Vertica::SourceFactory Class Reference

Produces and initializes UDSources to process input data sources. More...

Inheritance diagram for Vertica::SourceFactory:

Collaboration diagram for Vertica::SourceFactory:

Public Member Functions
virtual ssize_t	getDesiredThreads (ServerInterface &srvInterface, ExecutorPlanContext &planCtxt) __override__
UDXFactory::UDXType	getUDXFactoryType () __override__
virtual bool	isSourceApportionable ()
virtual void	plan (ServerInterface &srvInterface, NodeSpecifyingPlanContext &planCtxt) __override__
virtual SourceIterator *	prepare (ServerInterface &srvInterface, NodeSpecifyingPlanContext &planCtxt)
virtual SourceIterator *	prepareExecutor (ServerInterface &srvInterface, ExecutorPlanContext &planCtxt)
virtual std::vector< UDSource * >	prepareUDSources (ServerInterface &srvInterface, NodeSpecifyingPlanContext &planCtxt)
virtual std::vector< UDSource * >	prepareUDSourcesExecutor (ServerInterface &srvInterface, ExecutorPlanContext &planCtxt)

Detailed Description

Produces and initializes UDSources to process input data sources.

Subclassing SourceFactory is the preferred way to implement a factory corresponding to a UDSource. If you need more flexibility, implement both IterativeSourceFactory and SourceIterator. (See the documentation of IterativeSourceFactory.)

Factories should be registered using the RegisterFactory() macro, defined in Vertica.h.

Member Function Documentation

virtual ssize_t Vertica::SourceFactory::getDesiredThreads ( ServerInterface &  srvInterface, ExecutorPlanContext &  planCtxt  )

inlinevirtual

Request the number of threads that the SourceFactory wants to use. This function is called on each node before prepareUDSources().

'planCtxt' contains the same parameters stored by plan(). It exposes the maximum number of allowed threads, via 'getMaxAllowedThreads()', and also the actual number of threads, via 'getLoadConcurrency()'.

The default implementation of this invokes prepare(), saves the result in the ExecutorPlanContext, and returns the number of sources as the number of desired threads.

Reimplemented from Vertica::IterativeSourceFactory.

virtual void Vertica::SourceFactory::plan ( ServerInterface &  srvInterface, NodeSpecifyingPlanContext &  planCtxt  )

inlinevirtual

Execute any planning logic required at query plan time. This method is run once per query, during query initialization. Its job is to perform parameter validation, and to modify the set of nodes that the COPY statement will run on.

plan() runs exactly once per query, on the initiator node. If it throws an exception, the query will not proceed; it will be aborted prior to distributing the query to the other nodes and running prepare().

Parameters

srvInterface	Interface to server operations and functionality, including (not-per-column) parameter lookup
planCtxt	Context for storing and retrieving arbitrary data, for use just by this instance of this query. The same context is shared with plan(). Also provides functionality for specifying which nodes this query will run on.

Reimplemented from Vertica::IterativeSourceFactory.

virtual SourceIterator* Vertica::IterativeSourceFactory::prepareExecutor ( ServerInterface &  srvInterface, ExecutorPlanContext &  planCtxt  )

inlinevirtualinherited

Prepare this SourceFactory to start creating sources. This function will be called on each node, prior to the Load operator starting to execute and prior to any other virtual functions on this class being called.

'planCtxt' contains the same data that was placed there by the plan() and getDesiredThreads() methods.

virtual std::vector<UDSource*> Vertica::SourceFactory::prepareUDSourcesExecutor ( ServerInterface &  srvInterface, ExecutorPlanContext &  planCtxt  )

inlinevirtual

Create UDSources. This function will be called on each node, prior to the Load operator starting to execute and prior to any other virtual functions on this class being called.

'planCtxt' contains the same data that was placed there by the plan() and getDesiredThreads() methods. It also exposes the maximum execution parallelism with 'getMaxAllowedThreads()' and the actual load concurrency with 'getLoadConcurrency()'.

Unlike the standard SourceFactory, this method directly instantiates all of its UDSources, and returns a vector of them. This requires that all UDSources be resident in memory for the duration of the query, which is fine in the common case but which may not be acceptable for some resource-intensive UDSources.

Parameters

srvInterface	Interface to server operations and functionality, including (not-per-column) parameter lookup
planCtxt	Context for storing and retrieving arbitrary data, for use just by this instance of this query. The same context is shared with plan() and getDesiredThreads(). Also provides functionality for determining which nodes this query is running on and getting the number of available threads.

Returns

A vector of UDSources to use for this query. Sources will be loaded in a pooled manner, several at a time.