[WIP] Incorporate the official templates in core repository

[takezoe]

I'm considering about incorporate the official templates with main repository to reduce maintenance costs of them. My idea is:

• Check versions of PredictionIO/Scala/Spark/etc by TravisCI using a shell script and if these versions are not same among PredictionIO and templates, test will fail.
• Include the official templates and examples in the distribution which is made by make-distribution.sh.
• Documentations which refer these templates or examples will be fixed to be started from:

  $ cd $PIO_HOME/templates/template-scala-parallel-recommendation
  $ pio build
  ...
  

  instead of:

  $ git clone https://github.com/apache/predictionio-template-recommender.git
  $ cd predictionio-template-recommender
  $ pio build
  ...
  

Currently, for every release, we have to update versions of PredictionIO/Scala/Spark/etc, then create tag for each template repository by hand. If the official templates are in the same repository with PredictionIO itself, we will be able to finish this boring operation at once. Also, it will ensure that all official templates is synchronized with the latest PredictionIO version.

[Wei-1]

I think it causes some confusions when new users try to learn PIO and found that they will need to clone the template from another repository.
So it is definitely better for us to put it together with the main repo.
Since we already have those examples in the main repo, I think it makes a lot of sense.

One question that I will bring up is:
Are we still going to use the package org.example.recommendation?
or maybe we could change it to org.apache.predictionio.template?

[takezoe]

@Wei-1

Are we still going to use the package org.example.recommendation?
or maybe we could change it to org.apache.predictionio.template?

We should locate templates under org.apache.predictionio package if we will distribute templates with distribution. Thanks!

[takezoe]

Is the license file necessary for each template after they are incorporated into this repository?

[takezoe]

We will have to maintain README of each template after incorporating, I think it should be synchronized with PredictionIO release, but text-classifier template has inconsistent versioning with PredictionIO version. How should we handle that?

In addition, text-classifier template contains large data in its repository, I want to separate that data from the repository. Is there any suitable place to put them?

[Wei-1]

I think the test you added to check_templates.sh is a good idea.
People will know that they have to change the corresponding template version while modifying PIO's version.
As for the data, maybe we add another non-test branch?

[takezoe]

Add incorporated templates here to check by TravisCI.

[takezoe]

This script is run after unit tests on TravisCI to check version consistency and test templates.

[shimamoto]

@takezoe , sorry for the late response.

Basically, I agree with incorporating them into the main repository. I'm concerned about the following before this PR.

• What templates we incorporate into the main repository
  We might as well discuss whether we continue to maintain all of the current official templates.

• Namespace
  Though the official templates are maintained by committers, they are NOT as part of main modules but as part of examples. I think they should be under examples folder and have org.apache.predictionio.examples package.