Documentation ¶
Index ¶
- Variables
- func BytesToU64(data []byte) uint64
- func IsZeroKey(key Key) bool
- func NewDpaChunkStore(localStore, netStore ChunkStore) *dpaChunkStore
- func U64ToBytes(val uint64) []byte
- type Chunk
- type ChunkStore
- type Chunker
- type ChunkerParams
- type CloudStore
- type DPA
- type DbStore
- type DbSyncState
- type Hasher
- type Joiner
- type Key
- type LDBDatabase
- func (self *LDBDatabase) Close()
- func (self *LDBDatabase) Delete(key []byte) error
- func (self *LDBDatabase) Get(key []byte) ([]byte, error)
- func (self *LDBDatabase) LastKnownTD() []byte
- func (self *LDBDatabase) NewIterator() iterator.Iterator
- func (self *LDBDatabase) Put(key []byte, value []byte)
- func (self *LDBDatabase) Write(batch *leveldb.Batch) error
- type LazyChunkReader
- type LazySectionReader
- type LazyTestSectionReader
- type LocalStore
- type MemStore
- type NetStore
- type Node
- type Peer
- type PyramidChunker
- type RequestStatus
- type Splitter
- type StoreParams
- type Task
- type Tree
- type TreeChunker
Constants ¶
This section is empty.
Variables ¶
var ZeroKey = Key(common.Hash{}.Bytes())
Functions ¶
func BytesToU64 ¶
func NewDpaChunkStore ¶ added in v1.5.0
func NewDpaChunkStore(localStore, netStore ChunkStore) *dpaChunkStore
func U64ToBytes ¶
Types ¶
type Chunk ¶
type Chunk struct { Key Key // always SData []byte // nil if request, to be supplied by dpa Size int64 // size of the data covered by the subtree encoded in this chunk Source Peer // peer C chan bool // to signal data delivery by the dpa Req *RequestStatus // request Status needed by netStore // contains filtered or unexported fields }
Chunk also serves as a request object passed to ChunkStores in case it is a retrieval request, Data is nil and Size is 0 Note that Size is not the size of the data chunk, which is Data.Size() but the size of the subtree encoded in the chunk 0 if request, to be supplied by the dpa
func NewChunk ¶
func NewChunk(key Key, rs *RequestStatus) *Chunk
type ChunkStore ¶
type ChunkStore interface { Put(*Chunk) // effectively there is no error even if there is an error Get(Key) (*Chunk, error) Close() }
The ChunkStore interface is implemented by :
- MemStore: a memory cache - DbStore: local disk/db store - LocalStore: a combination (sequence of) memStore and dbStore - NetStore: cloud storage abstraction layer - DPA: local requests for swarm storage and retrieval
type ChunkerParams ¶
func NewChunkerParams ¶ added in v1.5.0
func NewChunkerParams() *ChunkerParams
type CloudStore ¶ added in v1.5.0
backend engine for cloud store It can be aggregate dispatching to several parallel implementations: bzz/network/forwarder. forwarder or IPFS or IPΞS
type DPA ¶ added in v1.5.0
type DPA struct { ChunkStore Chunker Chunker // contains filtered or unexported fields }
func NewDPA ¶ added in v1.5.0
func NewDPA(store ChunkStore, params *ChunkerParams) *DPA
func NewLocalDPA ¶ added in v1.5.0
for testing locally
func (*DPA) Retrieve ¶ added in v1.5.0
func (self *DPA) Retrieve(key Key) LazySectionReader
Public API. Main entry point for document retrieval directly. Used by the FS-aware API and httpaccess Chunk retrieval blocks on netStore requests with a timeout so reader will report error if retrieval of chunks within requested range time out.
type DbStore ¶ added in v1.5.0
type DbStore struct {
// contains filtered or unexported fields
}
func NewDbStore ¶ added in v1.5.0
func (*DbStore) NewSyncIterator ¶ added in v1.5.0
func (self *DbStore) NewSyncIterator(state DbSyncState) (si *dbSyncIterator, err error)
initialises a sync iterator from a syncToken (passed in with the handshake)
type DbSyncState ¶ added in v1.5.0
describes a section of the DbStore representing the unsynced
domain relevant to a peer Start - Stop designate a continuous area Keys in an address space typically the addresses closer to us than to the peer but not closer another closer peer in between From - To designates a time interval typically from the last disconnect till the latest connection (real time traffic is relayed)
type Hasher ¶
func MakeHashFunc ¶
type Joiner ¶ added in v1.5.0
type Joiner interface { /* Join reconstructs original content based on a root key. When joining, the caller gets returned a Lazy SectionReader, which is seekable and implements on-demand fetching of chunks as and where it is read. New chunks to retrieve are coming to caller via the Chunk channel, which the caller provides. If an error is encountered during joining, it appears as a reader error. The SectionReader. As a result, partial reads from a document are possible even if other parts are corrupt or lost. The chunks are not meant to be validated by the chunker when joining. This is because it is left to the DPA to decide which sources are trusted. */ Join(key Key, chunkC chan *Chunk) LazySectionReader }
type Key ¶ added in v1.5.0
type Key []byte
func (Key) MarshalJSON ¶ added in v1.5.0
func (*Key) UnmarshalJSON ¶ added in v1.5.0
type LDBDatabase ¶
type LDBDatabase struct {
// contains filtered or unexported fields
}
func NewLDBDatabase ¶
func NewLDBDatabase(file string) (*LDBDatabase, error)
func (*LDBDatabase) Close ¶
func (self *LDBDatabase) Close()
func (*LDBDatabase) Delete ¶
func (self *LDBDatabase) Delete(key []byte) error
func (*LDBDatabase) LastKnownTD ¶
func (self *LDBDatabase) LastKnownTD() []byte
func (*LDBDatabase) NewIterator ¶
func (self *LDBDatabase) NewIterator() iterator.Iterator
func (*LDBDatabase) Put ¶
func (self *LDBDatabase) Put(key []byte, value []byte)
type LazyChunkReader ¶
type LazyChunkReader struct {
// contains filtered or unexported fields
}
LazyChunkReader implements LazySectionReader
func (*LazyChunkReader) Read ¶
func (self *LazyChunkReader) Read(b []byte) (read int, err error)
Read keeps a cursor so cannot be called simulateously, see ReadAt
func (*LazyChunkReader) ReadAt ¶
func (self *LazyChunkReader) ReadAt(b []byte, off int64) (read int, err error)
read at can be called numerous times concurrent reads are allowed Size() needs to be called synchronously on the LazyChunkReader first
type LazySectionReader ¶
Size, Seek, Read, ReadAt
type LazyTestSectionReader ¶
type LazyTestSectionReader struct {
*io.SectionReader
}
type LocalStore ¶
type LocalStore struct { DbStore ChunkStore // contains filtered or unexported fields }
LocalStore is a combination of inmemory db over a disk persisted db implements a Get/Put with fallback (caching) logic using any 2 ChunkStores
func NewLocalStore ¶
func NewLocalStore(hash Hasher, params *StoreParams) (*LocalStore, error)
This constructor uses MemStore and DbStore as components
func (*LocalStore) Get ¶
func (self *LocalStore) Get(key Key) (chunk *Chunk, err error)
Get(chunk *Chunk) looks up a chunk in the local stores This method is blocking until the chunk is retrieved so additional timeout may be needed to wrap this call if ChunkStores are remote and can have long latency
func (*LocalStore) Put ¶
func (self *LocalStore) Put(chunk *Chunk)
LocalStore is itself a chunk store unsafe, in that the data is not integrity checked
type MemStore ¶
type MemStore struct {
// contains filtered or unexported fields
}
func NewMemStore ¶
type NetStore ¶
type NetStore struct {
// contains filtered or unexported fields
}
NetStore is a cloud storage access abstaction layer for swarm it contains the shared logic of network served chunk store/retrieval requests both local (coming from DPA api) and remote (coming from peers via bzz protocol) it implements the ChunkStore interface and embeds LocalStore
It is called by the bzz protocol instances via Depo (the store/retrieve request handler) a protocol instance is running on each peer, so this is heavily parallelised. NetStore falls back to a backend (CloudStorage interface) implemented by bzz/network/forwarder. forwarder or IPFS or IPΞS
func NewNetStore ¶
func NewNetStore(hash Hasher, lstore *LocalStore, cloud CloudStore, params *StoreParams) *NetStore
netstore contructor, takes path argument that is used to initialise dbStore, the persistent (disk) storage component of LocalStore the second argument is the hive, the connection/logistics manager for the node
type Peer ¶
type Peer interface{}
Peer is the recorded as Source on the chunk should probably not be here? but network should wrap chunk object
type PyramidChunker ¶
type PyramidChunker struct {
// contains filtered or unexported fields
}
func NewPyramidChunker ¶ added in v1.5.0
func NewPyramidChunker(params *ChunkerParams) (self *PyramidChunker)
type RequestStatus ¶ added in v1.5.0
each chunk when first requested opens a record associated with the request next time a request for the same chunk arrives, this record is updated this request status keeps track of the request ID-s as well as the requesting peers and has a channel that is closed when the chunk is retrieved. Multiple local callers can wait on this channel (or combined with a timeout, block with a select).
type Splitter ¶ added in v1.5.0
type Splitter interface { /* When splitting, data is given as a SectionReader, and the key is a hashSize long byte slice (Key), the root hash of the entire content will fill this once processing finishes. New chunks to store are coming to caller via the chunk storage channel, which the caller provides. wg is a Waitgroup (can be nil) that can be used to block until the local storage finishes The caller gets returned an error channel, if an error is encountered during splitting, it is fed to errC error channel. A closed error signals process completion at which point the key can be considered final if there were no errors. */ Split(io.Reader, int64, chan *Chunk, *sync.WaitGroup, *sync.WaitGroup) (Key, error) }
Chunker is the interface to a component that is responsible for disassembling and assembling larger data and indended to be the dependency of a DPA storage system with fixed maximum chunksize.
It relies on the underlying chunking model.
When calling Split, the caller provides a channel (chan *Chunk) on which it receives chunks to store. The DPA delegates to storage layers (implementing ChunkStore interface).
Split returns an error channel, which the caller can monitor. After getting notified that all the data has been split (the error channel is closed), the caller can safely read or save the root key. Optionally it times out if not all chunks get stored or not the entire stream of data has been processed. By inspecting the errc channel the caller can check if any explicit errors (typically IO read/write failures) occurred during splitting.
When calling Join with a root key, the caller gets returned a seekable lazy reader. The caller again provides a channel on which the caller receives placeholder chunks with missing data. The DPA is supposed to forward this to the chunk stores and notify the chunker if the data has been delivered (i.e. retrieved from memory cache, disk-persisted db or cloud based swarm delivery). As the seekable reader is used, the chunker then puts these together the relevant parts on demand.
type StoreParams ¶
func NewStoreParams ¶
func NewStoreParams(path string) (self *StoreParams)
type TreeChunker ¶
type TreeChunker struct {
// contains filtered or unexported fields
}
func NewTreeChunker ¶ added in v1.5.0
func NewTreeChunker(params *ChunkerParams) (self *TreeChunker)
func (*TreeChunker) Join ¶
func (self *TreeChunker) Join(key Key, chunkC chan *Chunk) LazySectionReader
implements the Joiner interface