Cache

default static final long UID_UNSET = -1

Returned by getUid if initialization failed before the unique identifier was read or generated.

abstract NavigableSet<CacheSpan> addListener(String key, Cache.Listener listener)

Registers a listener to listen for changes to a given resource.

No guarantees are made about the thread or threads on which the listener is called, but it is guaranteed that listener methods will be called in a serial fashion (i.e. one at a time) and in the same order as events occurred.

@WorkerThread
abstract void applyContentMetadataMutations(
    String key,
    ContentMetadataMutations mutations
)

Applies mutations to the ContentMetadata for the given resource. A new is added if there isn't one already for the resource.

This method may be slow and shouldn't normally be called on the main thread.

@WorkerThread
abstract void commitFile(File file, long length)

Commits a file into the cache. Must only be called when holding a corresponding hole obtained from startReadWrite.

This method may be slow and shouldn't normally be called on the main thread.

abstract long getCacheSpace()

Returns the total disk space in bytes used by the cache.

abstract long getCachedBytes(String key, long position, long length)

Returns the total number of cached bytes between position (inclusive) and (position + length) (exclusive) of a resource.

abstract long getCachedLength(String key, long position, long length)

Returns the length of continuously cached data starting from position, up to a maximum of maxLength, of a resource. If position isn't cached then -holeLength is returned, where holeLength is the length of continuously uncached data starting from position, up to a maximum of maxLength.

abstract NavigableSet<CacheSpan> getCachedSpans(String key)

Returns the cached spans for a given resource.

abstract ContentMetadata getContentMetadata(String key)

Returns a ContentMetadata for the given resource.

abstract Set<String> getKeys()

Returns the cache keys of all of the resources that are at least partially cached.

abstract long getUid()

Returns a non-negative unique identifier for the cache, or UID_UNSET if initialization failed before the unique identifier was determined.

Implementations are expected to generate and store the unique identifier alongside the cached content. If the location of the cache is deleted or swapped, it is expected that a new unique identifier will be generated when the cache is recreated.

abstract boolean isCached(String key, long position, long length)

Returns whether the specified range of data in a resource is fully cached.

@WorkerThread
abstract void release()

Releases the cache. This method must be called when the cache is no longer required. The cache must not be used after calling this method.

This method may be slow and shouldn't normally be called on the main thread.

abstract void releaseHoleSpan(CacheSpan holeSpan)

Releases a CacheSpan obtained from startReadWrite which corresponded to a hole in the cache.

abstract void removeListener(String key, Cache.Listener listener)

Unregisters a listener.

@WorkerThread
abstract void removeResource(String key)

Removes all CacheSpans for a resource, deleting the underlying files.

@WorkerThread
abstract void removeSpan(CacheSpan span)

Removes a cached CacheSpan from the cache, deleting the underlying file.

This method may be slow and shouldn't normally be called on the main thread.

@WorkerThread
abstract File startFile(String key, long position, long length)

Obtains a cache file into which data can be written. Must only be called when holding a corresponding hole CacheSpan obtained from startReadWrite.

This method may be slow and shouldn't normally be called on the main thread.

@WorkerThread
abstract CacheSpan startReadWrite(String key, long position, long length)

A caller should invoke this method when they require data starting from a given position in a given resource.

If there is a cache entry that overlaps the position, then the returned CacheSpan defines the file in which the data is stored. isCached is true. The caller may read from the cache file, but does not acquire any locks.

If there is no cache entry overlapping position, then the returned CacheSpan defines a hole in the cache starting at position into which the caller may write as it obtains the data from some other source. The returned CacheSpan serves as a lock. Whilst the caller holds the lock it may write data into the hole. It may split data into multiple files. When the caller has finished writing a file it should commit it to the cache by calling commitFile. When the caller has finished writing, it must release the lock by calling releaseHoleSpan.

This method may be slow and shouldn't normally be called on the main thread.

@WorkerThread
abstract @Nullable CacheSpan startReadWriteNonBlocking(String key, long position, long length)

Same as startReadWrite. However, if the cache entry is locked, then instead of blocking, this method will return null as the CacheSpan.

This method may be slow and shouldn't normally be called on the main thread.