mirror of
https://github.com/imjasonh/client-go2
synced 2026-07-08 00:55:56 +00:00
This PR adds support for resource-specific expansion methods and implements
missing standard client-go operations to achieve feature parity.
## New Features
### Expansion Methods
- Added PodClient with methods: GetLogs, Bind, Evict, ProxyGet
- Added ServiceClient with method: ProxyGet
- Both clients implement the official k8s.io/client-go expansion interfaces
- Runtime type assertions ensure type safety (panic if wrong type)
### Missing Standard Methods
- Watch: Watch resources with optional label/field selectors
- DeleteCollection: Delete multiple resources matching criteria
- UpdateStatus: Update only the status subresource
### Other Improvements
- Refactored all tests to use inline configuration pattern
- Added comprehensive e2e tests for new functionality
- Updated documentation with examples
- Fixed linting error in stream.Close() handling
## Design Decisions
The expansion methods follow a namespace-scoped pattern where calling
`client.PodClient("namespace")` returns a client that implements the
standard client-go PodExpansion interface. This maintains full API
compatibility while providing type safety through runtime assertions.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
4.1 KiB
4.1 KiB
Experimenting with k8s.io/client-go and Go generics
This is an experimental type-parameter-aware client that wraps k8s.io/client-go/rest, in about 570 lines of mostly-vibe-coded Go.
Features
- Type-safe generic client - Work with strongly-typed Kubernetes objects instead of
unstructured.Unstructured - Zero code generation - Uses Go generics instead of code generation
- Full CRUD operations - List, Get, Create, Update, Delete, Patch, Watch, DeleteCollection, and UpdateStatus support
- Informer support - Watch for changes with type-safe event handlers
- Automatic GVR inference - No need to manually specify GroupVersionResource for standard Kubernetes types
- Expansion methods - Resource-specific operations like Pod.GetLogs() and Service.ProxyGet()
- Label/Field selectors - Filter resources using Kubernetes selectors
- SubResource access - Generic method to access any subresource
Usage
See the example for comprehensive usage examples.
Quick Examples
Basic CRUD Operations
// Create a client with automatic GVR inference
client, err := generic.NewClient[*corev1.Pod](config)
// List pods with label selector
pods, err := client.List(ctx, "default", &metav1.ListOptions{
LabelSelector: "app=nginx",
})
// Get a specific pod
pod, err := client.Get(ctx, "default", "my-pod", nil)
Expansion Methods for Pods (client-go compatible)
// Start with a generic client for pods
client, err := generic.NewClient[*corev1.Pod](config)
// Get a namespace-scoped PodClient that implements typedcorev1.PodExpansion
podClient := client.PodClient("default") // Will panic if T is not *corev1.Pod
// Get pod logs (matches client-go API)
req := podClient.GetLogs("my-pod", &corev1.PodLogOptions{
Container: "nginx",
TailLines: &tailLines,
})
logs, err := req.DoRaw(ctx)
// Bind pod to node
err = podClient.Bind(ctx, binding, metav1.CreateOptions{})
// Evict pod
err = podClient.Evict(ctx, eviction)
// For cluster-scoped operations, use the generic client directly
pods, err := client.List(ctx, "default", nil)
Expansion Methods for Services (client-go compatible)
// Start with a generic client for services
client, err := generic.NewClient[*corev1.Service](config)
// Get a namespace-scoped ServiceClient that implements typedcorev1.ServiceExpansion
serviceClient := client.ServiceClient("default") // Will panic if T is not *corev1.Service
// Use proxy to access service endpoints
req := serviceClient.ProxyGet("http", "my-service", "80", "api/health", nil)
resp, err := req.DoRaw(ctx)
Generic SubResource Access
// Access any subresource using the generic method
req := client.SubResource("default", "my-pod", "status")
status, err := req.DoRaw(ctx)
Watch Resources
// Watch for pod changes with label selector
watcher, err := client.Watch(ctx, "default", &metav1.ListOptions{
LabelSelector: "app=nginx",
})
defer watcher.Stop()
for event := range watcher.ResultChan() {
pod := event.Object.(*corev1.Pod)
fmt.Printf("Event: %s Pod: %s\n", event.Type, pod.Name)
}
Delete Collection
// Delete all pods with specific label
err = client.DeleteCollection(ctx, "default", nil, &metav1.ListOptions{
LabelSelector: "app=test",
})
// Delete all pods in namespace
err = client.DeleteCollection(ctx, "default", nil, nil)
Update Status
// Update only the status subresource
pod.Status.Phase = corev1.PodRunning
pod.Status.Conditions = append(pod.Status.Conditions, corev1.PodCondition{
Type: corev1.PodReady,
Status: corev1.ConditionTrue,
})
updated, err := client.UpdateStatus(ctx, "default", pod, nil)
Testing
Unit tests against a mock REST client:
make test
End-to-end tests against a real K8s cluster:
make e2e
⚠️ THIS IS AN EXPERIMENT
This is just for demo purposes.
The name client-go2 is a placeholder, and a joke.