- Type Parameters:
T- type of emitted objects
public final class SourceBuilder.TimestampedStream<T> extends Object
Modifier and Type Method Description
()Builds and returns the timestamped stream source.
FunctionEx<? super C,? extends S> createSnapshotFn)(
ConsumerEx<? super C> pDestroyFn)(Sets the function that Jet will call when it is done cleaning up after an execution.
(int preferredLocalParallelism)Declares that you're creating a distributed source.
BiConsumerEx<? super C,? super SourceBuilder.TimestampedSourceBuffer<T_NEW>> fillBufferFn)(Sets the function that Jet will call whenever it needs more data from your source.
fillBufferFn@Nonnull public <T_NEW> SourceBuilder.TimestampedStream<T_NEW> fillBufferFn(@Nonnull BiConsumerEx<? super C,? super SourceBuilder.TimestampedSourceBuffer<T_NEW>> fillBufferFn)Sets the function that Jet will call whenever it needs more data from your source. The function receives the context object obtained from
createFnand Jet's buffer object. It should add some items to the buffer, ideally those it can produce without making any blocking calls. The buffer's
add()method takes two arguments: the item and the timestamp in milliseconds.
On any given invocation the function may also choose not to add any items. Jet will automatically employ an exponential backoff strategy to avoid calling your function in a tight loop, if the previous call didn't add any items to the buffer.
SourceBuilder.TimestampedSourceBufferisn't thread-safe, you shouldn't pass it to other threads. For example, you shouldn't add to it in a callback of an asynchronous operation.
- Type Parameters:
T_NEW- type of the emitted items
fillBufferFn- function that fills the buffer with source data. It must be stateless.
- this builder with the item type reset to the one inferred from
destroyFn@Nonnull public SourceBuilder.TimestampedStream<T> destroyFn(@Nonnull ConsumerEx<? super C> pDestroyFn)Sets the function that Jet will call when it is done cleaning up after an execution. It gives you the opportunity to release any resources that your context object may be holding. Jet also calls this function when the user cancels or restarts the job.
The function must be stateless.
distributedDeclares that you're creating a distributed source. On each member of the cluster Jet will create as many processors as you specify with the
preferredLocalParallelismparameter. If you call this, you must ensure that all the source processors are coordinated and not emitting duplicated data. The
processorContext.globalProcessorIndex(). Jet calls
createFnexactly once with each
globalProcessorIndexfrom 0 to
totalParallelism - 1and you can use this to make all the instances agree on which part of the data to emit.
If you don't call this method, there will be only one processor instance running on an arbitrary member.
preferredLocalParallelism- the requested number of processors on each cluster member
buildBuilds and returns the timestamped stream source.