Skip to content

Latest commit

 

History

History
49 lines (35 loc) · 2.16 KB

README.md

File metadata and controls

49 lines (35 loc) · 2.16 KB
Raw

go-stream

This library is a framework for stream processing analysis. It is meant to be used as a library for go programs that need to do stream processing of large volumes of data.

It is made up of a graph connecting a source to 1 or more operators, terminating at a sink. Operators pass data from one to another with go channels. An example graph to encode objects to snappy is:

var from *util.MemoryBuffer
// fill up from

var to *util.MemoryBuffer

ch := stream.NewOrderedChain()
ch.Add(source.NewNextReaderSource(from))
timingOp, _, dur := timing.NewTimingOp()
ch.Add(timingOp)
ch.Add(compress.NewSnappyEncodeOp())
ch.Add(sink.NewWriterSink(to))

ch.Start()

log.Printf("RES: Compress Snappy.\t\tRatio %v", float64(to.ByteSize())/float64(from.ByteSize()))
log.Printf("RES: Compress Snappy.\t\tBuffered Items: %d\tItem: %v\ttook: %v\trate: %d\tsize: %E\tsize/item: %E", to.Len(), *counter, *dur, int(    float64(*counter)/(*dur).Seconds()), float64(to.ByteSize()), float64(to.ByteSize())/float64(*counter))

Operators are the main components of a chain. They process tuples to produce results. Sources are operators with no output. Sinks are operators with no input. Operators implement stream.Operator. If it takes input implements stream.In; if it produces output implements stream.Out.

Mappers give a simple way to implement operators. mapper.NewOp() takes a function of the form func(input stream.Object, out Outputer) which processes the input and outputs it to the Outputer object. Mappers are automatically parallelized. Generators give a way to give mappers thread-local storage through closures. You can also give mappers special functionality after they have finished processing the last tuple.

You can also split the data of a chain into other chains. stream.Fanout takes input and copies them to N other chains. Distributor takes input and puts it onto 1 of N chains according to a mapping function.

Chains can be ordered or unordered. Ordered chains preserve the order of tuples from input to output (although the operators still use parallelism).
Installing:

go get "github.com/cevian/go-stream/stream

Compiling: go build

Testing: go test