Optimize lazy context tracking and descendant traversal performance

[andreyyudin]

Summary

• Added config.WithLazyContextTracking() for on-demand context-variable tracking.
• Preserved backward compatibility by keeping eager tracking as the default.
• Optimized descendant traversal using iterative descendApply to reduce allocations.
• Added reproducible benchmarks for descendant traversal and lazy tracking.

Benchmark Prerequisites

Install benchstat once:

go install golang.org/x/perf/cmd/benchstat@latest

If benchstat is not on your PATH, invoke it as:

$(go env GOPATH)/bin/benchstat ...

How To Benchmark (Reproducible)

1) Descendant traversal (descendApply) before/after (n=15)

# feature branch
git checkout feature/lazy-context-tracking
go test ./pkg/jsonpath -bench BenchmarkDescendantQuery -benchmem -run '^$' -benchtime=5s -count=15 > /tmp/feat_desc.txt

# main baseline (reuse benchmark file from feature branch)
git checkout main
git checkout feature/lazy-context-tracking -- pkg/jsonpath/yaml_query_test.go
go test ./pkg/jsonpath -bench BenchmarkDescendantQuery -benchmem -run '^$' -benchtime=5s -count=15 > /tmp/main_desc.txt
git restore --source=HEAD --staged --worktree pkg/jsonpath/yaml_query_test.go

# compare
$(go env GOPATH)/bin/benchstat /tmp/main_desc.txt /tmp/feat_desc.txt

2) Lazy tracking on/off (n=15)

git checkout feature/lazy-context-tracking

LAZY_TRACKING_MODE=eager go test ./pkg/jsonpath \
  -bench BenchmarkLazyContextTrackingStress -benchmem -run '^$' \
  -benchtime=1x -count=15 -tags 'lazytracking stress' > /tmp/lazy_eager.txt

LAZY_TRACKING_MODE=lazy go test ./pkg/jsonpath \
  -bench BenchmarkLazyContextTrackingStress -benchmem -run '^$' \
  -benchtime=1x -count=15 -tags 'lazytracking stress' > /tmp/lazy_lazy.txt

$(go env GOPATH)/bin/benchstat /tmp/lazy_eager.txt /tmp/lazy_lazy.txt

Performance

Descendant traversal (BenchmarkDescendantQuery, n=15)

• time: -20.69% (8.723ms -> 6.918ms, p=0.000)
• bytes: -38.55% (6.128MiB -> 3.766MiB, p=0.000)
• allocs: -49.90% (50.38k -> 25.24k, p=0.000)

Lazy tracking stress (BenchmarkLazyContextTrackingStress, n=15)

• time: -75.18% (433.2ms -> 107.5ms, p=0.000)
• bytes: -74.39% (158.62MiB -> 40.63MiB, p=0.000)
• allocs: -20.30% (1.658M -> 1.321M, p=0.000)