Documentation ¶
Index ¶
- Variables
- func NewByteCountingLinkSystem(lsys *linking.LinkSystem, bytesWritten func(count uint64)) *linking.LinkSystem
- type InProgressCids
- type IndexerRouting
- type MultiBlockstore
- func (mbs *MultiBlockstore) AddLinkSystem(id types.RetrievalID, lsys *linking.LinkSystem) error
- func (mbs *MultiBlockstore) AllKeysChan(ctx context.Context) (<-chan cid.Cid, error)
- func (mbs *MultiBlockstore) DeleteBlock(ctx context.Context, c cid.Cid) error
- func (mbs *MultiBlockstore) Get(ctx context.Context, c cid.Cid) (blocks.Block, error)
- func (mbs *MultiBlockstore) GetSize(ctx context.Context, c cid.Cid) (int, error)
- func (mbs *MultiBlockstore) Has(ctx context.Context, c cid.Cid) (bool, error)
- func (mbs *MultiBlockstore) HashOnRead(enabled bool)
- func (mbs *MultiBlockstore) Put(ctx context.Context, blk blocks.Block) error
- func (mbs *MultiBlockstore) PutMany(ctx context.Context, blks []blocks.Block) error
- func (mbs *MultiBlockstore) RemoveLinkSystem(id types.RetrievalID)
- type PreloadCachingStorage
- type PreloadStats
Constants ¶
This section is empty.
Variables ¶
var ErrAlreadyRegisterd = errors.New("already registered")
ErrAlreadyRegistered means something has already been registered for a retrieval id
var ErrNotRegistered = errors.New("not registered")
ErrAlreadyRegistered means there is nothing registered for a retrieval id
var ErrNotSupported = errors.New("not supported")
ErrNotSupported indicates an operation not supported by the MultiBlockstore
Functions ¶
func NewByteCountingLinkSystem ¶
func NewByteCountingLinkSystem(lsys *linking.LinkSystem, bytesWritten func(count uint64)) *linking.LinkSystem
Types ¶
type InProgressCids ¶
type InProgressCids struct {
// contains filtered or unexported fields
}
func NewInProgressCids ¶
func NewInProgressCids() *InProgressCids
func (*InProgressCids) Dec ¶
func (ipc *InProgressCids) Dec(c cid.Cid, retrievalID types.RetrievalID)
func (*InProgressCids) Get ¶
func (ipc *InProgressCids) Get(c cid.Cid) []types.RetrievalID
func (*InProgressCids) Inc ¶
func (ipc *InProgressCids) Inc(c cid.Cid, retrievalID types.RetrievalID)
type IndexerRouting ¶
type IndexerRouting struct { routinghelpers.Null // contains filtered or unexported fields }
IndexerRouting provides an interface that satisfies routing.Routing but only returns provider records based on a preset set of providers read from the context key. Bitswap will potentially make multiple FindProvidersAsync requests, and the cid passed will not always be the root As a result, we have to rely on the retrieval id within a context key Also while there is a delegated routing client that talks to the indexer, we use this cause we run it on top of the processing we're doing at a higher level with multiprotocol filtering
func NewIndexerRouting ¶
func NewIndexerRouting(toRetrievalID func(cid.Cid) []types.RetrievalID) *IndexerRouting
NewIndexerRouting makes a new indexer routing instance
func (*IndexerRouting) AddProviders ¶
func (ir *IndexerRouting) AddProviders(retrievalID types.RetrievalID, providers []types.RetrievalCandidate)
AddProviders adds provider records to the total list for a given retrieval id
func (*IndexerRouting) FindProvidersAsync ¶
func (ir *IndexerRouting) FindProvidersAsync(ctx context.Context, c cid.Cid, max int) <-chan peer.AddrInfo
FindProvidersAsync returns providers based on the retrieval id in a context key It returns a channel with up to `max` providers, keeping the others around for a future call TODO: there is a slight risk that go-bitswap, which dedups requests by CID across multiple sessions, could accidentally read the wrong retrieval id if two retrievals were running at the same time. Not sure how much of a risk this really is, cause when requests are deduped, both calls still receive the results. See go-bitswap ProviderQueryManager for more specifics
func (*IndexerRouting) RemoveProviders ¶
func (ir *IndexerRouting) RemoveProviders(retrievalID types.RetrievalID)
RemoveProviders removes all provider records for a given retrieval id
type MultiBlockstore ¶
type MultiBlockstore struct {
// contains filtered or unexported fields
}
MultiBlockstore creates a blockstore based on one or more linkystems, extracting the target linksystem for each request from the retrieval id context key
func NewMultiblockstore ¶
func NewMultiblockstore() *MultiBlockstore
NewMultiblockstore returns a new MultiBlockstore
func (*MultiBlockstore) AddLinkSystem ¶
func (mbs *MultiBlockstore) AddLinkSystem(id types.RetrievalID, lsys *linking.LinkSystem) error
AddLinkSystem registers a linksystem to use for a given retrieval id
func (*MultiBlockstore) AllKeysChan ¶
func (mbs *MultiBlockstore) AllKeysChan(ctx context.Context) (<-chan cid.Cid, error)
AllKeysChan is unsupported
func (*MultiBlockstore) DeleteBlock ¶
func (mbs *MultiBlockstore) DeleteBlock(ctx context.Context, c cid.Cid) error
DeleteBlock is not supported
func (*MultiBlockstore) Get ¶
Get returns a block only if the given ctx contains a retrieval ID as a value that references a known linksystem. If it does, it uses that linksystem to load the block
func (*MultiBlockstore) GetSize ¶
func (mbs *MultiBlockstore) GetSize(ctx context.Context, c cid.Cid) (int, error)
GetSize is unsupported
func (*MultiBlockstore) Has ¶
func (mbs *MultiBlockstore) Has(ctx context.Context, c cid.Cid) (bool, error)
Has is not supported
func (*MultiBlockstore) HashOnRead ¶
func (mbs *MultiBlockstore) HashOnRead(enabled bool)
HashOnRead is unsupported
func (*MultiBlockstore) Put ¶
Put writes a block only if the given ctx contains a retrieval ID as a value that references a known linksystem. If it does, it uses that linksystem to save the block
func (*MultiBlockstore) PutMany ¶
PutMany puts a slice of blocks at the same time, with the same rules as Put
func (*MultiBlockstore) RemoveLinkSystem ¶
func (mbs *MultiBlockstore) RemoveLinkSystem(id types.RetrievalID)
RemoveLinkSystem unregisters the link system for a given retrieval id
type PreloadCachingStorage ¶ added in v0.8.0
type PreloadCachingStorage struct { TraversalLinkSystem *linking.LinkSystem BitswapLinkSystem *linking.LinkSystem // contains filtered or unexported fields }
func NewPreloadCachingStorage ¶ added in v0.8.0
func NewPreloadCachingStorage( parentLinkSystem linking.LinkSystem, cacheLinkSystem linking.LinkSystem, fetcher linking.BlockReadOpener, workGroup groupworkpool.Group, ) (*PreloadCachingStorage, error)
NewPreloadCachingStorage creates a new PreloadCachingStorage.
The parentLinkSystem is used directly by the traversal for both reads (existing blocks, or blocks already traversed and stored) and writes (new blocks discovered during traversal). Writes will be properly ordered according to the traversal.
The cacheLinkSystem is used by the preloader as a temporary space for preloaded blocks. Writes are not properly ordered, only according to the preloader's ability to fetch blocks in its queue. When a block is requested by the traversal and it is already in the cacheLinkSystem, it is copied to the parentLinkSystem.
The fetcher is used by both the preloader and the loader to fetch blocks. It should be able to fetch blocks in a thread-safe manner.
func (*PreloadCachingStorage) GetStats ¶ added in v0.8.0
func (cs *PreloadCachingStorage) GetStats() PreloadStats
GetStats returns the current stats for the PreloadCachingStorage.
func (*PreloadCachingStorage) Loader ¶ added in v0.8.0
func (cs *PreloadCachingStorage) Loader(linkCtx linking.LinkContext, link ipld.Link) (io.Reader, error)
Loader is compatible with go-ipld-prime's StorageReadOpener. It is not intended to be used directly but is wired up as the StorageReadOpener for the TraversalLinkSystem.
func (*PreloadCachingStorage) Preloader ¶ added in v0.8.0
func (cs *PreloadCachingStorage) Preloader(preloadCtx preload.PreloadContext, link preload.Link)
Preloader is compatible with go-ipld-prime's linking/preload.Loader and should be provided to a traversal Config as the Preloader field.
type PreloadStats ¶ added in v0.8.0
type PreloadStats struct { // LoadCount is the number of times the Loader was called LoadCount int // ActivePreloads is the number of links currently being preloaded ActivePreloads int // number of links currently being preloaded // NotFound is the number of links that were a decisive "not found" (by the // preloader) NotFound int // PreloadedHits is the number of times a link was loaded from the cache // (already full preloaded) PreloadedHits int // PreloadingHits is the number of times a link was loaded that was in the // queue to be preloaded but hadn't yet been fully loaded. PreloadingHits int // PreloadMisses is the number of times a link was not found in the cache or // queued to be preloaded. This should only happen once per traversal (the // root). PreloadMisses int }
func (PreloadStats) PreloadedPercent ¶ added in v0.8.0
func (s PreloadStats) PreloadedPercent() uint64
PreloadedPercent returns the percentage of loads that were hits in the preloaded cache. This is a good indicator of how much of the traversal was able to be satisfied by the preloaded cache, however it is not able to capture the number of blocks that were actively being loaded when they were requested.
func (PreloadStats) PreloadingPercent ¶ added in v0.8.0
func (s PreloadStats) PreloadingPercent() uint64
PreloadingHitRate returns the fraction of loads that were hits in the queue of links to be preloaded. This means that the block was flagged as being needed by the preloader, but had not yet been preloaded, or was in the process of being preloaded when it was required by the traversal.
PreloadingHitRate() + PreloadedHitRate() should be close to a value of 1.0 (not exactly as the first block will be a complete preloader miss), together they provide a measure of the performance of the preloader for the traversal.
func (PreloadStats) Print ¶ added in v0.8.0
func (s PreloadStats) Print()
Print prints the stats to stdout