perf(error): cut intermediate allocations in LogValue, ToMap and formatVerbose

[samber]

Summary

• LogValue: build the []any slog.Group argument lists directly instead of lo.MapToSlice followed by lo.ToAnySlice, which allocated two intermediate slices per group (context, user, tenant).
• ToMap: use the snapshot-owned user/tenant maps directly instead of copying them with lo.Assign - snapshot() always returns freshly merged maps, so mutating them is safe.
• formatVerbose: replace the Split -> lo.Map(prefix) -> Join pipeline for request/response dumps with a single-pass writer into the existing strings.Builder (writePrefixedLines), preserving output byte-for-byte.
• writePrefixedLines: take its input as []byte instead of string. Both callers already hold the httputil dump as []byte and were converting it to string just for this call; accepting []byte directly removes that copy.

Benchmarks

benchstat vs main (darwin/arm64, Apple M3, -count=10 -cpu=1):

            │ before2.bench.txt │          after2.bench.txt          │
            │      sec/op       │   sec/op     vs base               │
ToMap               3.843µ ± 4%   3.812µ ± 6%       ~ (p=0.393 n=10)
LogValue            3.803µ ± 4%   3.857µ ± 3%       ~ (p=0.353 n=10)
MarshalJSON         5.852µ ± 5%   5.754µ ± 8%       ~ (p=0.353 n=10)
geomean             4.406µ        4.390µ       -0.37%

            │ before2.bench.txt │          after2.bench.txt           │
            │       B/op        │     B/op      vs base               │
ToMap              5.901Ki ± 0%   5.573Ki ± 0%  -5.56% (p=0.000 n=10)
LogValue           5.977Ki ± 0%   5.758Ki ± 0%  -3.66% (p=0.000 n=10)
MarshalJSON        6.665Ki ± 0%   6.665Ki ± 0%       ~ (p=1.000 n=10)
geomean            6.172Ki        5.980Ki       -3.10%

            │ before2.bench.txt │          after2.bench.txt           │
            │     allocs/op     │ allocs/op   vs base                 │
ToMap                72.00 ± 0%   70.00 ± 0%  -2.78% (p=0.000 n=10)
LogValue             70.00 ± 0%   66.00 ± 0%  -5.71% (p=0.000 n=10)
MarshalJSON          92.00 ± 0%   92.00 ± 0%       ~ (p=1.000 n=10)
geomean              77.40        75.19       -2.86%

Note: wall-clock deltas were not statistically significant on this run (background noise on the benchmark machine), but the B/op and allocs/op reductions are fully reproducible (0% variance, p<0.001). MarshalJSON's benchmark fixture does not set User/Tenant data, so it does not exercise the ToMap fast path.

For the writePrefixedLines []byte change specifically, benchstat vs the previous string-based version (same machine, -count=10 -cpu=1, ~100-byte dump of typical request headers), measured with a throwaway internal benchmark not included in this PR (existing benchmarks live in the external benchmarks/ package and writePrefixedLines is unexported):

                            │ wpl_before.txt │            wpl_after.txt            │
                            │     sec/op     │   sec/op     vs base                │
TempWritePrefixedLinesSmall      149.7n ± 2%   134.5n ± 4%  -10.15% (p=0.001 n=10)

                            │ wpl_before.txt │           wpl_after.txt            │
                            │      B/op      │    B/op     vs base                │
TempWritePrefixedLinesSmall       344.0 ± 0%   232.0 ± 0%  -32.56% (p=0.000 n=10)

                            │ wpl_before.txt │           wpl_after.txt            │
                            │   allocs/op    │ allocs/op   vs base                │
TempWritePrefixedLinesSmall       5.000 ± 0%   4.000 ± 0%  -20.00% (p=0.000 n=10)

Test plan

Verified with go build ./..., go vet ./..., go test ./..., and the benchmark runs above compared with benchstat.

[codecov]

Codecov Report

❌ Patch coverage is 93.10345% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.43%. Comparing base (470a3a2) to head (bf516c4).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
error.go	93.10%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #135      +/-   ##
==========================================
+ Coverage   90.17%   90.43%   +0.26%     
==========================================
  Files          15       15              
  Lines        1231     1234       +3     
==========================================
+ Hits         1110     1116       +6     
+ Misses         96       93       -3     
  Partials       25       25              

Flag	Coverage Δ
unittests	90.43% <93.10%> (+0.26%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR focuses on reducing allocations in OopsError’s logging/serialization and verbose formatting paths, primarily by avoiding intermediate slices/maps and by streaming line-prefix formatting directly into an existing strings.Builder.

Changes:

• LogValue: constructs slog.Group argument lists directly to avoid lo.MapToSlice + lo.ToAnySlice intermediate allocations.
• ToMap: reuses snapshot-owned user/tenant maps directly instead of copying them.
• formatVerbose: replaces Split/Map/Join request/response dump formatting with a single-pass helper writePrefixedLines.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.