Skip to content

Latest commit

 

History

History
361 lines (237 loc) · 12.5 KB

README.md

File metadata and controls

361 lines (237 loc) · 12.5 KB

norema - Node Reference Manual Search

About norema

norema is a web service to search node.js documentation. You can search node API documentation with norema.

norema can be deployed onto PaaS services, such as heroku.

norema uses Groonga CloudSearch, which is an Amazon CloudSearch compatible implementation, as its search back-end.

About Groonga CloudSearch, the key technology behind norema

norema is also built to be a living sample application of Groonga CloudSearch. The full source code of norema is available on Github. You can see how to build and deploy full text search powered web applications, on cloud. With Groonga CloudSearch, you can own your private search server with Amazon CloudSearch compatible APIs.

Installing Groonga CloudSearch on Ubuntu box is fairly simple. You will have working Groonga CloudSearch server in several minutes.

How norema works?

The search front-end is working on heroku, which is built with express. The search back-end is Groonga CloudSearch and running on a VPS.

The search requests from users are processed as follows:

  1. A search query is posted to the front-end via web UI.
  2. The query is translated into the Amazon CloudSearch style API request. (Some options are added. Pagination and facet selector is also considered.)
  3. The request sent to the back-end, Groonga CloudSearch.
  4. Search is executed.
  5. The search result is returned to the front-end as a JSON document.
  6. Render the search result and back to the user.

Indexing and configuring Groonga CloudSearch is needed to be done beforehand. It will be described later.

How to norema

This section describes how to setup norema on your environment step-by-step.

You will be able to deploy your norema copy on the cloud when you finished this section.

Prerequisites

In this document, we assume that you

  • are working on some *NIX environment, such as Mac OS X and Linux.
  • have an Ubuntu 12.04 (Precise Pangolin) box to be a Groonga CloudSearch server which is accessible from heroku (i.e. it has a global IP address and the access is not restricted).
  • are familiar with git, github, node.js and heroku.
  • have node.js configured on your workstation.
  • have heroku account.

Back-end

Setup Groonga CloudSearch on a VPS

First of all, add [groonga][]'s apt repository on your system, which provides Groonga CloudSearch packages.

Note: In this document, the prompt of Groonga CloudSearch server will be omitted so as to be copy and paste friendly. When you need to run some commands on your workstation, the prompt will be written explicitly as 'workstation$'.

Edit /etc/apt.sources.d/groonga.list:

sudo vi /etc/apt/sources.list.d/groonga.list

Write the file as follows:

deb http://packages.groonga.org/ubuntu/ precise universe
deb-src http://packages.groonga.org/ubuntu/ precise universe

Then, install GPG keys:

sudo apt-get update
sudo apt-get -y --allow-unauthenticated install groonga-keyring
sudo apt-get update

Now we're ready to install Groonga CloudSearch. Install gcs package:

sudo apt-get install -y gcs

That's all for the installation.

Check if Groonga Cloud is installed successfully:

gcs --version

If Groonga CloudSearch server successfully installed, you should see the version of Groonga CloudSearch.

See details in Install instructions of Groonga CloudSearch.

Create and configure norema search domain

In advance of indexing documents, we need to create a search domain.

Amazon CloudSearch API groups search target documents. Each group is searched separately. The group is called as search domain. Search domain is corresponding to table in RDBMS. A search domain has documents like a table in RDBMS has records.

Use gcs-create-domain command to create a search domain. This is the Groonga CloudSearch version of cs-create-domain command line utility.

Create search domain whose name is norema. We need to do this as the gcs user.

sudo -u gcs -H gcs-create-domain --domain-name norema

Then, configure index fields. This defines the schema of the search domain. See the detail of configuring index fields at Amazon CloudSearch document.

norema uses the following configurations:

name type options usage
title text search, result The title of the section in plain text.
desc text search, result The HTML fragments of the documentation.
text text search, result The plain text version of desc. Used for search.
path literal search, result The path to the document. Used to show breadcrumbs on search result.
path_facet literal facet The path to the document. Contains same data as path, but used for faceting.
type literal result The type of the section.
index uint search, result The appearing order of the section in the entire documents.

In order to apply the configuration, you need run the following commands:

sudo -u gcs -H gcs-configure-fields --domain-name norema --name title --type text --option result
sudo -u gcs -H gcs-configure-fields --domain-name norema --name desc --type text --option result
sudo -u gcs -H gcs-configure-fields --domain-name norema --name text --type text --option result
sudo -u gcs -H gcs-configure-fields --domain-name norema --name path --type literal --option result
sudo -u gcs -H gcs-configure-fields --domain-name norema --name path --type literal --option search
sudo -u gcs -H gcs-configure-fields --domain-name norema --name path_facet --type literal --option facet
sudo -u gcs -H gcs-configure-fields --domain-name norema --name type --type literal --option result
sudo -u gcs -H gcs-configure-fields --domain-name norema --name index --type uint

You can check the domain configuration by gcs-describe-domain command.

sudo -u gcs -H gcs-describe-domain

You will see something like this if the configuration is done correctly:

=== Domain Summary ===
Domain Name: norema
Document Service endpoint: doc-norema-xxxxxxxxxxxxxxxxxxxxxxxxxx.127.0.0.1.xip.io:7575
Search Service endpoint: search-norema-xxxxxxxxxxxxxxxxxxxxxxxxxx.127.0.0.1.xip.io:7575
SearchInstanceType: null
SearchPartitionCount: 0
SearchInstanceCount: 0
Searchable Documents: 0
Current configuration changes require a call to IndexDocuments: No

=== Domain Configuration ===

Fields:
=======
desc Active text (Search Result)
index Active uint (Search Result)
path Active literal (Search Result)
path_facet Active literal (Facet)
text Active text (Search Result)
title Active text (Search Result)
type Active literal (Result)
======================
*Note: the hostname and the port number is detected from the default options. If you run the service with your favorite host name and port number, then use it instead of default information.

Index documents

Here, we are ready to index the documents. The documents should be written in SDF format to index. See details about SDF in the section Creating SDF Batches in Amazon CloudSearch of the Amazon CloudSearch Developer Guide.

Node.js provides its document in JSON format. We convert it into the SDF (SDF is also written as JSON. A bit confusing...).

The converter is bundled with norema. Let's clone norema project on your workstation.

workstation$ git clone git://github.com/nroonga/norema.git

Directory norema will be created. Enter the directory:

workstation$ cd norema

Then, install the packages required for norema by using npm command:

workstation$ npm install

Fetch Node.js documentation (of JSON version):

workstation$ wget http://nodejs.org/docs/latest/api/all.json

Convert into SDF:

workstation$ ./tools/convert.js

Now you have all.sdf.json. Copy it onto your Groonga CloudSearch server by scp (or anything you like):

workstation$ scp all.sdf.json [your Groonga CloudSearch Server]:

And index it on the Groonga CloudSearch server (Note that you must run as user gcs):

sudo -u gcs -H gcs-post-sdf --domain-name norema --source all.sdf.json

If indexed successfully, you will see something like this:

Processing: /home/dara/all.sdf.json
Detected source format for all.sdf.json as json
Status: success
Added: 892
Deleted: 0

Now we are ready to search the documents.

Test the search by hand

Before proceeding on the frontend configuration, let us confirm the index works correctly by issuing search request by curl command.

In order to issue the request, we have to know the endpoints of the search domain.

We can have endpoint addresses by using gcs-describe-domain command on the Groonga CloudSeach server.

sudo -u gcs -H gcs-describe-domain

Document Service endpoint and Search Service endpoint are unique endpoints for the domain.

These endpoints should be like the followings, where domain-id is a random string:

Document Service endpoint: doc-norema-[domain-id].127.0.0.1.xip.io:7575
Search Service endpoint: search-norema-[domain-id].127.0.0.1.xip.io:7575

These addresses are using [xip.io][http://xip.io], which provides wildcard DNS, and directed to the localhost (127.0.0.1).

In order to issue request to the endpoints from the other machines, you need replace 127.0.0.1 with the globally accessible IP address of the server.

For example, when the IP address of your Groonga CloudSearch server 192.0.2.1, globally accessible endpoints are the followings:

Document Service endpoint: doc-norema-[domain-id].192.0.2.1.xip.io:7575
Search Service endpoint: search-norema-[domain-id].192.0.2.1.xip.io:7575

The URL to search console from the norema domain is http://search-norema-[domain-id].[gcs-server-ip].xip.io:7575/2011-02-01/search?q=console.

You can obtain the result by issuing GET request to the URL (Note that you need to replace 192.0.2.1 with your Groonga CloudSearch server's IP):

curl "http://search-norema-[domain-id].192.0.2.1.xip.io:7575/2011-02-01/search?q=console"

The response will be like this:

{"rank":"-text_relevance","match-expr":"(label 'console')","hits":{"found":172,"start":0,"hit":[{"id":"doc_3"},{"id":"doc_10"},{"id":"doc_11"},{"id":"doc_12"},{"id":"doc_13"},{"id":"doc_14"},{"id":"doc_17"},{"id":"doc_21"},{"id":"doc_22"},{"id":"doc_23"}]},"info":{"rid":"000000000000000000000000000000000000000000000000000000000000000","time-ms":19,"cpu-time-ms":0}}

Now your search back-end is fully configured. Move on the front-end.

Front-end

Try norema front-end locally

Work on your workstation, in norema directory, which is the directory norema project cloned.

You can run norema app locally by the followings:

workstation$ env SEARCH_ENDPOINT=search-norema-[domain-id].[gcs-server-ip].xip.io:7575 nom start

Now norema server will listen at the port 3000.

You will see norema server working on http://localhost:3000/ with your browser. Try it.

Deploy to heroku

Then, let's deploy the front-end onto the cloud. You need to have working heroku configuration to proceed.

Create heroku project:

workstation$ heroku create

Configure search endpoint:

workstation$ heroku config:add SEARCH_ENDPOINT=search-norema-[domain-id].[gcs-server-ip].xip.io:7575

And deploy the code:

workstation$ git push heroku master

That's all! Try on your browser:

workstation$ heroku open

Conclusion

norema is a web service to search Node.js documents. It is build on the top of Groonga CloudSearch and running in the cloud. The installation of norema is explained step by step. You can learn how Groonga CloudSearch applications is built.