Skip to content

anirbanroydas/elaster

Repository files navigation

elaster

An in-app full-text Search Engine using Elasticsearch (a document store search engine based on Lucene), tornado as web server, sockjs in client(browser) side javascript library, sockjs-tornado as sockjs implementation on server side and elasticsearch-py as elasticsearch python client library.

NOTE : Still in developement. Not ready for release. Info added prior just for record purpose.

NOTE : Meanwhile, you can check the Example - Webapp images and static codes for an overview.

Example - Webapp

  • Welcome Page

img1.png

  • Select app type

img2.png

Documentation

Link : http://elaster.readthedocs.io/en/latest/

Project Home Page

Link : https://www.github.com/anirbanroydas/elaster

Details

Author:Anirban Roy Das
Email:anirban.nick@gmail.com
Copyright(C):2017, Anirban Roy Das <anirban.nick@gmail.com>

Check elaster/LICENSE file for full Copyright notice.

Overview

elaster is an in-app full-text Search Engine based on Elasticsearch which can be set up to search your app, website, etc. for any text.

All you have to do is add data to the index, just the data files, all mappings, indexing, settings are taken care by the elaster server. You can start using the server right away without knowing anything about elasticsearch.

Even the cluster, nodes, etc. are taken care of.

If you want more power and control, change the configuration file at /usr/local/etc/elasticsearch/elasticsearch.yml.

It uses the Elasticsearch to implement the real time full text search functionality. Elasticsearch is a search server based on Lucene. It provides a distributed, multitenant-capable full-text search engine with an HTTP web interface and schema-free JSON documents. Elasticsearch is developed in Java and is released as open source under the terms of the Apache License. Elasticsearch is the most popular enterprise search engine followed by Apache Solr, also based on Lucene.

A website example is given as builtin. For the website , the connection is created using the sockjs protocol. SockJS is implemented in many languages, primarily in Javascript to talk to the servers in real time, which tries to create a duplex bi-directional connection between the Client(browser) and the Server. Ther server should also implement the sockjs protocol. Thus using the sockjs-tornado library which exposes the sockjs protocol in Tornado server.

It first tries to create a Websocket connection, and if it fails then it fallbacks to other transport mechanisms, such as Ajax, long polling, etc. After the connection is established, the tornado server (sockjs-tornado) connects to Elasticsearch via using the Elasticsearch Python Client Library, elasticsearch-py.

Thus the connection is web-browser to tornado to elasticsearch and vice versa.

For any other app (Android, iOS), use any android, iOS client library to connect to the elaster server. This software provides the website example builtin. Using command line elaster --example=webapp, you start the elaster website example. But to use it in your person app(Anroid, iOS) or web app, use elaster as a general server.

Technical Specs

sockjs-client (optional):Advanced Websocket Javascript Client used in webapp example
Tornado:Async Python Web Library + Web Server
sockjs-tornado:SockJS websocket server implementation for Tornado
Elasticsearch:A document store search engine based on Lucene
elasticsearch-py:Low-level elasticsearch python client
pytest:Python testing library and test runner with awesome test discobery
pytest-flask:Pytest plugin for flask apps, to test fask apps using pytest library.
Uber's Test-Double:Test Double library for python, a good alternative to the mock library
Jenkins (Optional):A Self-hosted CI server
Travis-CI (Optional):A hosted CI server free for open-source projecs
Docker:A containerization tool for better devops

Features

  • Search Engine
  • in-app (Androi, iOS, website)
  • Use as a standalone server or as a python library
  • Suggestion based search
  • Autocomplete based search
  • Add/Index your own dataset in json files
  • Microservice
  • Testing using Docker and Docker Compose
  • CI servers like Jenkins, Travis-CI

Installation

Prerequisite (Optional)

To safegurad secret and confidential data leakage via your git commits to public github repo, check git-secrets.

This git secrets project helps in preventing secrete leakage by mistake.

Dependencies

  1. Docker
  2. Make (Makefile)

See, there are so many technologies used mentioned in the tech specs and yet the dependencies are just two. This is the power of Docker.

Install

  • Step 1 - Install Docker

    Follow my another github project, where everything related to DevOps and scripts are mentioned along with setting up a development environemt to use Docker is mentioned.

    • Go to setup directory and follow the setup instructions for your own platform, linux/macos
  • Step 2 - Install Make

    # (Mac Os)
    $ brew install automake
    
    # (Ubuntu)
    $ sudo apt-get update
    $ sudo apt-get install make
    
  • Step 3 - Install Dependencies

    Install the following dependencies on your local development machine which will be used in various scripts.

    1. openssl
    2. ssh-keygen
    3. openssh

CI Setup

If you are using the project in a CI setup (like travis, jenkins), then, on every push to github, you can set up your travis build or jenkins pipeline. Travis will use the .travis.yml file and Jenknis will use the Jenkinsfile to do their jobs. Now, in case you are using Travis, then run the Travis specific setup commands and for Jenkins run the Jenkins specific setup commands first. You can also use both to compare between there performance.

The setup keys read the values from a .env file which has all the environment variables exported. But you will notice an example env file and not a .env file. Make sure to copy the env file to .env and change/modify the actual variables with your real values.

The .env files are not commited to git since they are mentioned in the .gitignore file to prevent any leakage of confidential data.

After you run the setup commands, you will be presented with a number of secure keys. Copy those to your config files before proceeding.

NOTE: This is a one time setup. NOTE: Check the setup scripts inside the scripts/ directory to understand what are the environment variables whose encrypted keys are provided. NOTE: Don't forget to Copy the secure keys to your .travis.yml or Jenkinsfile

NOTE: If you don't want to do the copy of env to .env file and change the variable values in .env with your real values then you can just edit the travis-setup.sh or jenknis-setup.sh script and update the values their directly. The scripts are in the scripts/ project level directory.

IMPORTANT: You have to run the travis-setup.sh script or the jenkins-setup.sh script in your local machine before deploying to remote server.

Travis Setup

These steps will encrypt your environment variables to secure your confidential data like api keys, docker based keys, deploy specific keys.

$ make travis-setup

Jenkins Setup

These steps will encrypt your environment variables to secure your confidential data like api keys, docker based keys, deploy specific keys.

$ make jenkins-setup

Usage

After having installed the above dependencies, and ran the Optional (If not using any CI Server) or Required (If using any CI Server) CI Setup Step, then just run the following commands to use it:

You can run and test the app in your local development machine or you can run and test directly in a remote machine. You can also run and test in a production environment.

Run

The below commands will start everythin in development environment. To start in a production environment, suffix -prod to every make command.

For example, if the normal command is make start, then for production environment, use make start-prod. Do this modification to each command you want to run in production environment.

Exceptions: You cannot use the above method for test commands, test commands are same for every environment. Also the make system-prune command is standalone with no production specific variation (Remains same in all environments).

  • Start Applcation

    $ make clean
    $ make build
    $ make start
    
    # OR
    
    $ docker-compose up -d
    
  • Stop Application

    $ make stop
    
    # OR
    
    $ docker-compose stop
    
  • Remove and Clean Application

    $ make clean
    
    # OR
    
    $ docker-compose rm --force -v
    $ echo "y" | docker system prune
    
  • Clean System

    $ make system-prune
    
    # OR
    
    $ echo "y" | docker system prune
    

Logging

  • To check the whole application Logs

    $ make check-logs
    
    # OR
    
    $ docker-compose logs --follow --tail=10
    
  • To check just the python app's logs

    $ make check-logs-app
    
    # OR
    
    $ docker-compose logs --follow --tail=10 identidock
    

Test

Now, testing is the main deal of the project. You can test in many ways, namely, using make commands as mentioned in the below commands, which automates everything and you don't have to know anything else, like what test library or framework is being used, how the tests are happening, either directly or via docker containers, or may be different virtual environments using tox. Nothing is required to be known.

On the other hand if you want fine control over the tests, then you can run them directly, either by using pytest commands, or via tox commands to run them in different python environments or by using docker-compose commands to run differetn tests.

But running the make commands is lawasy the go to strategy and reccomended approach for this project.

NOTE: Tox can be used directly, where docker containers will not be used. Although we can try to run tox inside our test contianers that we are using for running the tests using the make commands, but then we would have to change the Dockerfile and install all the python dependencies like python2.7, python3.x and then run tox commands from inside the docker containers which then run the pytest commands which we run now to perform our tests inside the current test containers.

CAVEAT: The only caveat of using the make commands directly and not using tox is we are only testing the project in a single python environment, nameley python 3.6.

  • To Test everything

    $ make test
    

    Any Other method without using make will involve writing a lot of commands. So use the make command preferrably

  • To perform Unit Tests

    $ make test-unit
    
  • To perform Component Tests

    $ make test-component
    
  • To perform Contract Tests

    $ make test-contract
    
  • To perform Integration Tests

    $ make test-integration
    
  • To perform End To End (e2e) or System or UI Acceptance or Functional Tests

    $ make test-e2e
    
    # OR
    
    $ make test-system
    
    # OR
    
    $ make test-ui-acceptance
    
    # OR
    
    $ make test-functional
    

Todo

  1. Add Blog post regarding this topic.
  2. Add Contract Tests using pact
  3. Add integration tests
  4. Add e2d tests