This is a private, internal application that allows users to post and answer questions. They may also search questions.
I went through this project very simply. I aimed to use the tools closest to my toolbox- straightforward REST/CRUD with Rails-based conventions and common technologies (devise, postgres, bootstrap, heroku deployment, RSpec, Capybara).
For markdown processing, markdown_helper.rb gives access to markdown(text) throughout the application. This is done with redcarpet.
For searching, QueryPad utilises the pg_search gem. It applies this gem to use PostgreSQL' built-in full text tsearch as well as the enabled PostgreSQL trigram extension to assist in 'fuzzy matching' search queries.
I used the decorator pattern within the application for user flair.
There is a user voting feature employing polymorphic associations with Vote to Question and Answer (associating to votable). Upvote vs. downvote is set on a type attribute on vote.
Other notable features:
- Users obtain a rolling 'score' that is simply the sum of upvotes on their questions and answers. This is set with
incrementcounter cache directly to the database. Removing a vote or adding a downvote decrements scores on users, questions and answers. - Users with a 'score' above 25 may downvote and become a 'power user' with flair.
- Users with a 'score' above 50 may 'delete' posts and become a 'moderator', also with flair.
- 'Deleted' posts are preserved in the database but hidden on the front end (using acts_as_paranoid).
- You can sort questions on index by recently created or highest score.
- CI build hooked up with Travis-ci
I built this application expecting a relatively small database (< 10k Q's+A's), and a commensurate request volume. I also limited myself to a seven-day build-cycle, working in evenings where permissible. This allowed me to fulfill the specifications and add a handful of extended features. I held myself to the standard that the application must pass a traditional CI build (RSpec/brakeman/bundler-audit) and have >95% test coverage.
This application observes the YAGNI and KISS principles. If QueryPad were to scale much larger than a normal internal tool, I would look to use searchkick for elasticsearch implementation. I could see upvote/downvote logic being a pinchpoint at large scales. That could be resolved using read/write to a redis store. Anything beyond the simple current authorities of 'power user' and 'moderator' would benefit from a CanCanCan or Pundit authorisation library.
- Ruby: 2.5.1
- Rails: 5.2.0
Run bin/setup to setup your development environment and install any needed dependencies.
Run rake.
Using Google Oauth with the omniauth-google-oauth2 gem alongside devise.
Uses Rails credentials introduced in 5.2.
- Rspec
- Brakeman
- Bundler Audit