Json Searcher

Descriptions:

A Json searcher support fast search on defined entities

OS: *Debian (Docker)

Get Started

Systems Diagram

Design Philosophy

Appendix

Get Started

How to: run the app in an env with jdk8 and sbt setup

Run Test:

sbt clean test

Start the console app:

sbt clean run

How to: run the app without jdk and sbt

Please be aware, the docker image will be built from scratch, which means the building time for the first run would be super long....

after the first run, two external volume would be created to cache all the loaded dependencies to make the next run much faster;

Run Test:

./auto/test

Start the console app:

./auto/run

Systems Diagram

 
 ________________________________                  ______________________________________________________
|    Console                    |                  |            Json-Searcher-Core                       |
|                               |                  |        __________________________________           |               
|  atStart                      |        |---------|-----> |- materialise denormolised-data   |          |
|  atInstruction                |        |         |       |- preload indices                 |          |
|  atSearchableFields           |        |         |       |__________________________________|          |                         
|  atStartSearch   -------------|--------|         |                                   |                 |
|  atEnterSearchTerm     --->|  |               |--|-->A In memory SearchStore <-------|                 |
|  atsEnterSearchValue   |<---| |               |  |                |                                    |
|  atOperateSearch   <---|      |               |  |                |                                    |
|               |               |               |  |                |                                    |
|               |_____________SearchTerm________|  |                |                                    |
|                               |                  |                |                                    |   
|     stdout    <---------------|----SearchResults-|----------------|                                    |
|_______________________________|                  |_____________________________________________________|

Design Philosophy

Core Principles:

• Searhable Data should be denormolised for fast search
• Data should be indexed by the available search key
• Searchable Data should be materialised into memory before SearchOperation
• Console should have finite states and each state should know its previous restatable state
• Exceptions should not break the user experiences and should have instructive notifications

Disadvantages and optimizations:

Out of Memonry when large volume of data exceeding single machine capacity:

Although diskIO by streaming in-memory-denormolised data into multiple partitions in disk by hash key for a given search key could guarantee the memory would not be explored given a big enough disk volume;

This appraoch would slow down the search dramactically even with HDFS support;

The ideal way to optimise this large volume data scenario would be applying distributed computing in a cluster with mutiple machines; each instance could be working on a partition of searchkey and later on assemble the result to serve back to the searcher; It is a well-known map-reduce application senario;

Appendix

Domain Analysis:

1. Organisation:

   All Fields are required

2. Tickets:

   Optional Fields: [type, description, assignee_id, organization_id, due_at]

3. User:

   Optional Fields: [alias, verified, locale, timezone, email, phone, organization_id]

Denormolised-View for search-results-entity:

• on user granularity:

  UserSearchResult {
      user: User,
      org: Option[Organisation],
      assigned_tickets: List[Ticket],
      submitted_tickets: List[Ticket]   
  }
  

• on ticket granularity:

  TicketSearchResult {
      ticket: Ticket,
      org: Option[Organisation],
      assignee: Option[user],
      submitter: Option[user] 
  }
  

• on organisation granularity:

  OrganisationSearchResult {
      org: Organisation,
      users: Array[User],
      tickets: Array[Ticket]
  }