Breeze-Bokeh
Graphs for breeze. Totally experimental at this point.
How to use
$ sbt
> run
[info] Running breeze.bokeh.Server
http://localhost:8080/plot/bef1d598-f19b-4234-9fa2-99e9ba286659
Copy that url into your browser. Yay for a BokehJS powered Scatterplot.
Core Concepts
At the core of Breeze-Bokeh is the registry, which is of type com.bayesianwitch.injera.misc.UUIDWeakReferenceRegistry. This is a data structure which assigns UUID's to objects sitting in memory. A server is then built which references this:
implicit val registry = new UUIDWeakReferenceRegistry[Plot]
val server = Server.start("localhost", 8080, registry)
Then you can create plots. Currently only a scatterplot is supported:
val x = DenseVector.range(0,32)
val plot = plots.Scatter(x,x,None,"xlabel", "ylabel",Some("my plot"))
println("URL is: http://localhost:8080" + plot.urlPath)
Copy&past the printed URL into a browser.
Currently the only plot type is plot.Scatterplot. The author does recognize the irony.
Garbage Collection and memory leaks
Creating a Plot object (e.g. Scatter(...)) does not hinder garbage collection. The weak registry from injera uses weak references to ensure that objects can be garbage collected when they fall out of scope. Whether or not this actually occurs is up to the JVM.
The plotting URL will continue to work until the garbage collection actually occurs, so don't build anything security critical on top of this.
Longer term vision
Those of us wanting to live in scalaland can hijack Continuum Analytics great work on Bokeh to get a good plotting library.
This will enable us to run a server and embed plots in the browser. We won't need to worry about constantly keeping parity with the Python world - all we'll need to do is import their javascript libs. This is also useful for web-based reporting - half the motivation for building this project is to embed web-based plots in the BayesianWitch webapp.
To build a Mathematica-like environment, we can also exploit Bridgewater's Scala Notebook. By building Breeze-Bokeh widgets for Scala Notebook, we can have plot embedded directly in repl.