Move truncindex in separate package in pkg/

Docker-DCO-1.1-Signed-off-by: Alexandr Morozov <lk4d4math@gmail.com> (github: LK4D4)

truncindex/truncindex.go

@@ -0,0 +1,102 @@
+package truncindex
+
+import (
+	"fmt"
+	"index/suffixarray"
+	"strings"
+	"sync"
+)
+
+// TruncIndex allows the retrieval of string identifiers by any of their unique prefixes.
+// This is used to retrieve image and container IDs by more convenient shorthand prefixes.
+type TruncIndex struct {
+	sync.RWMutex
+	index *suffixarray.Index
+	ids   map[string]bool
+	bytes []byte
+}
+
+func NewTruncIndex(ids []string) (idx *TruncIndex) {
+	idx = &TruncIndex{
+		ids:   make(map[string]bool),
+		bytes: []byte{' '},
+	}
+	for _, id := range ids {
+		idx.ids[id] = true
+		idx.bytes = append(idx.bytes, []byte(id+" ")...)
+	}
+	idx.index = suffixarray.New(idx.bytes)
+	return
+}
+
+func (idx *TruncIndex) addId(id string) error {
+	if strings.Contains(id, " ") {
+		return fmt.Errorf("Illegal character: ' '")
+	}
+	if _, exists := idx.ids[id]; exists {
+		return fmt.Errorf("Id already exists: %s", id)
+	}
+	idx.ids[id] = true
+	idx.bytes = append(idx.bytes, []byte(id+" ")...)
+	return nil
+}
+
+func (idx *TruncIndex) Add(id string) error {
+	idx.Lock()
+	defer idx.Unlock()
+	if err := idx.addId(id); err != nil {
+		return err
+	}
+	idx.index = suffixarray.New(idx.bytes)
+	return nil
+}
+
+func (idx *TruncIndex) AddWithoutSuffixarrayUpdate(id string) error {
+	idx.Lock()
+	defer idx.Unlock()
+	return idx.addId(id)
+}
+
+func (idx *TruncIndex) UpdateSuffixarray() {
+	idx.Lock()
+	defer idx.Unlock()
+	idx.index = suffixarray.New(idx.bytes)
+}
+
+func (idx *TruncIndex) Delete(id string) error {
+	idx.Lock()
+	defer idx.Unlock()
+	if _, exists := idx.ids[id]; !exists {
+		return fmt.Errorf("No such id: %s", id)
+	}
+	before, after, err := idx.lookup(id)
+	if err != nil {
+		return err
+	}
+	delete(idx.ids, id)
+	idx.bytes = append(idx.bytes[:before], idx.bytes[after:]...)
+	idx.index = suffixarray.New(idx.bytes)
+	return nil
+}
+
+func (idx *TruncIndex) lookup(s string) (int, int, error) {
+	offsets := idx.index.Lookup([]byte(" "+s), -1)
+	//log.Printf("lookup(%s): %v (index bytes: '%s')\n", s, offsets, idx.index.Bytes())
+	if offsets == nil || len(offsets) == 0 || len(offsets) > 1 {
+		return -1, -1, fmt.Errorf("No such id: %s", s)
+	}
+	offsetBefore := offsets[0] + 1
+	offsetAfter := offsetBefore + strings.Index(string(idx.bytes[offsetBefore:]), " ")
+	return offsetBefore, offsetAfter, nil
+}
+
+func (idx *TruncIndex) Get(s string) (string, error) {
+	idx.RLock()
+	defer idx.RUnlock()
+	before, after, err := idx.lookup(s)
+	//log.Printf("Get(%s) bytes=|%s| before=|%d| after=|%d|\n", s, idx.bytes, before, after)
+	if err != nil {
+		return "", err
+	}
+	return string(idx.bytes[before:after]), err
+}

truncindex/truncindex_test.go

@@ -0,0 +1,105 @@
+package truncindex
+
+import "testing"
+
+// Test the behavior of TruncIndex, an index for querying IDs from a non-conflicting prefix.
+func TestTruncIndex(t *testing.T) {
+	ids := []string{}
+	index := NewTruncIndex(ids)
+	// Get on an empty index
+	if _, err := index.Get("foobar"); err == nil {
+		t.Fatal("Get on an empty index should return an error")
+	}
+
+	// Spaces should be illegal in an id
+	if err := index.Add("I have a space"); err == nil {
+		t.Fatalf("Adding an id with ' ' should return an error")
+	}
+
+	id := "99b36c2c326ccc11e726eee6ee78a0baf166ef96"
+	// Add an id
+	if err := index.Add(id); err != nil {
+		t.Fatal(err)
+	}
+	// Get a non-existing id
+	assertIndexGet(t, index, "abracadabra", "", true)
+	// Get the exact id
+	assertIndexGet(t, index, id, id, false)
+	// The first letter should match
+	assertIndexGet(t, index, id[:1], id, false)
+	// The first half should match
+	assertIndexGet(t, index, id[:len(id)/2], id, false)
+	// The second half should NOT match
+	assertIndexGet(t, index, id[len(id)/2:], "", true)
+
+	id2 := id[:6] + "blabla"
+	// Add an id
+	if err := index.Add(id2); err != nil {
+		t.Fatal(err)
+	}
+	// Both exact IDs should work
+	assertIndexGet(t, index, id, id, false)
+	assertIndexGet(t, index, id2, id2, false)
+
+	// 6 characters or less should conflict
+	assertIndexGet(t, index, id[:6], "", true)
+	assertIndexGet(t, index, id[:4], "", true)
+	assertIndexGet(t, index, id[:1], "", true)
+
+	// 7 characters should NOT conflict
+	assertIndexGet(t, index, id[:7], id, false)
+	assertIndexGet(t, index, id2[:7], id2, false)
+
+	// Deleting a non-existing id should return an error
+	if err := index.Delete("non-existing"); err == nil {
+		t.Fatalf("Deleting a non-existing id should return an error")
+	}
+
+	// Deleting id2 should remove conflicts
+	if err := index.Delete(id2); err != nil {
+		t.Fatal(err)
+	}
+	// id2 should no longer work
+	assertIndexGet(t, index, id2, "", true)
+	assertIndexGet(t, index, id2[:7], "", true)
+	assertIndexGet(t, index, id2[:11], "", true)
+
+	// conflicts between id and id2 should be gone
+	assertIndexGet(t, index, id[:6], id, false)
+	assertIndexGet(t, index, id[:4], id, false)
+	assertIndexGet(t, index, id[:1], id, false)
+
+	// non-conflicting substrings should still not conflict
+	assertIndexGet(t, index, id[:7], id, false)
+	assertIndexGet(t, index, id[:15], id, false)
+	assertIndexGet(t, index, id, id, false)
+}
+
+func assertIndexGet(t *testing.T, index *TruncIndex, input, expectedResult string, expectError bool) {
+	if result, err := index.Get(input); err != nil && !expectError {
+		t.Fatalf("Unexpected error getting '%s': %s", input, err)
+	} else if err == nil && expectError {
+		t.Fatalf("Getting '%s' should return an error", input)
+	} else if result != expectedResult {
+		t.Fatalf("Getting '%s' returned '%s' instead of '%s'", input, result, expectedResult)
+	}
+}
+
+func BenchmarkTruncIndexAdd(b *testing.B) {
+	ids := []string{"banana", "bananaa", "bananab"}
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		index := NewTruncIndex([]string{})
+		for _, id := range ids {
+			index.Add(id)
+		}
+	}
+}
+
+func BenchmarkTruncIndexNew(b *testing.B) {
+	ids := []string{"banana", "bananaa", "bananab"}
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		NewTruncIndex(ids)
+	}
+}