chore: avoid unnecessary deep copy

[crenshaw-dev]

I'm not convinced we need to deep copy these objects at all. But to be cautious, I think we can save some resources by only deep copying if we need to mutate the objects.

Here's the benchmark result:

before: Benchmark_threeWayMergePatch-16            23341             52004 ns/op           39070 B/op        810 allocs/op
after:  Benchmark_threeWayMergePatch-16            24838             48706 ns/op           37709 B/op        800 allocs/op

We save a little memory and CPU.

Here's the benchmark code:

func Benchmark_threeWayMergePatch(b *testing.B) {
	orig := []byte(`
apiVersion: v1
kind: Service
metadata:
  name: my-service
  namespace: default
`)
	config := []byte(`
apiVersion: v1
kind: Service
metadata:
  name: my-service
  namespace: default
`)
	live := []byte(`
apiVersion: v1
kind: Service
metadata:
  name: my-service
  namespace: default
`)

	origUnstructured := unstructured.Unstructured{}
	err := yaml.Unmarshal(orig, &origUnstructured)
	require.NoError(b, err)
	configUnstructured := unstructured.Unstructured{}
	err = yaml.Unmarshal(config, &configUnstructured)
	require.NoError(b, err)
	liveUnstructured := unstructured.Unstructured{}
	err = yaml.Unmarshal(live, &liveUnstructured)
	require.NoError(b, err)

	b.ResetTimer()

	for i := 0; i < b.N; i++ {
		_, err := ThreeWayDiff(&origUnstructured, &configUnstructured, &liveUnstructured)
		require.NoError(b, err)
	}
}

[sonarcloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarCloud

[crenshaw-dev]

Fixing the test failure makes the change more costly than the original. Leaving it as-is for now.