Snappy compression and decompression Pekko Steams Flows.
Uses Snappy's framing format.
In your build.sbt:
libraryDependencies += "me.maciejb.snappyflows" %% "snappy-flows" % "0.3.0"Snappy flows are available for Scala 2.11, 2.12 and 2.13. We use a recent Pekko 1.0.1 branch version
Previous releases:
| Snappy flows | Akka | Scala |
|---|---|---|
0.2.0 - 0.2.1 |
Akka >2.4.2 | 2.11 |
0.1.3 - 0.1.5 |
Akka >2.4.2 | 2.11 |
0.1.2 |
Akka 2.4.2-RC3 | 2.11 |
0.1.1 |
Akka Streams 2.0 | 2.11 |
<= 0.1.0 |
Akka Streams 2.0-M2 | 2.11 |
Sync and async versions of the flows are available:
import me.maciejb.snappyflows.SnappyFlows
val Parallelism = 4
// Let's take a source we want to decompress
val sourceWithCompressedData: Source[ByteString] = Source(...)
sourceWithCompressedData.via(SnappyFlows.decompress())
// or
sourceWithCompressedData.via(SnappyFlows.decompressAsync(Parallelism))
// Now, one we want to compress
val sourceWithRawBytes: Source[ByteString] = Source(...)
sourceWithRawBytes.via(SnappyFlows.compress())
// or
sourceWithRawBytes.via(SnappyFlows.compressAsync(Parallelism))Initial benchmarks show that the non-async Snappy Flows achieve ~90% of performance of
SnappyFramedInputStream and SnappyFramedOutputStream. async versions provide very good speedup giving 550-600MiB/s throughput on the test data on a a late 2013 15’ MBP with Intel(R) Core(TM) i7-4960HQ CPU @ 2.60GHz, 16GiB.
For details refer to benchmarks/README.md.
