[ZEPPELIN-1219] Add searching feature to Zeppelin docs site

[AhyoungRyu]

What is this PR for?

As more and more document pages are added, it's really hard to find specific pages. So I added searching feature to Zeppelin documentation site(jekyll based site) using lunr.js.

• How does it work?

  I created search_data.json which is used for docs info template. lunr.js combines all of the text from all of the docs in docs/ into _site/search_data.json. It looks like below.
  
  All the info are comes from Jekyll YAML front matter variables. (i.e. title, group, description.. that's why I rewrote all docs' title and description.)
  search.js will do this job using this data!

What type of PR is it?

Improvement & Feature

Todos

• - Keep consistency for all docs pages' Title
• - Add some overview sentences to all docs pages' Description section (this will be used as the result preview)
• - Add apache license header to all docs page (some pages are missing the license header currently)
• - Add LICENSE for lunr.min.js

What is the Jira issue?

ZEPPELIN-1219

How should this be tested?

1. Apply this patch and build ZEPPELIN_HOME/docs dir -> please see docs/README.md#build-documentation
2. Click search icon in navbar and go to search.html page
3. Type anything you want to search in the search bar (i.e. type python, spark, dynamic ... )

Screenshots (if appropriate)

Questions:

• Does the licenses files need update? Yes, for lunr.min.js
• Is there breaking changes for older versions? no
• Does this needs documentation? no

[bzz]

This page will help you to get started and guide you through installation of Apache Zeppelin, running it in the command line and basic configuration options.

[bzz]

Great improvement @AhyoungRyu ! Posted few suggestions above.

[AhyoungRyu]

@bzz Really appreciate for your all precise comments! Updated several docs as you suggested :)

[AhyoungRyu]

It seems the CI failure is irrelevant with this PR. Can someone review this one?

[bzz]

@AhyoungRyu it's better to post CI failure to help getting feedback faster AND search existing JIRA issues, to see if that is something know or being addressed elsewhere.

Here is an example of what I mean for this PR:

---

Single CI profile Spark 1.4\Scala 2.10 is failing here with

[INFO] Building Zeppelin: Server 0.7.0-SNAPSHOT
...
Running org.apache.zeppelin.rest.ZeppelinSparkClusterTest
....
org.apache.zeppelin.interpreter.LazyOpenInterpreter@20ea2ef9

No output has been received in the last 10 minutes, this potentially indicates a stalled build or something wrong with the build itself.

The build has been terminated

Searching JIRA for ZeppelinSparkClusterTest brings ZEPPELIN-862 which looks very similar and is not relevant to this changes. This PR has been added to that issue to track this flaky tests

---

Hope this helps! In this particular case (single test failure) it might be work to close and re-open to PR to see if the test failure is persistent\reproducible.

In my experience if PR author try to follow such steps - that speeds up the review process by order of magnitude AND helps the project to make CI more stable.

Documentation update + search looks great to me, merging to master after CI is re-triggered by @AhyoungRyu , if there is no further discussion.

[AhyoungRyu]

@bzz

In my experience if PR author try to follow such steps - that speeds up the review process by order of magnitude AND helps the project to make CI more stable.

Yeah right, I should've done that. It was my bad. Let me re-trigger CI then :)

[AhyoungRyu]

@bzz After re-triggered CI build, it fails again only in spark 1.4\ scala 2.10 with

Running org.apache.zeppelin.rest.ZeppelinSparkClusterTest
...
No output has been received in the last 10 minutes, this potentially indicates a stalled build or something wrong with the build itself.

The build has been terminated

[bzz]

Yeah right, I should've done that. It was my bad. Let me re-trigger CI then :)

No worries! A lot of people do not do this and ask no excuses. So let's just show them what excellence means, by annotating further CI failures in case of any, as you did for the last one here :)

Different profile fails with the same flaky test, not relevant to the changeset.

Looks great to me, and merging to master if there is no further discussion!