dmitriy-kiriyenko / abacus_count

`ActiveRecord::Base#count` and other calculations as subqueries.

Abacus Count

Synopsis

ActiveRecord::Base#count and other calculations as subqueries. Instead of nice Rails grouping calculation feature, uses subqueries to return the result with any complex query.

There is a use case when ActiveRecord calculations fail on a valid relation. It's when you use alias values you select, and then use aliases in where or having conditions. That happens because ActiveRecord does calculations by throwing away select values. And when really do need to use custom aliases outside select, you definitely do not want this kind of optimization. More on that bug in this pull request.

The issue is fixed, so Rails master and 3.1.1 will be free of it. However, if you are using earlier version of Rails 3å, we can use an ultimate way of preventing any calculation issues, suggested by abacus_count. I mean using subqueries.

Installation

In your Gemfile:

  gem "abacus_count"

and then run bundle.

Usage

You now have abacus scope on any relation. This scope extends a relation with a calculations patch. With it, each calculation will be performed through subquery and so they won't fail in any case.

  users = User.select("users.id, avg(transactions.amount) as avg_amount").joins(:transactions).group("user_id").having("avg_amount >= 15")
  users.count # will fail
  users.abacus.count # will do

  # with Kaminari
  users.page(params[:page]).per(10) # will fail in count
  users.abacus.page(params[:page]).per(10) # will do

Caveats

In abacus calculations never return a hash, always a total result. Sometimes, this is just what you want, however.

Support and contribution

Tested with ruby 1.8.7 and ruby 1.9.3. Tested with ActiveRecord 3.0.9 and 3.1.0.rc5. Great thanks to a wonderful Thoughtbot's Appraisal gem. Tested with sqlite3 and mysql. Currently test suite runs only under sqlite3, gonna implement automated acceptance testing with mysql and maybe postgres.

Contribution

Install to your app and use it! If something bad happens, please, submit an issue or a pull request.