通过一次查询中的自我连接获取has_many的所有子级

Question

I am having a library full of Articles. An article references other articles through citations. So it's self-referencing through citations, like this:

class Article < ActiveRecord::Base
  has_many :citations
  has_many :referenced_articles, through: :citations
end

class Citation < ActiveRecord::Base
  belongs_to :Article
  belongs_to :referenced_article, :class_name => "Article"
end

The question: How can I, given some article, fetch all citated articles: both directly citated and indirectly citated by it's referenced articles, etc. until reaching the articles that do not have any citations (in my case that will probably be no more than 15 levels deep).

My current method: Given some article "main_article", I can get it's referenced_articles by

sub_articles = main_article.referenced_articles

And to go one level deeper I do:

sub_articles = Article.eager_load(:referenced_articles).where(id: sub_articles)
sub_sub_articles = sub_articles.collect {|x| x.referenced_articles}.flatten

Putting this in a loop gives the desired result, going one level deeper every iteration. But i am hoping for a nicer solution, as I am currently fetching every article twice.

Boundary condition: I don't want to add a referencing instance to every article in my library telling it where it is used(citated). So I guess the 'ancestry' gem is not an option :(

Answer 1

I don't have any experience with recursive queries, but if your database engine supports them you might want to look at common-table-expressions with recursion:

http://en.wikipedia.org/wiki/Hierarchical_and_recursive_queries_in_SQL#Common_table_expression

Oracle, sqlserver, and postgres all seem to support recursive cte's.

You'll probably have to craft the raw sql for this yourself, I doubt you'll find much support in rails for doing this with a single query.