1 files changed, 69 insertions, 0 deletions
diff --git a/skate/zipkey/batch.go b/skate/zipkey/batch.go
new file mode 100644
index 0000000..d81897d
--- /dev/null
+++ b/skate/zipkey/batch.go
@@ -0,0 +1,69 @@
+package zipkey
+
+import (
+	"runtime"
+	"sync"
+)
+
+// Batcher runs reducers in parallel on batches of groups.
+type Batcher struct {
+	Size       int
+	NumWorkers int
+	gf         groupFunc
+	batch      []*Group
+	queue      chan []*Group
+	wg         sync.WaitGroup
+	err        error
+}
+
+// NewBatcher set ups a new Batcher.
+func NewBatcher(gf groupFunc) *Batcher {
+	batcher := Batcher{
+		gf:         gf,
+		Size:       1000,
+		NumWorkers: runtime.NumCPU(),
+		queue:      make(chan []*Group),
+	}
+	for i := 0; i < batcher.NumWorkers; i++ {
+		batcher.wg.Add(1)
+		go batcher.worker(batcher.queue, &batcher.wg)
+	}
+	return &batcher
+}
+
+func (b *Batcher) Close() error {
+	g := make([]*Group, len(b.batch))
+	copy(g, b.batch)
+	b.queue <- g
+	b.batch = nil
+	close(b.queue)
+	b.wg.Wait()
+	return b.err
+}
+
+// GroupFunc implement the groupFunc type.
+func (b *Batcher) GroupFunc(g *Group) error {
+	b.batch = append(b.batch, g)
+	if len(b.batch) == b.Size {
+		g := make([]*Group, len(b.batch))
+		copy(g, b.batch)
+		b.queue <- g
+		b.batch = nil
+	}
+	return nil
+}
+
+// worker will wind down after a first error encountered.
+func (b *Batcher) worker(queue chan []*Group, wg *sync.WaitGroup) {
+	defer wg.Done()
+OUTER:
+	for batch := range queue {
+		for _, g := range batch {
+			err := b.gf(g)
+			if err != nil {
+				b.err = err
+				break OUTER
+			}
+		}
+	}
+}