Rethinkdb client in pure elixir (JSON protocol)
Latest commit 4dad32a Dec 14, 2016 @hamiltop committed on GitHub Merge pull request #119 from slofurno/polygon-holes
fix parsing polygons with multiple holes

RethinkDB Build Status

Multiplexed RethinkDB client in pure Elixir.

If you are coming here from elixir-rethinkdb, welcome! If you were expecting Exrethinkdb you are in the right place. We decided to change the name to just RethinkDB and the repo to rethinkdb-elixir. Sorry if it has caused confusion. Better now in the early stages than later!

I just set up a channel on the Elixir slack, so if you are on there join #rethinkdb.

Recent changes


  • Extract Changefeed out into separate package
  • Accept keyword options with queries

Getting Started

See API documentation for more details.


Connections are managed by a process. Start the process by calling start_link/1. See documentation for Connection.start_link/1 for supported options.

Basic Remote Connection

{:ok, conn} = RethinkDB.Connection.start_link([host: "", port: 28015])

Named Connection

{:ok, conn} = RethinkDB.Connection.start_link([name: :foo])

Supervised Connection

Start the supervisor with:

worker(RethinkDB.Connection, [[name: :foo]])
worker(RethinkDB.Connection, [[name: :bar, host: 'localhost', port: 28015]])

Default Connection

An RethinkDB.Connection does parallel queries via pipelining. It can and should be shared among multiple processes. Because of this, it is common to have one connection shared in your application. To create a default connection, we create a new module and use RethinkDB.Connection.

defmodule FooDatabase do
  use RethinkDB.Connection

This connection can be supervised without a name (it will assume the module as the name).

worker(FooDatabase, [])

Queries can be run without providing a connection (it will use the name connection).

import RethinkDB.Query
table("people") |>

Connection Pooling

To use a connection pool, add Poolboy to your dependencies:

{:poolboy, "~> 1.5"}

Then, in your supervision tree, add:

worker(:poolboy, [[name: {:local, :rethinkdb_pool}, worker_module: RethinkDB.Connection, size: 10, max_overflow: 0], [])

NOTE: If you want to use changefeeds or any persistent queries, max_overflow: 0 is required.

Then use it in your code:

db = :poolboy.checkout(:rethinkdb_pool)
table("people") |> db
:poolboy.checkin(:rethinkdb_pool, db)

Query accepts a process as the second argument (to facilitate piping).


q = Query.table("people")
  |> Query.insert(%{first_name: "John", last_name: "Smith"})
  |> conn


q = Query.table("people")
  |> Query.filter(%{last_name: "Smith"})
  |> conn


RethinkDB supports RethinkDB functions in queries. There are two approaches you can take:

Use RethinkDB operators

import RethinkDB.Query

make_array([1,2,3]) |> map(fn (x) -> add(x, 1) end)

Use Elixir operators via the lambda macro

require RethinkDB.Lambda
import RethinkDB.Lambda

make_array([1,2,3]) |> map(lambda fn (x) -> x + 1 end)


require RethinkDB.Lambda
import Query
import RethinkDB.Lambda

conn = RethinkDB.connect

  |> has_fields(["first_name", "last_name"])
  |> map(lambda fn (person) ->
    person[:first_name] + " " + person[:last_name]
  end) |> conn

See query.ex for more basic queries. If you don't see something supported, please open an issue. We're moving fast and any guidance on desired features is helpful.


# Simple indexes
# create
result = Query.table("people")
  |> Query.index_create("first_name", Lambda.lambda fn(row) -> row["first_name"] end)
  |> conn

# retrieve
result = Query.table("people")
  |> Query.get_all(["Will"], index: "first_name")
  |> conn

# Compound indexes
# create
result = Query.table("people")
  |> Query.index_create("full_name", Lambda.lambda fn(row) -> [row["first_name"], row["last_name"]] end)
  |> conn

# retrieve
result = Query.table("people")
  |> Query.get_all([["Will", "Smith"], ["James", "Bond"]], index: "full_name")
  |> conn

One limitation we have in Elixir is that we don't support varargs. So in JavaScript you would do getAll(key1, key2, {index: "uniqueness"}). In Elixir we have to do get_all([key1, key2], index: "uniqueness"). With a single key it becomes get_all([key1], index: "uniqueness") and when key1 is [partA, partB] you have to do get_all([[partA, partB]], index: "uniqueness")


Change feeds can be consumed either incrementally (by calling or via the Enumerable Protocol.

results = Query.table("people")
  |> Query.filter(%{last_name: "Smith"})
  |> Query.changes
  |> conn
# get one result
first_change = results
# get stream, chunked in groups of 5, Inspect
results |> Stream.chunk(5) |> Enum.each &IO.inspect/1

Supervised Changefeeds

Supervised Changefeeds (an OTP behavior for running a changefeed as a process) have been moved to their own repo to enable independent release cycles. See


Version 1.0.0 will be limited to individual connections and implement the entire documented ReQL (as of rethinkdb 2.0)

While not provided by this library, we will also include example code for:

  • Connection Pooling

The goal for 1.0.0 is to be stable. Issues have been filed for work that needs to be completed before 1.0.0 and tagged with the 1.0.0 milestone.

Example Apps

Checkout the wiki page for various example apps


Contributions are welcome. Take a look at the Issues. Anything that is tagged Help Wanted or Feedback Wanted is a good candidate for contributions. Even if you don't know where to start, respond to an interesting issue and you will be pointed in the right direction.


Be intentional. Whether you are writing production code or tests, make sure there is value in the test being written.