec. Its default options and per-call encoder and decoder ownership
are preserved.
### Checklist
- [x] I confirm the target branch is `develop`
- [ ] I have run `make fmt` to format my code
- [ ] I have run `make test` to run local tests
- [x] I have added tests that prove my fix is effective or that my feature
works
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]