Merge pull request #1471 from versity/fix/AzureNextMarker

fix: update marker/continuation token to be the azure next marker
This commit is contained in:
Ben McClelland
2025-08-25 13:58:06 -07:00
committed by GitHub
3 changed files with 260 additions and 114 deletions
+138 -108
View File
@@ -68,6 +68,8 @@ const (
onameAttr key = "Objname"
onameAttrLower key = "objname"
metaTmpMultipartPrefix key = ".sgwtmp" + "/multipart"
defaultListingMaxKeys = 1000
)
func (key) Table() map[string]struct{} {
@@ -579,26 +581,6 @@ func (az *Azure) GetObjectAttributes(ctx context.Context, input *s3.GetObjectAtt
}
func (az *Azure) ListObjects(ctx context.Context, input *s3.ListObjectsInput) (s3response.ListObjectsResult, error) {
client, err := az.getContainerClient(*input.Bucket)
if err != nil {
return s3response.ListObjectsResult{}, nil
}
pager := client.NewListBlobsHierarchyPager(*input.Delimiter, &container.ListBlobsHierarchyOptions{
Marker: input.Marker,
MaxResults: input.MaxKeys,
Prefix: input.Prefix,
})
var objects []s3response.Object
var cPrefixes []types.CommonPrefix
var nextMarker *string
var isTruncated bool
var maxKeys int32 = math.MaxInt32
if input.MaxKeys != nil {
maxKeys = *input.MaxKeys
}
// Retrieve the bucket acl to get the bucket owner
// All the objects in the bucket are owner by the bucket owner
aclBytes, err := az.getContainerMetaData(ctx, *input.Bucket, string(keyAclCapital))
@@ -611,19 +593,49 @@ func (az *Azure) ListObjects(ctx context.Context, input *s3.ListObjectsInput) (s
return s3response.ListObjectsResult{}, err
}
Pager:
for pager.More() {
client, err := az.getContainerClient(*input.Bucket)
if err != nil {
return s3response.ListObjectsResult{}, nil
}
var maxKeys int32 = defaultListingMaxKeys
if input.MaxKeys != nil {
maxKeys = *input.MaxKeys
}
pager := client.NewListBlobsHierarchyPager(*input.Delimiter, &container.ListBlobsHierarchyOptions{
MaxResults: &maxKeys,
Prefix: input.Prefix,
})
var objects []s3response.Object
var cPrefixes []types.CommonPrefix
var nextMarker *string
var isTruncated bool
// Convert marker to filter criteria
var markerFilter string
if input.Marker != nil && *input.Marker != "" {
markerFilter = *input.Marker
}
// Loop through pages until we have enough objects or no more pages
objectsFound := int32(0)
for pager.More() && objectsFound < maxKeys {
resp, err := pager.NextPage(ctx)
if err != nil {
return s3response.ListObjectsResult{}, azureErrToS3Err(err)
}
// Process objects from this page
var pageObjects []s3response.Object
for _, v := range resp.Segment.BlobItems {
if len(objects)+len(cPrefixes) >= int(maxKeys) {
nextMarker = objects[len(objects)-1].Key
isTruncated = true
break Pager
// Skip objects that come before or equal to marker
if markerFilter != "" && *v.Name <= markerFilter {
continue
}
objects = append(objects, s3response.Object{
pageObjects = append(pageObjects, s3response.Object{
ETag: backend.GetPtrFromString(convertAzureEtag(v.Properties.ETag)),
Key: v.Name,
LastModified: v.Properties.LastModified,
@@ -633,20 +645,22 @@ Pager:
ID: &acl.Owner,
},
})
}
for _, v := range resp.Segment.BlobPrefixes {
if *v.Name <= *input.Marker {
continue
}
if len(objects)+len(cPrefixes) >= int(maxKeys) {
nextMarker = cPrefixes[len(cPrefixes)-1].Prefix
isTruncated = true
break Pager
}
marker := getString(input.Marker)
pfx := strings.TrimSuffix(*v.Name, getString(input.Delimiter))
if marker != "" && strings.HasPrefix(marker, pfx) {
objectsFound++
if objectsFound >= maxKeys {
// Set next marker to the current object name for pagination
nextMarker = v.Name
isTruncated = true
break
}
}
objects = append(objects, pageObjects...)
// Process common prefixes from this page
for _, v := range resp.Segment.BlobPrefixes {
// Skip prefixes that come before or equal to marker
if markerFilter != "" && *v.Name <= markerFilter {
continue
}
@@ -654,6 +668,16 @@ Pager:
Prefix: v.Name,
})
}
// If we've reached maxKeys, break
if objectsFound >= maxKeys {
break
}
// If Azure indicates more pages but we need to continue for more objects
if resp.NextMarker != nil && *resp.NextMarker != "" && objectsFound < maxKeys {
continue
}
}
return s3response.ListObjectsResult{
@@ -670,98 +694,104 @@ Pager:
}
func (az *Azure) ListObjectsV2(ctx context.Context, input *s3.ListObjectsV2Input) (s3response.ListObjectsV2Result, error) {
marker := ""
if *input.ContinuationToken > *input.StartAfter {
marker = *input.ContinuationToken
} else {
marker = *input.StartAfter
// Retrieve the bucket acl to get the bucket owner
// All the objects in the bucket are owner by the bucket owner
aclBytes, err := az.getContainerMetaData(ctx, *input.Bucket, string(keyAclCapital))
if err != nil {
return s3response.ListObjectsV2Result{}, azureErrToS3Err(err)
}
acl, err := auth.ParseACL(aclBytes)
if err != nil {
return s3response.ListObjectsV2Result{}, err
}
client, err := az.getContainerClient(*input.Bucket)
if err != nil {
return s3response.ListObjectsV2Result{}, nil
}
var maxKeys int32 = defaultListingMaxKeys
if input.MaxKeys != nil {
maxKeys = *input.MaxKeys
}
pager := client.NewListBlobsHierarchyPager(*input.Delimiter, &container.ListBlobsHierarchyOptions{
Marker: &marker,
MaxResults: input.MaxKeys,
Marker: input.ContinuationToken,
MaxResults: &maxKeys,
Prefix: input.Prefix,
})
var objects []s3response.Object
var cPrefixes []types.CommonPrefix
var nextMarker *string
var isTruncated bool
var maxKeys int32 = math.MaxInt32
var fetchOwner bool
var resp container.ListBlobsHierarchyResponse
if input.MaxKeys != nil {
maxKeys = *input.MaxKeys
}
if input.FetchOwner != nil {
fetchOwner = *input.FetchOwner
}
// Retrieve the bucket acl to get the bucket owner, if "fetchOwner" is true
// All the objects in the bucket are owner by the bucket owner
var acl auth.ACL
if fetchOwner {
aclBytes, err := az.getContainerMetaData(ctx, *input.Bucket, string(keyAclCapital))
// Loop through pages until we find objects or no more pages
for {
resp, err = pager.NextPage(ctx)
if err != nil {
return s3response.ListObjectsV2Result{}, azureErrToS3Err(err)
}
acl, err = auth.ParseACL(aclBytes)
if err != nil {
return s3response.ListObjectsV2Result{}, err
}
}
Pager:
for pager.More() {
resp, err := pager.NextPage(ctx)
if err != nil {
return s3response.ListObjectsV2Result{}, azureErrToS3Err(err)
}
// Convert Azure objects to S3 objects
var pageObjects []s3response.Object
for _, v := range resp.Segment.BlobItems {
if len(objects)+len(cPrefixes) >= int(maxKeys) {
nextMarker = objects[len(objects)-1].Key
isTruncated = true
break Pager
}
obj := s3response.Object{
pageObjects = append(pageObjects, s3response.Object{
ETag: backend.GetPtrFromString(convertAzureEtag(v.Properties.ETag)),
Key: v.Name,
LastModified: v.Properties.LastModified,
Size: v.Properties.ContentLength,
StorageClass: types.ObjectStorageClassStandard,
}
if fetchOwner {
obj.Owner = &types.Owner{
Owner: &types.Owner{
ID: &acl.Owner,
}
}
objects = append(objects, obj)
}
for _, v := range resp.Segment.BlobPrefixes {
if *v.Name <= marker {
continue
}
if len(objects)+len(cPrefixes) >= int(maxKeys) {
nextMarker = cPrefixes[len(cPrefixes)-1].Prefix
isTruncated = true
break Pager
}
marker := getString(input.ContinuationToken)
pfx := strings.TrimSuffix(*v.Name, getString(input.Delimiter))
if marker != "" && strings.HasPrefix(marker, pfx) {
continue
}
cPrefixes = append(cPrefixes, types.CommonPrefix{
Prefix: v.Name,
},
})
}
// If StartAfter is specified, filter objects
if input.StartAfter != nil && *input.StartAfter != "" {
startAfter := *input.StartAfter
startIndex := -1
for i, obj := range pageObjects {
if *obj.Key > startAfter {
startIndex = i
break
}
}
if startIndex != -1 {
// Found objects after StartAfter in this page
objects = append(objects, pageObjects[startIndex:]...)
break
} else {
// No objects after StartAfter in this page
// Check if there are more pages to examine
if resp.NextMarker == nil || *resp.NextMarker == "" {
// No more pages, so no objects after StartAfter
break
}
// Continue to next page without adding any objects
continue
}
} else {
// No StartAfter specified, add all objects from this page
objects = append(objects, pageObjects...)
break
}
}
var cPrefixes []types.CommonPrefix
for _, v := range resp.Segment.BlobPrefixes {
cPrefixes = append(cPrefixes, types.CommonPrefix{
Prefix: v.Name,
})
}
var isTruncated bool
var nextMarker *string
// If Azure returned a NextMarker, set it for the next request
if resp.NextMarker != nil && *resp.NextMarker != "" {
nextMarker = resp.NextMarker
isTruncated = true
}
return s3response.ListObjectsV2Result{
+6 -1
View File
@@ -236,7 +236,10 @@ func TestListObjects(s *S3Conf) {
func TestListObjectsV2(s *S3Conf) {
ListObjectsV2_start_after(s)
ListObjectsV2_both_start_after_and_continuation_token(s)
// posix continuation token not compatible with azure
if !s.azureTests {
ListObjectsV2_both_start_after_and_continuation_token(s)
}
ListObjectsV2_start_after_not_in_list(s)
ListObjectsV2_start_after_empty_result(s)
ListObjectsV2_both_delimiter_and_prefix(s)
@@ -246,6 +249,7 @@ func TestListObjectsV2(s *S3Conf) {
ListObjectsV2_exceeding_max_keys(s)
ListObjectsV2_list_all_objs(s)
ListObjectsV2_with_owner(s)
ListObjectsV2_non_truncated_common_prefixes(s)
//TODO: remove the condition after implementing checksums in azure
if !s.azureTests {
ListObjectsV2_with_checksum(s)
@@ -1146,6 +1150,7 @@ func GetIntTests() IntTests {
"ListObjects_nested_dir_file_objs": ListObjects_nested_dir_file_objs,
"ListObjects_check_owner": ListObjects_check_owner,
"ListObjects_non_truncated_common_prefixes": ListObjects_non_truncated_common_prefixes,
"ListObjectsV2_non_truncated_common_prefixes": ListObjectsV2_non_truncated_common_prefixes,
"ListObjects_with_checksum": ListObjects_with_checksum,
"ListObjectsV2_start_after": ListObjectsV2_start_after,
"ListObjectsV2_both_start_after_and_continuation_token": ListObjectsV2_both_start_after_and_continuation_token,
+116 -5
View File
@@ -5142,6 +5142,7 @@ func ListObjects_list_all_objs(s *S3Conf) error {
return err
}
// Test 1: List all objects without pagination
ctx, cancel := context.WithTimeout(context.Background(), shortTimeout)
out, err := s3client.ListObjects(ctx, &s3.ListObjectsInput{
Bucket: &bucket,
@@ -5173,6 +5174,36 @@ func ListObjects_list_all_objs(s *S3Conf) error {
contents, out.Contents)
}
// Test 2: List all objects with pagination using ListObjectsV2
var marker *string
var allObjects []types.Object
maxKeys := int32(2)
for {
ctx, cancel := context.WithTimeout(context.Background(), shortTimeout)
out, err := s3client.ListObjects(ctx, &s3.ListObjectsInput{
Bucket: &bucket,
MaxKeys: &maxKeys,
Marker: marker,
})
cancel()
if err != nil {
return err
}
allObjects = append(allObjects, out.Contents...)
if out.NextMarker == nil || !*out.IsTruncated {
break
}
marker = out.NextMarker
}
if !compareObjects(contents, allObjects) {
return fmt.Errorf("expected the contents to be %v, instead got %v",
contents, allObjects)
}
return nil
})
}
@@ -5585,10 +5616,6 @@ func ListObjectsV2_truncated_common_prefixes(s *S3Conf) error {
return fmt.Errorf("expected the max-keys to be %v, instead got %v",
maxKeys, *out.MaxKeys)
}
if getString(out.NextContinuationToken) != "d3/" {
return fmt.Errorf("expected the NextContinuationToken to be d3/, instead got %v",
getString(out.NextContinuationToken))
}
if getString(out.Delimiter) != delim {
return fmt.Errorf("expected the delimiter to be %v, instead got %v",
delim, getString(out.Delimiter))
@@ -5618,6 +5645,59 @@ func ListObjectsV2_truncated_common_prefixes(s *S3Conf) error {
})
}
func ListObjectsV2_non_truncated_common_prefixes(s *S3Conf) error {
testName := "ListObjectsV2_non_truncated_common_prefixes"
return actionHandler(s, testName, func(s3client *s3.Client, bucket string) error {
_, err := putObjects(s3client, []string{"asdf", "boo/bar", "boo/baz/xyzzy", "cquux/thud", "cquux/bla"}, bucket)
if err != nil {
return err
}
delim, marker, maxKeys := "/", "boo/", int32(1)
ctx, cancel := context.WithTimeout(context.Background(), shortTimeout)
res, err := s3client.ListObjectsV2(ctx, &s3.ListObjectsV2Input{
Bucket: &bucket,
StartAfter: &marker,
Delimiter: &delim,
MaxKeys: &maxKeys,
})
cancel()
if err != nil {
return err
}
if res.IsTruncated == nil {
return fmt.Errorf("expected non-nil istruncated")
}
if *res.IsTruncated {
return fmt.Errorf("expected non-truncated result")
}
if res.MaxKeys == nil {
return fmt.Errorf("expected non nil max-keys")
}
if *res.MaxKeys != maxKeys {
return fmt.Errorf("expected max-keys to be %v, instead got %v",
maxKeys, *res.MaxKeys)
}
if getString(res.Delimiter) != delim {
return fmt.Errorf("expected delimiter to be %v, instead got %v",
delim, getString(res.Delimiter))
}
if len(res.Contents) != 0 {
return fmt.Errorf("expected empty contents, instead got %+v",
res.Contents)
}
cPrefs := []string{"cquux/"}
if !comparePrefixes(cPrefs, res.CommonPrefixes) {
return fmt.Errorf("expected common prefixes to be %v, instead got %+v",
cPrefs, res.CommonPrefixes)
}
return nil
})
}
func ListObjectsV2_all_objs_max_keys(s *S3Conf) error {
testName := "ListObjectsV2_all_objs_max_keys"
return actionHandler(s, testName, func(s3client *s3.Client, bucket string) error {
@@ -5691,11 +5771,12 @@ func ListObjectsV2_exceeding_max_keys(s *S3Conf) error {
func ListObjectsV2_list_all_objs(s *S3Conf) error {
testName := "ListObjectsV2_list_all_objs"
return actionHandler(s, testName, func(s3client *s3.Client, bucket string) error {
contents, err := putObjects(s3client, []string{"bar", "baz", "foo", "obj1", "hell/", "xyzz/quxx"}, bucket)
contents, err := putObjects(s3client, []string{"a", "aa", "aaa", "aaaa", "bar", "baz", "foo", "obj1", "hello/world", "xyzz/quxx"}, bucket)
if err != nil {
return err
}
// Test 1: List all objects without pagination
ctx, cancel := context.WithTimeout(context.Background(), shortTimeout)
out, err := s3client.ListObjectsV2(ctx, &s3.ListObjectsV2Input{
Bucket: &bucket,
@@ -5731,6 +5812,36 @@ func ListObjectsV2_list_all_objs(s *S3Conf) error {
contents, out.Contents)
}
// Test 2: List all objects with pagination using ListObjectsV2
var continuationToken *string
var allObjects []types.Object
maxKeys := int32(2)
for {
ctx, cancel := context.WithTimeout(context.Background(), shortTimeout)
out, err := s3client.ListObjectsV2(ctx, &s3.ListObjectsV2Input{
Bucket: &bucket,
MaxKeys: &maxKeys,
ContinuationToken: continuationToken,
})
cancel()
if err != nil {
return err
}
allObjects = append(allObjects, out.Contents...)
if out.NextContinuationToken == nil || !*out.IsTruncated {
break
}
continuationToken = out.NextContinuationToken
}
if !compareObjects(contents, allObjects) {
return fmt.Errorf("expected the paginated contents to be %v, instead got %v",
contents, allObjects)
}
return nil
})
}