Last revive fixes

See: https://github.com/mgechev/revive/blob/master/RULES_DESCRIPTIONS.md#exported

As an aside we really should make sure we always return exported types.
UNexported are really terrible to work with: you can not switch
type/type assert them - this is good for things like if we are very anal
about interface breakage, but in the grand scheme of things it makes
dealing with some code pretty painful at times.

Signed-off-by: Milos Gajdos <[email protected]>

Author: milosgajdos

.golangci.yml

@@ -12,7 +12,7 @@ linters:
     - bodyclose
     - tparallel
     - errcheck
-    #- revive
+    - revive
 
 issues:
   exlude-dirs:

api/pkg/scheduler/allocator.go

@@ -29,24 +29,26 @@ type WorkloadAllocator interface {
 // TimeoutFunc defines a function type that determines if a runner has timed out based on the last activity.
 type TimeoutFunc func(runnerID string, lastActivityTime time.Time) bool
 
-// allocator implements the WorkloadAllocator interface, managing runners, slots, and workload allocation.
-type allocator struct {
+// Allocator implements the WorkloadAllocator interface, managing runners, slots, and workload allocation.
+type Allocator struct {
 	slots           *xsync.MapOf[uuid.UUID, *Slot] // Maps slot ID to Slot details.
 	modelStaleFunc  TimeoutFunc                    // Function to check if models are stale
 	slotTimeoutFunc TimeoutFunc                    // Function to check if slots have timed out due to error
 }
 
+var _ WorkloadAllocator = &Allocator{}
+
 // NewWorkloadAllocator creates a new allocator instance with timeout functions for models and runners.
-func NewWorkloadAllocator(staleFunc TimeoutFunc, slotTimeoutFunc TimeoutFunc) *allocator {
-	return &allocator{
+func NewWorkloadAllocator(staleFunc TimeoutFunc, slotTimeoutFunc TimeoutFunc) *Allocator {
+	return &Allocator{
 		slots:           xsync.NewMapOf[uuid.UUID, *Slot](),
 		modelStaleFunc:  staleFunc,
 		slotTimeoutFunc: slotTimeoutFunc,
 	}
 }
 
 // AllocateSlot assigns a workload to a specific slot, validating the model and slot before scheduling.
-func (a *allocator) AllocateSlot(slotID uuid.UUID, req *Workload) error {
+func (a *Allocator) AllocateSlot(slotID uuid.UUID, req *Workload) error {
 	// Validate model
 	if _, err := model.GetModel(req.ModelName().String()); err != nil {
 		return fmt.Errorf("unable to get model (%s): %v", req.ModelName(), err)
@@ -81,7 +83,7 @@ func (a *allocator) AllocateSlot(slotID uuid.UUID, req *Workload) error {
 }
 
 // AllocateNewSlot creates a new slot for a workload and allocates it to the best available runner.
-func (a *allocator) AllocateNewSlot(runnerID string, req *Workload) (*Slot, error) {
+func (a *Allocator) AllocateNewSlot(runnerID string, req *Workload) (*Slot, error) {
 	// Create a new slot and schedule the workload.
 	slot := NewSlot(runnerID, req, a.modelStaleFunc, a.slotTimeoutFunc)
 	log.Trace().
@@ -100,7 +102,7 @@ func (a *allocator) AllocateNewSlot(runnerID string, req *Workload) (*Slot, erro
 }
 
 // ReleaseSlot frees the resources allocated to a specific slot.
-func (a *allocator) ReleaseSlot(slotID uuid.UUID) error {
+func (a *Allocator) ReleaseSlot(slotID uuid.UUID) error {
 	// Find the slot.
 	slot, ok := a.slots.Load(slotID)
 	if !ok {
@@ -121,7 +123,7 @@ func (a *allocator) ReleaseSlot(slotID uuid.UUID) error {
 }
 
 // ReconcileSlots updates the state of a runner and reconciles its slots with the allocator's records.
-func (a *allocator) ReconcileSlots(props *types.RunnerState) error {
+func (a *Allocator) ReconcileSlots(props *types.RunnerState) error {
 	// Log runner state update.
 	l := log.With().
 		Str("runner_id", props.ID).
@@ -216,7 +218,7 @@ func (a *allocator) ReconcileSlots(props *types.RunnerState) error {
 }
 
 // WarmSlots returns a list of available slots with warm models waiting for work.
-func (a *allocator) WarmSlots(req *Workload) []*Slot {
+func (a *Allocator) WarmSlots(req *Workload) []*Slot {
 	cosyWarm := make([]*Slot, 0, a.slots.Size())
 
 	a.slots.Range(func(id uuid.UUID, slot *Slot) bool {
@@ -268,7 +270,7 @@ func (a *allocator) WarmSlots(req *Workload) []*Slot {
 }
 
 // RunnerSlots returns all slots associated with a specific runner ID.
-func (a *allocator) RunnerSlots(id string) []*Slot {
+func (a *Allocator) RunnerSlots(id string) []*Slot {
 	allSlots := Values(a.slots)
 	// Filter slots to include only those belonging to the specified runner.
 	return Filter(allSlots, func(s *Slot) bool {
@@ -278,7 +280,7 @@ func (a *allocator) RunnerSlots(id string) []*Slot {
 
 // DeadSlots checks for any runners that have timed out and removes them.
 // It returns the slots associated with the dead runners.
-func (a *allocator) DeadSlots(deadRunnerIDs []string) []*Slot {
+func (a *Allocator) DeadSlots(deadRunnerIDs []string) []*Slot {
 	deadSlots := make([]*Slot, 0)
 	// Iterate through runners to check if any have timed out.
 	for _, runnerID := range deadRunnerIDs {
@@ -303,7 +305,7 @@ func (a *allocator) DeadSlots(deadRunnerIDs []string) []*Slot {
 }
 
 // StartSlot marks scheduled work as in progress
-func (a *allocator) StartSlot(slotID uuid.UUID) error {
+func (a *Allocator) StartSlot(slotID uuid.UUID) error {
 	// Find the slot.
 	slot, ok := a.slots.Load(slotID)
 	if !ok {
@@ -326,6 +328,6 @@ func (a *allocator) StartSlot(slotID uuid.UUID) error {
 	return nil
 }
 
-func (a *allocator) DeleteSlot(slotID uuid.UUID) {
+func (a *Allocator) DeleteSlot(slotID uuid.UUID) {
 	a.slots.Delete(slotID)
 }

api/pkg/scheduler/cluster.go

@@ -20,6 +20,11 @@ type cluster struct {
 	runnerTimeoutFunc TimeoutFunc                   // Function to check if runners have timed out.
 }
 
+var _ Cluster = &cluster{}
+
+// NOTE(milosgajdos): we really should make sure we return exported types.
+// If we want the type fields to be inaccessible we should make them unexported.
+// nolint:revive
 func NewCluster(runnerTimeoutFunc TimeoutFunc) *cluster {
 	return &cluster{
 		runners:           xsync.NewMapOf[string, *runner](),

api/pkg/scheduler/scheduler.go

@@ -50,6 +50,10 @@ var _ Scheduler = &scheduler{}
 
 // NewScheduler creates a new scheduler with a workload allocator.
 // This also starts a goroutine to process the queue in the background.
+//
+// NOTE(milosgajdos): we really should make sure we return exported types.
+// If we want the type fields to be inaccessible we should make them unexported.
+// nolint:revive
 func NewScheduler(ctx context.Context, cfg *config.ServerConfig, onSchedulingErr func(work *Workload, err error)) *scheduler {
 	scheduler := newSchedulerWithoutGoroutines(cfg, onSchedulingErr)