Documentation
¶
Overview ¶
Package object implements repository support for content-addressable objects of arbitrary size.
Index ¶
Constants ¶
This section is empty.
Variables ¶
var ErrObjectNotFound = errors.New("object not found")
ErrObjectNotFound is returned when an object cannot be found.
Functions ¶
Types ¶
type Format ¶
type Format struct {
Splitter string `json:"splitter,omitempty"` // splitter used to break objects into pieces of content
}
Format describes the format of objects in a repository.
type HasObjectID ¶
type HasObjectID interface {
ObjectID() ID
}
HasObjectID exposes the identifier of an object.
type ID ¶
type ID string
ID is an identifier of a repository object. Repository objects can be stored.
- In a single content block, this is the most common case for small objects.
- In a series of content blocks with an indirect block pointing at them (multiple indirections are allowed). This is used for larger files. Object IDs using indirect blocks start with "I"
func Compressed ¶ added in v0.4.0
Compressed returns object ID with 'Z' prefix indicating it's compressed.
func DirectObjectID ¶
DirectObjectID returns direct object ID based on the provided block ID.
func IndirectObjectID ¶
IndirectObjectID returns indirect object ID based on the underlying index object ID.
func (ID) IndexObjectID ¶
IndexObjectID returns the object ID of the underlying index object.
type Manager ¶
type Manager struct {
Format Format
// contains filtered or unexported fields
}
Manager implements a content-addressable storage on top of blob storage.
func NewObjectManager ¶
NewObjectManager creates an ObjectManager with the specified content manager and format.
func (*Manager) Concatenate ¶ added in v0.7.0
Concatenate creates an object that's a result of concatenation of other objects. This is more efficient than reading and rewriting the objects because Concatenate can efficiently merge index entries without reading the underlying contents.
This function exists primarily to facilitate efficient parallel uploads of very large files (>1GB). Due to bottleneck of splitting which is inherently sequential, we can only one use CPU core for each Writer, which limits throughput.
For example when uploading a 100 GB file it is beneficial to independently upload sections of [0..25GB), [25..50GB), [50GB..75GB) and [75GB..100GB) and concatenate them together as this allows us to run four splitters in parallel utilizing more CPU cores. Because some split points now start at fixed bounaries and not content-specific, this causes some slight loss of deduplication at concatenation points (typically 1-2 contents, usually <10MB), so this method should only be used for very large files where this overhead is relatively small.
type Reader ¶
Reader allows reading, seeking, getting the length of and closing of a repository object.
type Writer ¶
type Writer interface {
io.WriteCloser
// Checkpoint returns ID of an object consisting of all contents written to storage so far.
// This may not include some data buffered in the writer.
// In case nothing has been written yet, returns empty object ID.
Checkpoint() (ID, error)
// Result returns object ID representing all bytes written to the writer.
Result() (ID, error)
}
Writer allows writing content to the storage and supports automatic deduplication and encryption of written data.
type WriterOptions ¶
type WriterOptions struct {
Description string
Prefix content.ID // empty string or a single-character ('g'..'z')
Compressor compression.Name
AsyncWrites int // allow up to N content writes to be asynchronous
}
WriterOptions can be passed to Repository.NewWriter().
Source Files