On 20 Apr 2012, at 17:27, Etienne Dumoulin wrote:
> Like it is a bit difficult to start from scratch a first time,
> I copied/pasted the getting started code.
> I do not understand why there is 3 template classes when
> TextVertexInputFormat has 4
> (from the javadoc and eclipse).
> For example:
> "public static class SimpleShortestPathsVertexInputFormat extends
> TextVertexInputFormat<LongWritable, DoubleWritable, FloatWritable>"
> Is there something I miss?
Personally, I dont think that the code to which the getting started guide
is the best place to start *understanding* Giraph.
But its the best code to execute in order to see if your Hadoop setup actually
works for Giraph.
A better example for understanding Giraph IMHO is the
ConnectedComponentsVertexTest class (in the test directory of the code).
It shows that you need to implement a Vertex class, an TextInputFormat and a
TextOutputFormat class in order to define your own job.
Then you can use InternalVertexRunner.run() in order to test your code inside
of Giraph, and figure out Hadoop later.
I was puzzled about your "3 templates versus 4" error, but then I tried it out
in Eclipse. If you are also using eclipse, then I know what you mean ;)
If you mouse over e.g. SimpleShortesPathsVertex, then eclipse will display a
tool-tip. However, thats just the top part of the javadoc with 3 template
parameters, you need to scroll down,
in order to see all 4 elements of the template signature. To do that, you can
either press F2, to see the scroll bar or resize the window.
Or you click on the class, and then you can read the source of the javadoc.
However, keep in mind, that some specific classes which are used in the
TextInputFormat and the TextOutputFormat have template signatures
which really only have 2 or 3 elements.
As a second thing, I would look at the SimpleShortestPathVertexTest class
(again in the test directory).
Then you can look at the PageRankeBenchmark class in giraph.benchmark.
Let me/us know if you have more questions.