ActsAsRecursiveTree

Use the power of recursive SQL statements in your Rails application.

When you have tree based data in your application, you always to struggle with retrieving data. There are solutions, but the always come at a price:

• Nested Set is fast at retrieval, but when inserting you might have to rearrange bounds, which can be very complex
• Closure_Tree stores additional data in a separate table, which has be kept up to date

Luckily, there is already a SQL standard that makes it very easy to retrieve data in the traditional parent/child relation. Currently this is only supported in sqlite and Postgres. With this it is possible to query complete trees without the need of extra tables or indices.

ActsAsRecursiveTree currently supports following ActiveRecord versions and is tested for compatibility:

• ActiveRecord 7.0.x
• ActiveRecord 7.1.x
• ActiveRecord 7.2.x
• ActiveRecord NEXT (from git)

ActsAsRecursiveTree is tested with following rubies:

• MRuby 3.1
• MRuby 3.2
• MRuby 3.3

Other Ruby implementations are not tested, but should also work.

Add this line to your application's Gemfile:

gem 'acts_as_recursive_tree'

And then execute:

$ bundle

Or install it yourself as:

$ gem install acts_as_recursive_tree

In your model class add following line:

class Node  < ActiveRecord::Base
  recursive_tree
end

That's it. This will assume that your model has a column named parent_id which will be used for traversal. If your column is something different, then you can specify it in the call to recursive_tree:

recursive_tree parent_key: :some_other_column

Some extra special stuff - if your parent relation is also polymorphic, then specify the polymorphic column:

recursive_tree parent_type_column: :some_other_type_column

Controlling deletion behaviour:

By default, it is up to the user code to delete all child nodes in a tree when a parent node gets deleted. This can be controlled by the :dependent option, which will be set on the children association (see #has_many in the Rails doc).

recursive_tree dependent: :nullify # or :destroy, etc.

After you set up a model for usage, there are now several methods you can use.

You have access to following associations:

• parent - the parent of this instance
• children - all children (parent_id = self.id)
• self_and_siblings - all node where parent_id = self.parent_id

• roots - all root elements (parent_id = nil)
• self_and_descendants_of(reference) - the complete tree of reference including reference in the result
• descendants_of(reference) - the complete tree of reference excluding reference in the result
• leaves_of(reference) - special case of descendants where only those elements are returned, that do not have any children
• self_and_ancestors_of(reference) - the complete ancestor list of reference including reference in the result
• ancestors_of(reference) - the complete ancestor list of reference excluding reference in the result
• roots_of(reference) - special case of ancestors where only those elements are returned, that do not have any parent

You can pass in following argument types for reference, that will be accepted:

• integer - simple integer value

    Node.descendants_of(1234)

• array - array of integer value

    Node.descendants_of([1234, 5678])

• ActiveRecord::Base - instance of an AR::Model class

    Node.descendants_of(some_node)

• ActiveRecord::Relation - an AR::Relation form the same type

    Node.descendants_of(Node.where(foo: :bar))

For nearly all mentioned scopes and associations there is a corresponding instance method:

• root - returns the root element of this node
• self_and_descendants - the complete tree including self in the result
• descendants - the complete tree excluding self in the result
• leaves - only leaves of this node
• self_and_ancestors - the complete ancestor list including self in the result
• ancestors - the complete ancestor list excluding self in the result

Those methods simply delegate to the corresponding scope and pass self as reference.

Additional methods:

• siblings - return all elements where parent_id = self.parent_id excluding self
• self_and_children - return all children and self as a Relation

Utility methods:

• root? - returns true if this node is a root node
• leaf? - returns true if this node is a leave node
• preload_tree - fetches all descendants of this node and assigns the proper parent/children associations. You are then able to traverse the tree through the children/parent association without querying the database again. You can also pass arguments to includes which will be forwarded when fetching records.

    node.preload_tree(includes: [:association, :another_association])

All ancestors and descendants methods/scopes can take an additional block argument. The block receives ans opts argument with which you are able to customize the recursion.

Depth

Specify a depth condition. Only the elements matching the depth are returned. Supported operations are:

• == exact match - can be Integer or Range or Array. When specifying a Range this will result in a depth BETWEEN min AND max query.
• != except - can be Integer or Array
• > greater than - only Integer
• >= greater than or equals - only Integer
• < less than - only Integer
• <= less than or equals - only Integer

Node.descendants_of(1){|opts| opts.depth == 3..6 }
node_instance.descendants{ |opts| opts.depth <= 4 }
node_instance.descendants{ |opts| opts.depth != [4, 7] }

NOTE: depth == 1 is the same as children/parent

Condition

Pass in an additional relation. Only those elements are returned where the condition query matches.

Node.descendants_of(1){|opts| opts.condition = Node.where(active: true) }
node_instance.descendants{ |opts| opts.condition = Node.where(active: true) }

NOTE: In contrast to depth, which first gathers the complete tree and then discards all non matching results, this will stop the recursive traversal when the relation is not met. Following two lines are completely different when executed:

node_instance.descendants.where(active: true) # => returns the complete tree and filters than out only the active ones
node_instance.descendants{ |opts| opts.condition = Node.where(active: true) } # => stops the recursion when encountering a non active node, which may return less results than the one above

Ordering All the ancestor methods will order the result depending on the depth of the recursion. Ordering for the descendants methods is disabled by default, but can be enabled if needed.

Node.descendants_of(1){|opts| opts.ensure_ordering! }
node_instance.descendants{ |opts| opts.ensure_ordering! }

NOTE: if there are many descendants this may cause a severe increase in execution time!

STI works out of the box. Consider following classes:

class Node  < ActiveRecord::Base
  recursive_tree
end

class SubNode  < Node
  
end

When calling ClassMethods the results depend on the class on which you call the method:

Node.descendants_of(123) # => returns Node and SubNode instances
SubNode.descendants_of(123) # => returns SubNode instances only

Instance Methods make no difference of the class from which they are called:

sub_node_instance.descendants # => returns Node and SubNode instances

As of now it is up to the user code to guarantee there will be no cycles created in the parent/child entries. If not, your DB might run into an endless recursion. Inserting/updating records that will cause a cycle is not prevented by some validation checks, so you have to do this by your own. This might change in a future version.

If you want to make sure to not run into an endless recursion when querying, then there are following options:

1. Add a maximum depth to the query options. If an cycle is present in your data, the recursion will stop when reaching the max depth and stop further traversing.
2. When you are on recent version of PostgreSQL (14+) you are lucky. Postgres added the CYCLE detection feature to detect cycles and prevent endless recursion. Our query builder will add this feature if your DB does support this.

1. Fork it ( https://github.com/1and1/acts_as_recursive_tree/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request