Both methods return the same output (an array of strings).

Article.all.map{|a| a.title}

Article.pluck(:title)

But they do that in different ways.

SQL query generated by all:

SELECT "articles".* FROM "articles"

It fetches all atributes, then it converts the collection with map method.

SQL query generated by pluck (or .all.pluck):

SELECT "articles"."title" FROM "articles"

pluck is faster.

My model will be called Node, and I want to connect them with other nodes in a way that is called “directed graph”.

An example of a directed graph from Wikimedia Commons.

Each node can have zero, one or many children, and zero, one or many parents. To do this I need an another model which will represent connections between nodes. I assume that Node model already exists and I will create the NodeConnection model by generating a scaffold.

rails g scaffold NodeConnection child:references parent:references --no-stylesheets

Nodes will be connected with other nodes, but I can’t make two column with the same “node” name. Instead of that the names are “child” and “parent”. I think the same result can be achived by child_id:integer parent_id:integer instead of child:references parent:references.

An auto-generated migration is OK, I don’t need to add anything at this moment, so I can migrate.

rake db:migrate

The associations in the NodeConnection model file should have additional arguments.

class NodeConnection < ActiveRecord::Base
  belongs_to :parent, class_name: 'Node', foreign_key: 'parent_id'
  belongs_to :child, class_name: 'Node', foreign_key: 'child_id'
end

And an assiociation in the Node model file:

class Node < ActiveRecord::Base
  has_many :parent_connections, foreign_key: 'child_id', class_name: 'NodeConnection'
  has_many :parents, through: :parent_connections

  has_many :child_connections, foreign_key: 'parent_id', class_name: 'NodeConnection'
  has_many :children, through: :child_connections
end

For each node there are now methods parents and children avaliable.