Skip to content

Latest commit

 

History

History
211 lines (151 loc) · 7.81 KB

README.md

File metadata and controls

211 lines (151 loc) · 7.81 KB

as_nested_set

Build Status Coveralls Coverage Hex.pm

An ecto based Nested set model implementation for database

Installation

Add as_nested_set to your list of dependencies in mix.exs:

  # use the stable version
  def deps do
    [{:as_nested_set, "~> 3.4"}]
  end

  # use the latest version
  def deps do
    [{:as_nested_set, github: "https://github.com/secretworry/as_nested_set.git"}]
  end

Usage

To make use of as_nested_set, 4 fields( id, lft, rgt and parent_id) are required for your model. The name of those fields are configurable.

defmodule AsNestedSet.TestRepo.Migrations.MigrateAll do
  use Ecto.Migration

  def change do
    create table(:taxons) do
      add :name, :string
      add :taxonomy_id, :id
      add :parent_id, :id
      add :lft, :integer
      add :rgt, :integer

      timestamps
    end

  end
end

Enable the nested set functionality by use AsNestedSet on your model

defmodule AsNestedSet.Taxon do
  use AsNestedSet, scope: [:taxonomy_id]
  # ...
end

Options

You can config the name of required fields through attributes:

defmodule AsNestedSet.Taxon do
  @right_column :right
  @left_column :left
  @node_id_column :node_id
  @parent_id_column :pid
  # ...
end
  • @right_column: column name for the right boundary (defaults to lft, since left is a reserved keyword for Mysql)
  • @left_column: column name for the left boundary (defaults to rgt, reserved too)
  • @node_id_column: specifies the name for the node id column (defaults to id, i.e. the id of the model, change this if you want to have a different id, or you id field is different)
  • @parent_id_column: specifies the name for the parent id column (defaults to parent_id)

You can also pass following arguments to modify its behavior:

defmodule AsNestedSet.Taxon do
  use AsNestedSet, scope: [:taxonomy_id]
  # ...
end
  • scope: (optional) a list of column names which restrict what are to be considered within the same tree(same scope). When ignored, all the nodes will be considered under the same tree.

Model Operations

Once you have set up you model, you can then

Add new nodes

target = Repo.find!(Taxon, 1)
# add to left
%Taxon{name: "left", taxonomy_id: 1} |> AsNestedSet.create(target, :left) |> AsNestedSet.execute(TestRepo)
# add to right
%Taxon{name: "right", taxonomy_id: 1} |> AsNestedSet.create(target, :right) |> AsNestedSet.execute(TestRepo)
# add as first child
%Taxon{name: "child", taxonomy_id: 1} |> AsNestedSet.create(target, :child) |> AsNestedSet.execute(TestRepo)
# add as parent
%Taxon{name: "parent", taxonomy_id: 1} |> AsNestedSet.create(target, :parent) |> AsNestedSet.execute(TestRepo)

# add as root
%Taxon{name: "root", taxonomy_id: 1} |> AsNestedSet.create(:root) |> AsNestedSet.execute(TestRepo)

# move a node to a new position

node |> AsNestedSet.move(:root) |> AsNestedSet.execute(TestRepo) // move the node to be a new root
node |> AsNestedSet.move(target, :left) |> AsNestedSet.execute(TestRepo) // move the node to the left of the target
node |> AsNestedSet.move(target, :right) |> AsNestedSet.execute(TestRepo) // move the node to the right of the target
node |> AsNestedSet.move(target, :child) |> AsNestedSet.execute(TestRepo) // move the node to be the right-most child of target

Remove a specified node and all its descendants

target = Repo.find!(Taxon, 1)
AsNestedSet.delete(target) |> AsNestedSet.execute(TestRepo)

Query different nodes

# find all roots
AsNestedSet.roots(Taxon, %{taxonomy_id: 1}) |> AsNestedSet.execute(TestRepo)

# find all children of target
AsNestedSet.children(target) |> AsNestedSet.execute(TestRepo)

# find all the leaves for given scope
AsNestedSet.leaves(Taxon, %{taxonomy_id: 1}) |> AsNestedSet.execute(TestRepo)

# find all descendants
AsNestedSet.descendants(target) |> AsNestedSet.execute(TestRepo)
# include self
AsNestedSet.self_and_descendants(target) |> AsNestedSet.execute(TestRepo)

# find all ancestors
AsNestedSet.ancestors(target) |> AsNestedSet.execute(TestRepo)

#find all siblings (self included)
AsNestedSet.self_and_siblings(target) |> AsNestedSet.execute(TestRepo)

Traverse the tree

# traverse a tree with 3 args post callback
AsNestedSet.traverse(Taxon, %{taxonomy_id}, context, fn node, context -> {node, context}, end, fn node, children, context -> {node, context} end) |> AsNestedSet.execute(TestRepo)
# traverse a tree with 2 args post callback
AsNestedSet.traverse(Taxon, %{taxonomy_id}, context, fn node, context -> {node, context}, end, fn node, context -> {node, context} end) |> AsNestedSet.execute(TestRepo)

# traverse a subtree with 3 args post callback
AsNestedSet.traverse(target, context, fn node, context -> {node, context}, end, fn node, children, context -> {node, context} end) |> AsNestedSet.execute(TestRepo)
# traverse a tree with 2 args post callback
AsNestedSet.traverse(target, context, fn node, context -> {node, context}, end, fn node, context -> {node, context} end) |> AsNestedSet.execute(TestRepo)

FAQ

How to ensure the consistency

We recommend users to use a transaction to wrap all the operations in the production environment

We introduced the @type executable( a delayed execution ) as the return value of each API, so using transaction or not and how granular the transaction should be are all up to users.

In general, almost all modifications of a nested set can be done in one SQL, but we can't express some of them using ecto's DSL( ecto doesn't support case-when in update query ), so users having concurrent modifications must wrap AsNestedSet.execute(call, repo) in a Transaction, for example

exec = node |> AsNestedSet.move(:root)
Repo.transaction fn -> AsNestedSet.execute(exec, Repo) end

The node passed in as an argument might has changed after loaded from db, we will reload it from DB before use it, so there is no need to wrap the load and the execute in the same transaction

# This is not necessary
Repo.transaction fn ->
	node = loadNode()
	node |> AsNestedSet.move(:root) |> AsNestedSet.execute(Repo) # We will reload the node passed in
end

But if you want to ensure consistency across multiple executes , to avoid the racing condition, you have to isolate them by wrap them in different transactions.

How to move a node to be the n-th child of a target

Be default, after using AsNestedSet.move(node, target, :child), you move the node to be the right-most child of the target, because we can know the left and right of the target right way, but to find out the proper right and left for n-th child requires more operations.

To achieve the goal, you should:

  1. Query the n-th child or (n-1)th child of the target by AsNestedSet.children(target),
  2. Use move(node, n_th_child, :left) and move(node, n_1_th_child, :right) respectively.

Ecto 2.x

Ecto is upgrading to 3.0, with a clear API and a lot lot of bug fixes. Please consider to upgrade for your projects too:) We will not support 2.x in our public releases, but if you are using Ecto 2.x, you can get the latest updates by using branch ecto-2.x

Contributors

Special thanks

  • Thanks Travis CI for providing free and convenient integration test