Merge pull request #242 from barakmich/unique_iterator

Implement the Unique iterator

graph/iterator.go

@@ -237,6 +237,7 @@ const (
 	Not
 	Optional
 	Materialize
+	Unique
 )
 
 var (
@@ -258,6 +259,7 @@ var (
 		"not",
 		"optional",
 		"materialize",
+		"unique",
 	}
 )
 

graph/iterator/unique_iterator.go

@@ -0,0 +1,162 @@
+package iterator
+
+import (
+	"github.com/google/cayley/graph"
+)
+
+// Unique iterator removes duplicate values from it's subiterator.
+type Unique struct {
+	uid      uint64
+	tags     graph.Tagger
+	subIt    graph.Iterator
+	result   graph.Value
+	runstats graph.IteratorStats
+	err      error
+	seen     map[graph.Value]bool
+}
+
+func NewUnique(subIt graph.Iterator) *Unique {
+	return &Unique{
+		uid:   NextUID(),
+		subIt: subIt,
+		seen:  make(map[graph.Value]bool),
+	}
+}
+
+func (it *Unique) UID() uint64 {
+	return it.uid
+}
+
+// Reset resets the internal iterators and the iterator itself.
+func (it *Unique) Reset() {
+	it.result = nil
+	it.subIt.Reset()
+	it.seen = make(map[graph.Value]bool)
+}
+
+func (it *Unique) Tagger() *graph.Tagger {
+	return &it.tags
+}
+
+func (it *Unique) TagResults(dst map[string]graph.Value) {
+	for _, tag := range it.tags.Tags() {
+		dst[tag] = it.Result()
+	}
+
+	for tag, value := range it.tags.Fixed() {
+		dst[tag] = value
+	}
+
+	if it.subIt != nil {
+		it.subIt.TagResults(dst)
+	}
+}
+
+func (it *Unique) Clone() graph.Iterator {
+	uniq := NewUnique(it.subIt.Clone())
+	uniq.tags.CopyFrom(it)
+	return uniq
+}
+
+// SubIterators returns a slice of the sub iterators. The first iterator is the
+// primary iterator, for which the complement is generated.
+func (it *Unique) SubIterators() []graph.Iterator {
+	return []graph.Iterator{it.subIt}
+}
+
+// DEPRECATED
+func (it *Unique) ResultTree() *graph.ResultTree {
+	tree := graph.NewResultTree(it.Result())
+	tree.AddSubtree(it.subIt.ResultTree())
+	return tree
+}
+
+// Next advances the subiterator, continuing until it returns a value which it
+// has not previously seen.
+func (it *Unique) Next() bool {
+	graph.NextLogIn(it)
+	it.runstats.Next += 1
+
+	for graph.Next(it.subIt) {
+		curr := it.subIt.Result()
+		if ok := it.seen[curr]; !ok {
+			it.result = curr
+			it.seen[curr] = true
+			return graph.NextLogOut(it, it.result, true)
+		}
+	}
+	it.err = it.subIt.Err()
+	return graph.NextLogOut(it, nil, false)
+}
+
+func (it *Unique) Err() error {
+	return it.err
+}
+
+func (it *Unique) Result() graph.Value {
+	return it.result
+}
+
+// Contains checks whether the passed value is part of the primary iterator,
+// which is irrelevant for uniqueness.
+func (it *Unique) Contains(val graph.Value) bool {
+	graph.ContainsLogIn(it, val)
+	it.runstats.Contains += 1
+	return graph.ContainsLogOut(it, val, it.subIt.Contains(val))
+}
+
+// NextPath for unique always returns false. If we were to return multiple
+// paths, we'd no longer be a unique result, so we have to choose only the first
+// path that got us here. Unique is serious on this point.
+func (it *Unique) NextPath() bool {
+	return false
+}
+
+// Close closes the primary iterators.
+func (it *Unique) Close() error {
+	it.seen = nil
+	return it.subIt.Close()
+}
+
+func (it *Unique) Type() graph.Type { return graph.Unique }
+
+func (it *Unique) Optimize() (graph.Iterator, bool) {
+	newIt, optimized := it.subIt.Optimize()
+	if optimized {
+		it.subIt = newIt
+	}
+	return it, false
+}
+
+const uniquenessFactor = 2
+
+func (it *Unique) Stats() graph.IteratorStats {
+	subStats := it.subIt.Stats()
+	return graph.IteratorStats{
+		NextCost:     subStats.NextCost * uniquenessFactor,
+		ContainsCost: subStats.ContainsCost,
+		Size:         subStats.Size / uniquenessFactor,
+		Next:         it.runstats.Next,
+		Contains:     it.runstats.Contains,
+		ContainsNext: it.runstats.ContainsNext,
+	}
+}
+
+func (it *Unique) Size() (int64, bool) {
+	return it.Stats().Size, false
+}
+
+func (it *Unique) Describe() graph.Description {
+	subIts := []graph.Description{
+		it.subIt.Describe(),
+	}
+
+	return graph.Description{
+		UID:       it.UID(),
+		Type:      it.Type(),
+		Tags:      it.tags.Tags(),
+		Iterators: subIts,
+	}
+}
+
+var _ graph.Nexter = &Unique{}

graph/iterator/unique_iterator_test.go

@@ -0,0 +1,31 @@
+package iterator
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestUniqueIteratorBasics(t *testing.T) {
+	allIt := NewFixed(Identity)
+	allIt.Add(1)
+	allIt.Add(2)
+	allIt.Add(3)
+	allIt.Add(3)
+	allIt.Add(2)
+
+	u := NewUnique(allIt)
+
+	expect := []int{1, 2, 3}
+	for i := 0; i < 2; i++ {
+		if got := iterated(u); !reflect.DeepEqual(got, expect) {
+			t.Errorf("Failed to iterate Unique correctly on repeat %d: got:%v expected:%v", i, got, expect)
+		}
+		u.Reset()
+	}
+
+	for _, v := range []int{1, 2, 3} {
+		if !u.Contains(v) {
+			t.Errorf("Failed to find a correct value in the unique iterator.")
+		}
+	}
+}