Rack::Reducer

Declaratively filter data via URL params, in any Rack app, with any ORM.

Add rack-reducer to your Gemfile:

gem 'rack-reducer', require: 'rack/reducer'

Rack::Reducer has no dependencies beyond Rack itself.

If your app needs to render a list of database records, you probably want those records to be filterable via URL params, like so:

GET /artists => all artists
GET /artists?name=blake` => artists named 'blake'
GET /artists?genre=electronic&name=blake => electronic artists named 'blake'

Rack::Reducer can help. It applies incoming URL params to an array of filter functions you define, runs only the relevant filters, and returns your filtered data. Here’s how you might use it in a Rails controller:

# app/controllers/artists_controller.rb
class ArtistsController < ApplicationController

  # Step 1: Instantiate a reducer
  ArtistReducer = Rack::Reducer.new(
    Artist.all,
    ->(name:) { where('lower(name) like ?', "%#{name.downcase}%") },
    ->(genre:) { where(genre: genre) },
  )

  # Step 2: Apply the reducer to incoming requests
  def index
    @artists = ArtistReducer.apply(params)
    render json: @artists
  end
end

This example app would handle requests as follows:

# GET /artists => All artists:
[
  { "name": "Blake Mills", "genre": "alternative" },
  { "name": "Björk", "genre": "electronic" },
  { "name": "James Blake", "genre": "electronic" },
  { "name": "Janelle Monae", "genre": "alt-soul" },
  { "name": "SZA", "genre": "alt-soul" }
]

# GET /artists?name=blake => Artists named "blake":
[
  { "name": "Blake Mills", "genre": "alternative" },
  { "name": "James Blake", "genre": "electronic" }
]

# GET /artists?name=blake&genre=electronic => Electronic artists named "blake"
[{ "name": "James Blake", "genre": "electronic" }]

https://www.rubydoc.info/gems/rack-reducer

These examples apply Rack::Reducer in different frameworks and ORMs. The pairings of ORMs and frameworks are arbitrary, just to demonstrate a few possible stacks.

This example uses Sinatra to handle requests, and Sequel as an ORM.

# config.ru
class SinatraExample < Sinatra::Base
  DB = Sequel.connect ENV['DATABASE_URL']

  # dataset is a Sequel::Dataset, so filters use Sequel query methods
  ArtistReducer = Rack::Reducer.new(
    DB[:artists],
    ->(genre:) { where(genre: genre) },
    ->(name:) { grep(:name, "%#{name}%", case_insensitive: true) },
  )

  get '/artists' do
    @artists = ArtistReducer.apply(params).all
    @artists.to_json
  end
end

This example runs a raw Rack app with Rack::Reducer mounted as middleware. It doesn't use an ORM at all -- it just stores data in a ruby array.

# config.ru
require 'rack'
require 'rack/reducer'
require 'json'

ARTISTS = [
  { name: 'Blake Mills', genre: 'alternative' },
  { name: 'Björk', genre: 'electronic' },
  { name: 'James Blake', genre: 'electronic' },
  { name: 'Janelle Monae', genre: 'alt-soul' },
  { name: 'SZA', genre: 'alt-soul' },
]

app = Rack::Builder.new do
  # dataset is an Array, so filter functions use Array methods
  use Rack::Reducer::Middleware, dataset: ARTISTS, filters: [
    ->(genre:) { select { |item| item[:genre].match(/#{genre}/i) } },
    ->(name:) { select { |item| item[:name].match(/#{name}/i) } },
    ->(sort:) { sort_by { |item| item[sort.to_sym] } },
  ]
  run ->(env) { [200, {}, [env['rack.reduction'].to_json]] }
end

run app

When Rack::Reducer is mounted as middleware, it stores its filtered data in env['rack.reduction'], then calls the next app in the middleware stack. You can change the env key by passing a new name as option to use:

use Rack::Reducer::Midleware, key: 'custom.key', dataset: ARTISTS, filters: [
  # an array of lambdas
]

The Rails quickstart example created a reducer inside a controller, but if your filters use lots of ActiveRecord scopes, it might make more sense to keep your reducers in your models instead.

# app/models/artist.rb
class Artist < ApplicationRecord
  # filters get instance_exec'd against the dataset you provide -- in this case
  # it's `self.all` -- so filters can use query methods, scopes, etc
  Reducer = Rack::Reducer.new(
    self.all,
    ->(name:) { by_name(name) },
    ->(genre:) { where(genre: genre) },
    ->(sort:) { order(sort.to_sym) }
  )

  scope :by_name, lambda { |name|
    where('lower(name) like ?', "%#{name.downcase}%")
  }
end

# app/controllers/artists_controller.rb
class ArtistsController < ApplicationController
  def index
    @artists = Artist::Reducer.apply(params)
    render json: @artists
  end
end

Most of the time it makes sense to use required keyword arguments for each filter, and skip running the filter altogether when the keyword argments aren't present.

But sometimes you'll want to run a filter with a default value, even when the required params are missing. The code below will order by params[:sort] when it exists, and by name otherwise.

class ArtistsController < ApplicationController
  ArtistReducer = Rack::Reducer.new(
    Artist.all,
    ->(genre:) { where(genre: genre) },
    ->(sort: 'name') { order(sort.to_sym) }
  )

  def index
    @artists = ArtistReducer.apply(params)
    render json: @artists
  end
end

For a slight performance penalty (~5%), you can skip instantiating a reducer via ::new and just call Rack::Reducer as a function. This can be useful when prototyping, mostly because you don't need to think about naming anything.

# app/controllers/artists_controller.rb
class ArtistsController < ApplicationController
  # Step 1: there is no step 2
  def index
    @artists = Rack::Reducer.call(params, dataset: Artist.all, filters: [
      ->(name:) { where('lower(name) like ?', "%#{name.downcase}%") },
      ->(genre:) { where(genre: genre) },
    ])
    render json: @artists
  end
end

Rack::Reducer takes a dataset, an array of lambdas, and a params hash.

To return filtered data, it calls Enumerable#reduce on your array of lambdas, with the reduction's initial value set to dataset.

Each reduction looks for keys in the params hash that match the current lambda's keyword arguments. If the keys exist, it instance_execs the lambda against the dataset, passing just those keys as arguments, and finally passes the filtered dataset on to the next lambda.

Lambdas that don't find all their required keyword arguments in params don't execute at all, and just pass the unaltered dataset down the chain.

The reason Reducer works with any ORM is that you supply the dataset and filter functions. Reducer doesn't need to know anything about ActiveRecord, Sequel, Mongoid, etc -- it just instance_execs your own code against your own dataset.

For requests with empty params, Rack::Reducer has no measurable performance impact. For requests with populated params, Rack::Reducer is about 10% slower than a set of hand-coded conditionals, according to spec/benchmarks.rb.

 Conditionals (full)   530.000  i/100ms
      Reducer (full)   432.000  i/100ms
Conditionals (empty)   780.000  i/100ms
     Reducer (empty)   808.000  i/100ms
Calculating -------------------------------------
 Conditionals (full)      4.864k (± 2.3%) i/s -     24.380k in   5.015551s
      Reducer (full)      4.384k (± 1.3%) i/s -     22.032k in   5.026651s
Conditionals (empty)      7.889k (± 1.7%) i/s -     39.780k in   5.043797s
     Reducer (empty)      8.129k (± 1.7%) i/s -     41.208k in   5.070453s

Comparison:
     Reducer (empty):     8129.5 i/s
Conditionals (empty):     7889.3 i/s - same-ish: difference falls within error
 Conditionals (full):     4863.7 i/s - 1.67x  slower
      Reducer (full):     4383.8 i/s - 1.85x  slower

In Rails, note that params is never empty, so use request.query_parameters instead if you want to handle parameterless requests at top speed.

# app/controllers/artists_controller.rb
class ArtistController < ApplicationController
  # ArtistReducer = Rack::Reducer.new(...etc etc)

  def index
    @artists = ArtistReducer.apply(request.query_parameters)
    render json: @artists
  end
end

If you're working in Rails, Plataformatec's excellent HasScope has been solving this problem since 2009. I prefer keeping my request logic all in one place, though, instead of spreading it across my controllers and models.

Periscope, by Steve Richert, seems like another solid Rails option. It is Rails-only, but it supports more than just ActiveRecord.

For Sinatra, Simon Courtois has a Sinatra port of has_scope. It depends on ActiveRecord.

Thank you @danielpuglisi, @nicolasleger, @jeremyshearer, and @shanecav84 for helping improve Rack::Reducer!

Please open an issue on Github.

PRs are welcome, and I'll do my best to review them promptly.

Copyright 2018 Chris Frank

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.