snapshot of https://github.com/seaweedfs/seaweedfs.git (master, 2026-09-16; blobs > 131072B omitted)

weed/filer/arangodb/arangodb_store.go

@@ -0,0 +1,356 @@
+package arangodb
+
+import (
+	"context"
+	"crypto/tls"
+	"fmt"
+	"strconv"
+	"sync"
+	"time"
+
+	"github.com/arangodb/go-driver"
+	"github.com/arangodb/go-driver/http"
+	"github.com/seaweedfs/seaweedfs/weed/filer"
+	"github.com/seaweedfs/seaweedfs/weed/glog"
+	"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
+	"github.com/seaweedfs/seaweedfs/weed/util"
+)
+
+func init() {
+	filer.Stores = append(filer.Stores, &ArangodbStore{})
+}
+
+var (
+	BUCKET_PREFIX      = "/buckets"
+	DEFAULT_COLLECTION = "seaweed_no_bucket"
+	KVMETA_COLLECTION  = "seaweed_kvmeta"
+)
+
+type ArangodbStore struct {
+	connect      driver.Connection
+	client       driver.Client
+	database     driver.Database
+	kvCollection driver.Collection
+
+	buckets map[string]driver.Collection
+	mu      sync.RWMutex
+
+	databaseName string
+}
+
+type Model struct {
+	Key       string `json:"_key"`
+	Directory string `json:"directory,omitempty"`
+	Name      string `json:"name,omitempty"`
+	Ttl       string `json:"ttl,omitempty"`
+
+	//arangodb does not support binary blobs
+	//we encode byte slice into uint64 slice
+	//see helpers.go
+	Meta []uint64 `json:"meta"`
+}
+
+func (store *ArangodbStore) GetName() string {
+	return "arangodb"
+}
+
+func (store *ArangodbStore) Initialize(configuration util.Configuration, prefix string) (err error) {
+	store.buckets = make(map[string]driver.Collection, 3)
+	store.databaseName = configuration.GetString(prefix + "db_name")
+	return store.connection(configuration.GetStringSlice(prefix+"servers"),
+		configuration.GetString(prefix+"username"),
+		configuration.GetString(prefix+"password"),
+		configuration.GetBool(prefix+"insecure_skip_verify"),
+	)
+}
+
+func (store *ArangodbStore) connection(uris []string, user string, pass string, insecure bool) (err error) {
+	ctx, cn := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cn()
+	store.connect, err = http.NewConnection(http.ConnectionConfig{
+		Endpoints: uris,
+		TLSConfig: &tls.Config{
+			InsecureSkipVerify: insecure,
+		},
+	})
+	if err != nil {
+		return err
+	}
+	store.client, err = driver.NewClient(driver.ClientConfig{
+		Connection:     store.connect,
+		Authentication: driver.BasicAuthentication(user, pass),
+	})
+	if err != nil {
+		return err
+	}
+	ok, err := store.client.DatabaseExists(ctx, store.databaseName)
+	if err != nil {
+		return err
+	}
+	if ok {
+		store.database, err = store.client.Database(ctx, store.databaseName)
+	} else {
+		store.database, err = store.client.CreateDatabase(ctx, store.databaseName, &driver.CreateDatabaseOptions{})
+	}
+	if err != nil {
+		return err
+	}
+	if store.kvCollection, err = store.ensureCollection(ctx, KVMETA_COLLECTION); err != nil {
+		return err
+	}
+	return err
+}
+
+type key int
+
+const (
+	transactionKey key = 0
+)
+
+func (store *ArangodbStore) BeginTransaction(ctx context.Context) (context.Context, error) {
+	keys := make([]string, 0, len(store.buckets)+1)
+	for k := range store.buckets {
+		keys = append(keys, k)
+	}
+	keys = append(keys, store.kvCollection.Name())
+	txn, err := store.database.BeginTransaction(ctx, driver.TransactionCollections{
+		Exclusive: keys,
+	}, &driver.BeginTransactionOptions{})
+	if err != nil {
+		return nil, err
+	}
+
+	return context.WithValue(driver.WithTransactionID(ctx, txn), transactionKey, txn), nil
+}
+
+func (store *ArangodbStore) CommitTransaction(ctx context.Context) error {
+	val := ctx.Value(transactionKey)
+	cast, ok := val.(driver.TransactionID)
+	if !ok {
+		return fmt.Errorf("txn cast fail %s:", val)
+	}
+	err := store.database.CommitTransaction(ctx, cast, &driver.CommitTransactionOptions{})
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+func (store *ArangodbStore) RollbackTransaction(ctx context.Context) error {
+	val := ctx.Value(transactionKey)
+	cast, ok := val.(driver.TransactionID)
+	if !ok {
+		return fmt.Errorf("txn cast fail %s:", val)
+	}
+	err := store.database.AbortTransaction(ctx, cast, &driver.AbortTransactionOptions{})
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+func (store *ArangodbStore) InsertEntry(ctx context.Context, entry *filer.Entry) (err error) {
+	dir, name := entry.FullPath.DirAndName()
+	meta, err := entry.EncodeAttributesAndChunks()
+	if err != nil {
+		return fmt.Errorf("encode %s: %s", entry.FullPath, err)
+	}
+
+	if len(entry.GetChunks()) > filer.CountEntryChunksForGzip {
+		meta = util.MaybeGzipData(meta)
+	}
+	model := &Model{
+		Key:       hashString(string(entry.FullPath)),
+		Directory: dir,
+		Name:      name,
+		Meta:      bytesToArray(meta),
+	}
+	if entry.TtlSec > 0 {
+		model.Ttl = time.Now().Add(time.Second * time.Duration(entry.TtlSec)).Format(time.RFC3339)
+	} else {
+		model.Ttl = ""
+	}
+
+	targetCollection, err := store.extractBucketCollection(ctx, entry.FullPath)
+	if err != nil {
+		return err
+	}
+	_, err = targetCollection.CreateDocument(ctx, model)
+	if driver.IsConflict(err) {
+		return store.UpdateEntry(ctx, entry)
+	}
+
+	if err != nil {
+		return fmt.Errorf("InsertEntry %s: %v", entry.FullPath, err)
+	}
+
+	return nil
+
+}
+
+func (store *ArangodbStore) UpdateEntry(ctx context.Context, entry *filer.Entry) (err error) {
+	dir, name := entry.FullPath.DirAndName()
+	meta, err := entry.EncodeAttributesAndChunks()
+	if err != nil {
+		return fmt.Errorf("encode %s: %s", entry.FullPath, err)
+	}
+
+	if len(entry.GetChunks()) > filer.CountEntryChunksForGzip {
+		meta = util.MaybeGzipData(meta)
+	}
+	model := &Model{
+		Key:       hashString(string(entry.FullPath)),
+		Directory: dir,
+		Name:      name,
+		Meta:      bytesToArray(meta),
+	}
+	if entry.TtlSec > 0 {
+		model.Ttl = time.Now().Add(time.Duration(entry.TtlSec) * time.Second).Format(time.RFC3339)
+	} else {
+		model.Ttl = "none"
+	}
+	targetCollection, err := store.extractBucketCollection(ctx, entry.FullPath)
+	if err != nil {
+		return err
+	}
+	_, err = targetCollection.UpdateDocument(ctx, model.Key, model)
+	if err != nil {
+		return fmt.Errorf("UpdateEntry %s: %v", entry.FullPath, err)
+	}
+
+	return nil
+}
+
+func (store *ArangodbStore) FindEntry(ctx context.Context, fullpath util.FullPath) (entry *filer.Entry, err error) {
+	var data Model
+	targetCollection, err := store.extractBucketCollection(ctx, fullpath)
+	if err != nil {
+		return nil, err
+	}
+	_, err = targetCollection.ReadDocument(ctx, hashString(string(fullpath)), &data)
+	if err != nil {
+		if driver.IsNotFound(err) {
+			return nil, filer_pb.ErrNotFound
+		}
+		glog.ErrorfCtx(ctx, "find %s: %v", fullpath, err)
+		return nil, filer_pb.ErrNotFound
+	}
+	if len(data.Meta) == 0 {
+		return nil, filer_pb.ErrNotFound
+	}
+	entry = &filer.Entry{
+		FullPath: fullpath,
+	}
+	err = entry.DecodeAttributesAndChunks(util.MaybeDecompressData(arrayToBytes(data.Meta)))
+	if err != nil {
+		return entry, fmt.Errorf("decode %s : %v", entry.FullPath, err)
+	}
+
+	return entry, nil
+}
+
+func (store *ArangodbStore) DeleteEntry(ctx context.Context, fullpath util.FullPath) (err error) {
+	targetCollection, err := store.extractBucketCollection(ctx, fullpath)
+	if err != nil {
+		return err
+	}
+	_, err = targetCollection.RemoveDocument(ctx, hashString(string(fullpath)))
+	if err != nil && !driver.IsNotFound(err) {
+		glog.ErrorfCtx(ctx, "find %s: %v", fullpath, err)
+		return fmt.Errorf("delete %s : %v", fullpath, err)
+	}
+	return nil
+}
+
+// this runs in log time
+func (store *ArangodbStore) DeleteFolderChildren(ctx context.Context, fullpath util.FullPath) (err error) {
+	targetCollection, err := store.extractBucketCollection(ctx, fullpath)
+	if err != nil {
+		return err
+	}
+	query := `
+	for d in @@collection
+	filter starts_with(d.directory, @dirPrefix)  || d.directory == @dir
+	remove d._key in @@collection`
+	cur, err := store.database.Query(ctx, query, map[string]interface{}{
+		"@collection": targetCollection.Name(),
+		"dirPrefix":   string(fullpath) + "/",
+		"dir":         string(fullpath),
+	})
+	if err != nil {
+		return fmt.Errorf("delete %s : %v", fullpath, err)
+	}
+	defer cur.Close()
+	return nil
+}
+
+func (store *ArangodbStore) ListDirectoryEntries(ctx context.Context, dirPath util.FullPath, startFileName string, includeStartFile bool, limit int64, eachEntryFunc filer.ListEachEntryFunc) (lastFileName string, err error) {
+	return store.ListDirectoryPrefixedEntries(ctx, dirPath, startFileName, includeStartFile, limit, "", eachEntryFunc)
+}
+
+func (store *ArangodbStore) ListDirectoryPrefixedEntries(ctx context.Context, dirPath util.FullPath, startFileName string, includeStartFile bool, limit int64, prefix string, eachEntryFunc filer.ListEachEntryFunc) (lastFileName string, err error) {
+	targetCollection, err := store.extractBucketCollection(ctx, dirPath+"/")
+	if err != nil {
+		return lastFileName, err
+	}
+	bindVars := map[string]interface{}{
+		"@collection": targetCollection.Name(),
+		"dir":         dirPath,
+		"startFile":   startFileName,
+	}
+	query := "for d in @@collection"
+	if includeStartFile {
+		query = query + " filter d.name >= @startFile "
+	} else {
+		query = query + " filter d.name > @startFile "
+	}
+	if prefix != "" {
+		query = query + "&& starts_with(d.name, @prefix)"
+		bindVars["prefix"] = prefix
+	}
+	query = query + `
+filter d.directory == @dir
+sort d.name asc
+`
+	if limit > 0 {
+		query = query + "limit " + strconv.Itoa(int(limit))
+	}
+	query = query + "\n return d"
+	cur, err := store.database.Query(ctx, query, bindVars)
+	if err != nil {
+		return lastFileName, fmt.Errorf("failed to list directory entries: find error: %w", err)
+	}
+	defer cur.Close()
+	for cur.HasMore() {
+		var data Model
+		_, err = cur.ReadDocument(ctx, &data)
+		if err != nil {
+			break
+		}
+		entry := &filer.Entry{
+			FullPath: util.NewFullPath(data.Directory, data.Name),
+		}
+		lastFileName = data.Name
+		converted := arrayToBytes(data.Meta)
+		if decodeErr := entry.DecodeAttributesAndChunks(util.MaybeDecompressData(converted)); decodeErr != nil {
+			err = decodeErr
+			glog.V(0).InfofCtx(ctx, "list %s : %v", entry.FullPath, err)
+			break
+		}
+
+		resEachEntryFunc, resEachEntryFuncErr := eachEntryFunc(entry)
+		if resEachEntryFuncErr != nil {
+			err = fmt.Errorf("failed to process eachEntryFunc: %w", resEachEntryFuncErr)
+			break
+		}
+
+		if !resEachEntryFunc {
+			break
+		}
+
+	}
+	return lastFileName, err
+}
+
+func (store *ArangodbStore) Shutdown() {
+}

weed/filer/arangodb/arangodb_store_bucket.go

@@ -0,0 +1,42 @@
+package arangodb
+
+import (
+	"context"
+	"time"
+
+	"github.com/arangodb/go-driver"
+	"github.com/seaweedfs/seaweedfs/weed/filer"
+	"github.com/seaweedfs/seaweedfs/weed/glog"
+)
+
+var _ filer.BucketAware = (*ArangodbStore)(nil)
+
+func (store *ArangodbStore) OnBucketCreation(bucket string) {
+	timeout, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	// create the collection && add to cache
+	_, err := store.ensureBucket(timeout, bucket)
+	if err != nil {
+		glog.Errorf("bucket create %s: %v", bucket, err)
+	}
+}
+func (store *ArangodbStore) OnBucketDeletion(bucket string) {
+	timeout, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	collection, err := store.ensureBucket(timeout, bucket)
+	if err != nil {
+		glog.Errorf("bucket delete %s: %v", bucket, err)
+		return
+	}
+	err = collection.Remove(timeout)
+	if err != nil && !driver.IsNotFound(err) {
+		glog.Errorf("bucket delete %s: %v", bucket, err)
+		return
+	}
+	store.mu.Lock()
+	delete(store.buckets, bucket)
+	store.mu.Unlock()
+}
+func (store *ArangodbStore) CanDropWholeBucket() bool {
+	return true
+}

weed/filer/arangodb/arangodb_store_kv.go

@@ -0,0 +1,54 @@
+package arangodb
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/arangodb/go-driver"
+	"github.com/seaweedfs/seaweedfs/weed/filer"
+	"github.com/seaweedfs/seaweedfs/weed/glog"
+)
+
+func (store *ArangodbStore) KvPut(ctx context.Context, key []byte, value []byte) (err error) {
+	model := &Model{
+		Key:       hashString(".kvstore." + string(key)),
+		Directory: ".kvstore." + string(key),
+		Meta:      bytesToArray(value),
+	}
+
+	exists, err := store.kvCollection.DocumentExists(ctx, model.Key)
+	if err != nil {
+		return fmt.Errorf("kv put: %w", err)
+	}
+	if exists {
+		_, err = store.kvCollection.UpdateDocument(ctx, model.Key, model)
+	} else {
+		_, err = store.kvCollection.CreateDocument(ctx, model)
+	}
+	if err != nil {
+		return fmt.Errorf("kv put: %w", err)
+	}
+
+	return nil
+}
+func (store *ArangodbStore) KvGet(ctx context.Context, key []byte) (value []byte, err error) {
+	var model Model
+	_, err = store.kvCollection.ReadDocument(ctx, hashString(".kvstore."+string(key)), &model)
+	if driver.IsNotFound(err) {
+		return nil, filer.ErrKvNotFound
+	}
+	if err != nil {
+		glog.ErrorfCtx(ctx, "kv get: %s %v", string(key), err)
+		return nil, filer.ErrKvNotFound
+	}
+	return arrayToBytes(model.Meta), nil
+}
+
+func (store *ArangodbStore) KvDelete(ctx context.Context, key []byte) (err error) {
+	_, err = store.kvCollection.RemoveDocument(ctx, hashString(".kvstore."+string(key)))
+	if err != nil {
+		glog.ErrorfCtx(ctx, "kv del: %v", err)
+		return filer.ErrKvNotFound
+	}
+	return nil
+}

weed/filer/arangodb/arangodb_store_test.go

@@ -0,0 +1,195 @@
+package arangodb
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"testing"
+	"time"
+
+	"github.com/arangodb/go-driver"
+	"github.com/seaweedfs/seaweedfs/weed/filer"
+	"github.com/seaweedfs/seaweedfs/weed/util"
+)
+
+func newTestStore(t *testing.T) *ArangodbStore {
+	t.Helper()
+
+	if os.Getenv("RUN_ARANGODB_TESTS") != "1" {
+		t.Skip("arangodb tests are disabled. Start an arangodb server and set RUN_ARANGODB_TESTS=1 to enable, ARANGODB_ADDR defaults to http://127.0.0.1:8529.")
+	}
+
+	addr := os.Getenv("ARANGODB_ADDR")
+	if addr == "" {
+		addr = "http://127.0.0.1:8529"
+	}
+	user := os.Getenv("ARANGODB_USER")
+	if user == "" {
+		user = "root"
+	}
+
+	store := &ArangodbStore{databaseName: fmt.Sprintf("seaweed_test_%d", time.Now().UnixNano())}
+	store.buckets = make(map[string]driver.Collection, 3)
+	if err := store.connection([]string{addr}, user, os.Getenv("ARANGODB_PASSWORD"), true); err != nil {
+		t.Fatalf("connect to arangodb at %s: %v", addr, err)
+	}
+	t.Cleanup(func() {
+		if err := store.database.Remove(context.Background()); err != nil {
+			t.Errorf("drop test database: %v", err)
+		}
+	})
+	return store
+}
+
+func insertTestEntry(t *testing.T, store *ArangodbStore, path string) {
+	t.Helper()
+	if err := store.InsertEntry(context.Background(), &filer.Entry{FullPath: util.FullPath(path)}); err != nil {
+		t.Fatalf("insert %s: %v", path, err)
+	}
+}
+
+func countDocuments(t *testing.T, store *ArangodbStore, bucket string) int64 {
+	t.Helper()
+	collection, err := store.ensureBucket(context.Background(), bucket)
+	if err != nil {
+		t.Fatalf("ensure bucket %s: %v", bucket, err)
+	}
+	count, err := collection.Count(context.Background())
+	if err != nil {
+		t.Fatalf("count %s: %v", bucket, err)
+	}
+	return count
+}
+
+func listNames(t *testing.T, store *ArangodbStore, dir util.FullPath, prefix string) []string {
+	t.Helper()
+	var names []string
+	_, err := store.ListDirectoryPrefixedEntries(context.Background(), dir, "", true, 100, prefix,
+		func(entry *filer.Entry) (bool, error) {
+			names = append(names, entry.Name())
+			return true, nil
+		})
+	if err != nil {
+		t.Fatalf("list %s prefix %q: %v", dir, prefix, err)
+	}
+	return names
+}
+
+// AQL operators in a list prefix must be matched literally, never executed.
+func TestListDirectoryPrefixedEntriesAqlInjection(t *testing.T) {
+	store := newTestStore(t)
+
+	insertTestEntry(t, store, "/buckets/victim/secret")
+	insertTestEntry(t, store, "/buckets/tenant/regular")
+
+	injection := "x\" && false || (FOR q IN `victim` LIMIT 1 UPDATE q WITH {injected:\"yes\"} IN `victim` RETURN \"\") || \""
+	if names := listNames(t, store, util.FullPath("/buckets/tenant"), injection); len(names) != 0 {
+		t.Errorf("expected no match for injected prefix, got %v", names)
+	}
+
+	var victim Model
+	collection, err := store.ensureBucket(context.Background(), "victim")
+	if err != nil {
+		t.Fatalf("ensure bucket victim: %v", err)
+	}
+	if _, err := collection.ReadDocument(context.Background(), hashString("/buckets/victim/secret"), &victim); err != nil {
+		t.Fatalf("read victim document: %v", err)
+	}
+	if victim.Name != "secret" {
+		t.Errorf("victim document was modified across buckets: %+v", victim)
+	}
+}
+
+// A prefix full of AQL operators must not be evaluated as a query.
+func TestListDirectoryPrefixedEntriesPrefixNotEvaluated(t *testing.T) {
+	store := newTestStore(t)
+
+	insertTestEntry(t, store, "/buckets/tenant/a")
+	insertTestEntry(t, store, "/buckets/tenant/b")
+	insertTestEntry(t, store, "/buckets/tenant/c")
+
+	start := time.Now()
+	if names := listNames(t, store, util.FullPath("/buckets/tenant"), "x\" && false || TO_STRING(SLEEP(1)) && false || \""); len(names) != 0 {
+		t.Errorf("expected no match for injected prefix, got %v", names)
+	}
+	if elapsed := time.Since(start); elapsed > time.Second {
+		t.Errorf("SLEEP() in prefix was executed, listing took %v", elapsed)
+	}
+}
+
+// Quotes and newlines are ordinary object-name characters and must round-trip.
+func TestListDirectoryPrefixedEntriesQuotedNames(t *testing.T) {
+	store := newTestStore(t)
+
+	names := []string{"plain", "quo\"te", "new\nline", "back\\slash"}
+	for _, name := range names {
+		insertTestEntry(t, store, "/buckets/tenant/"+name)
+	}
+
+	for _, name := range names {
+		if got := listNames(t, store, util.FullPath("/buckets/tenant"), name); len(got) != 1 || got[0] != name {
+			t.Errorf("prefix %q listed %v, want exactly [%q]", name, got, name)
+		}
+	}
+
+	if got := listNames(t, store, util.FullPath("/buckets/tenant"), ""); len(got) != len(names) {
+		t.Errorf("empty prefix listed %v, want %d entries", got, len(names))
+	}
+}
+
+// A directory name full of AQL operators must only drop that directory's own children.
+func TestDeleteFolderChildrenAqlInjection(t *testing.T) {
+	store := newTestStore(t)
+
+	for i := 0; i < 5; i++ {
+		insertTestEntry(t, store, fmt.Sprintf("/buckets/tenant/keep%d", i))
+	}
+	injection := "x\" || true || \""
+	insertTestEntry(t, store, "/buckets/tenant/"+injection+"/child")
+
+	if err := store.DeleteFolderChildren(context.Background(), util.FullPath("/buckets/tenant/"+injection)); err != nil {
+		t.Fatalf("delete folder children: %v", err)
+	}
+
+	if got := countDocuments(t, store, "tenant"); got != 5 {
+		t.Errorf("tenant collection holds %d documents, want the 5 unrelated ones", got)
+	}
+}
+
+func TestDeleteFolderChildrenRemovesSubtree(t *testing.T) {
+	store := newTestStore(t)
+
+	insertTestEntry(t, store, "/buckets/tenant/dir/file")
+	insertTestEntry(t, store, "/buckets/tenant/dir/sub/deep")
+	insertTestEntry(t, store, "/buckets/tenant/dirX/sibling")
+	insertTestEntry(t, store, "/buckets/tenant/other")
+
+	if err := store.DeleteFolderChildren(context.Background(), util.FullPath("/buckets/tenant/dir")); err != nil {
+		t.Fatalf("delete folder children: %v", err)
+	}
+
+	if got := countDocuments(t, store, "tenant"); got != 2 {
+		t.Errorf("tenant collection holds %d documents, want /buckets/tenant/dirX/sibling and /buckets/tenant/other", got)
+	}
+}
+
+// startFileName is request-controlled too and must not break out of the filter.
+func TestListDirectoryEntriesQuotedStartFileName(t *testing.T) {
+	store := newTestStore(t)
+
+	insertTestEntry(t, store, "/buckets/tenant/a")
+	insertTestEntry(t, store, "/buckets/tenant/z")
+
+	var names []string
+	_, err := store.ListDirectoryEntries(context.Background(), util.FullPath("/buckets/tenant"), "z\" || true || \"", false, 100,
+		func(entry *filer.Entry) (bool, error) {
+			names = append(names, entry.Name())
+			return true, nil
+		})
+	if err != nil {
+		t.Fatalf("list with quoted start file: %v", err)
+	}
+	if len(names) != 0 {
+		t.Errorf("listed %v, want nothing sorted after the literal start name", names)
+	}
+}

weed/filer/arangodb/helpers.go

@@ -0,0 +1,159 @@
+package arangodb
+
+import (
+	"context"
+	"crypto/md5"
+	"encoding/binary"
+	"encoding/hex"
+	"io"
+	"strings"
+
+	"github.com/arangodb/go-driver"
+	"github.com/seaweedfs/seaweedfs/weed/util"
+)
+
+// convert a string into arango-key safe hex bytes hash
+func hashString(dir string) string {
+	h := md5.New()
+	io.WriteString(h, dir)
+	b := h.Sum(nil)
+	return hex.EncodeToString(b)
+}
+
+// convert slice of bytes into slice of uint64
+// the first uint64 indicates the length in bytes
+func bytesToArray(bs []byte) []uint64 {
+	out := make([]uint64, 0, 2+len(bs)/8)
+	out = append(out, uint64(len(bs)))
+	for len(bs)%8 != 0 {
+		bs = append(bs, 0)
+	}
+	for i := 0; i < len(bs); i = i + 8 {
+		out = append(out, binary.BigEndian.Uint64(bs[i:]))
+	}
+	return out
+}
+
+// convert from slice of uint64 back to bytes
+// if input length is 0 or 1, will return nil
+func arrayToBytes(xs []uint64) []byte {
+	if len(xs) < 2 {
+		return nil
+	}
+	first := xs[0]
+	out := make([]byte, len(xs)*8) // i think this can actually be len(xs)*8-8, but i dont think an extra 8 bytes hurts...
+	for i := 1; i < len(xs); i = i + 1 {
+		binary.BigEndian.PutUint64(out[((i-1)*8):], xs[i])
+	}
+	return out[:first]
+}
+
+// gets the collection the bucket points to from filepath
+func (store *ArangodbStore) extractBucketCollection(ctx context.Context, fullpath util.FullPath) (c driver.Collection, err error) {
+	bucket, _ := extractBucket(fullpath)
+	if bucket == "" {
+		bucket = DEFAULT_COLLECTION
+	}
+	c, err = store.ensureBucket(ctx, bucket)
+	if err != nil {
+		return nil, err
+	}
+	return c, err
+}
+
+// called by extractBucketCollection
+func extractBucket(fullpath util.FullPath) (string, string) {
+	if !strings.HasPrefix(string(fullpath), BUCKET_PREFIX+"/") {
+		return "", string(fullpath)
+	}
+	if strings.Count(string(fullpath), "/") < 3 {
+		return "", string(fullpath)
+	}
+	bucketAndObjectKey := string(fullpath)[len(BUCKET_PREFIX+"/"):]
+	t := strings.Index(bucketAndObjectKey, "/")
+	bucket := bucketAndObjectKey
+	shortPath := "/"
+	if t > 0 {
+		bucket = bucketAndObjectKey[:t]
+		shortPath = string(util.FullPath(bucketAndObjectKey[t:]))
+	}
+	// Dot-prefixed entries directly under /buckets (e.g. .system) are internal
+	// folders, not S3 buckets; keep them in the default collection.
+	if strings.HasPrefix(bucket, ".") {
+		return "", string(fullpath)
+	}
+	return bucket, shortPath
+}
+
+// get bucket collection from cache. if not exist, creates the buckets collection and grab it
+func (store *ArangodbStore) ensureBucket(ctx context.Context, bucket string) (bc driver.Collection, err error) {
+	var ok bool
+	store.mu.RLock()
+	bc, ok = store.buckets[bucket]
+	store.mu.RUnlock()
+	if ok && bc != nil {
+		return bc, nil
+	}
+	store.mu.Lock()
+	defer store.mu.Unlock()
+	store.buckets[bucket], err = store.ensureCollection(ctx, bucket)
+	if err != nil {
+		return nil, err
+	}
+	return store.buckets[bucket], nil
+}
+
+// transform to an arango compliant name
+func bucketToCollectionName(s string) string {
+	if len(s) == 0 {
+		return ""
+	}
+	// replace all "." with _
+	s = strings.ReplaceAll(s, ".", "_")
+
+	// if starts with number or '.' then add a special prefix
+	if (s[0] >= '0' && s[0] <= '9') || (s[0] == '.' || s[0] == '_' || s[0] == '-') {
+		s = "xN--" + s
+	}
+	return s
+}
+
+// creates collection if not exist, ensures indices if not exist
+func (store *ArangodbStore) ensureCollection(ctx context.Context, bucket_name string) (c driver.Collection, err error) {
+	// convert the bucket to collection name
+	name := bucketToCollectionName(bucket_name)
+
+	ok, err := store.database.CollectionExists(ctx, name)
+	if err != nil {
+		return
+	}
+	if ok {
+		c, err = store.database.Collection(ctx, name)
+	} else {
+		c, err = store.database.CreateCollection(ctx, name, &driver.CreateCollectionOptions{})
+	}
+	if err != nil {
+		return
+	}
+	// ensure indices
+	if _, _, err = c.EnsurePersistentIndex(ctx, []string{"directory", "name"},
+		&driver.EnsurePersistentIndexOptions{
+			Name: "directory_name_multi", Unique: true,
+		}); err != nil {
+		return
+	}
+	if _, _, err = c.EnsurePersistentIndex(ctx, []string{"directory"},
+		&driver.EnsurePersistentIndexOptions{Name: "IDX_directory"}); err != nil {
+		return
+	}
+	if _, _, err = c.EnsureTTLIndex(ctx, "ttl", 1,
+		&driver.EnsureTTLIndexOptions{Name: "IDX_TTL"}); err != nil {
+		return
+	}
+	if _, _, err = c.EnsurePersistentIndex(ctx, []string{"name"}, &driver.EnsurePersistentIndexOptions{
+		Name: "IDX_name",
+	}); err != nil {
+		return
+	}
+	return c, nil
+}

weed/filer/arangodb/readme.md

@@ -0,0 +1,85 @@
+##arangodb
+
+database: https://github.com/arangodb/arangodb
+go driver: https://github.com/arangodb/go-driver
+
+options:
+
+```
+[arangodb]
+enabled=true
+db_name="seaweedfs"
+servers=["http://localhost:8529"]
+#basic auth
+user="root"
+pass="test"
+
+# tls settings
+insecure_skip_verify=true
+```
+
+i test using this dev database:
+`docker run -p 8529:8529 -e ARANGO_ROOT_PASSWORD=test arangodb/arangodb:3.9.0`
+
+
+
+## database structure
+
+
+arangodb has a few restrictions which require the use of a few tricks in order to losslessly store the data.
+
+### filer store
+
+arangodb does not support []byte, and will store such as a uint64 array. this would be a waste of space. to counteract this, we store the data as a length prefixed uint64 byteset.
+
+### filer kv
+
+same as above
+
+### filer buckets
+
+s3 buckets are implemented through arangodb collection. this allows us to do very fast bucket deletion by simply deleting the collection
+
+
+arangodb collection name rules is character set `azAZ09_-` with a 256 character max. however the first character must be a letter.
+
+
+s3 bucket name rule is the set `azAZ09.-` with a 63 characters max.
+
+the rules for collection names is then the following:
+
+1. if the bucket name is a valid arangodb collection name, then nothing is done.
+2. if the bucket name contains a ".", the "." is replaced with "_"
+3. if the bucket name now begins with a number or "_", the prefix "xN--" is prepended to the collection name
+
+this allows for these collection names to be used.
+
+
+## features i don't personally need but are missing
+ [ ] provide tls cert to arango
+ [ ] authentication that is not basic auth
+ [ ] synchronise endpoint interval config
+ [ ] automatic creation of custom index
+ [ ] configure default arangodb collection sharding rules
+ [ ] configure default arangodb collection replication rules
+
+
+## complexity
+
+ok, so if https://www.arangodb.com/docs/stable/indexing-index-basics.html#persistent-index is correct
+
+O(1)
+- InsertEntry
+- UpdateEntry
+- FindEntry
+- DeleteEntry
+- KvPut
+- KvGet
+- KvDelete
+
+O(log(BUCKET_SIZE))
+- DeleteFolderChildren
+
+O(log(DIRECTORY_SIZE))
+- ListDirectoryEntries
+- ListDirectoryPrefixedEntries