From 48db56ddade2ef076fae4498ca9357a280270bc5 Mon Sep 17 00:00:00 2001
From: gfx <86091021+gfxlabs@users.noreply.github.com>
Date: Fri, 9 Sep 2022 11:43:42 -0500
Subject: [PATCH] arangodb s3 bucket name compatibility   (#3588)

* Update arangodb_store.go

* update readme, properly escape queries, add name patching

* use underscore

* use underscore

* better comment

* fix readme

Co-authored-by: a <a@a.a>
---
 weed/filer/arangodb/arangodb_store.go | 10 ++++----
 weed/filer/arangodb/helpers.go        | 22 ++++++++++++++++--
 weed/filer/arangodb/readme.md         | 33 +++++++++++++++++++++++++++
 3 files changed, 58 insertions(+), 7 deletions(-)

diff --git a/weed/filer/arangodb/arangodb_store.go b/weed/filer/arangodb/arangodb_store.go
index 25ef60bf0..ab5f8db4f 100644
--- a/weed/filer/arangodb/arangodb_store.go
+++ b/weed/filer/arangodb/arangodb_store.go
@@ -66,8 +66,8 @@ func (store *ArangodbStore) Initialize(configuration util.Configuration, prefix
 }
 
 func (store *ArangodbStore) connection(uris []string, user string, pass string, insecure bool) (err error) {
-	ctx, _ := context.WithTimeout(context.Background(), 10*time.Second)
-
+	ctx, cn := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cn()
 	store.connect, err = http.NewConnection(http.ConnectionConfig{
 		Endpoints: uris,
 		TLSConfig: &tls.Config{
@@ -274,10 +274,10 @@ func (store *ArangodbStore) DeleteFolderChildren(ctx context.Context, fullpath u
 	for d in %s
 	filter starts_with(d.directory, "%s/")  || d.directory == "%s"
 	remove d._key in %s`,
-		targetCollection.Name(),
+		"`"+targetCollection.Name()+"`",
 		strings.Join(strings.Split(string(fullpath), "/"), ","),
 		string(fullpath),
-		targetCollection.Name(),
+		"`"+targetCollection.Name()+"`",
 	)
 	cur, err := store.database.Query(ctx, query, nil)
 	if err != nil {
@@ -296,7 +296,7 @@ func (store *ArangodbStore) ListDirectoryPrefixedEntries(ctx context.Context, di
 	if err != nil {
 		return lastFileName, err
 	}
-	query := "for d in " + targetCollection.Name()
+	query := "for d in " + "`" + targetCollection.Name() + "`"
 	if includeStartFile {
 		query = query + " filter d.name >= \"" + startFileName + "\" "
 	} else {
diff --git a/weed/filer/arangodb/helpers.go b/weed/filer/arangodb/helpers.go
index 3f36acb0a..776e6d1b8 100644
--- a/weed/filer/arangodb/helpers.go
+++ b/weed/filer/arangodb/helpers.go
@@ -12,7 +12,7 @@ import (
 	"github.com/seaweedfs/seaweedfs/weed/util"
 )
 
-//convert a string into arango-key safe hex bytes hash
+// convert a string into arango-key safe hex bytes hash
 func hashString(dir string) string {
 	h := md5.New()
 	io.WriteString(h, dir)
@@ -98,8 +98,26 @@ func (store *ArangodbStore) ensureBucket(ctx context.Context, bucket string) (bc
 	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, name string) (c driver.Collection, err error) {
+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
diff --git a/weed/filer/arangodb/readme.md b/weed/filer/arangodb/readme.md
index e189811fb..57a594592 100644
--- a/weed/filer/arangodb/readme.md
+++ b/weed/filer/arangodb/readme.md
@@ -22,6 +22,39 @@ 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