ReqSandbox - Ecto SQL Sandbox requests for external HTTP clients

Photo of a toy truck in a sandbox
Photo by Takeshi Morisato on Unsplash

Req plugin for Phoenix.Ecto.SQL.Sandbox.

ReqSandbox simplifies making concurrent, transactional requests to a Phoenix server. Just before making a request, the sandbox metadata is applied via the specified request header. If there is no metadata available, then ReqSandbox creates a new sandbox session and saves the metadata for future requests. This is mostly useful in client test environments to ensure logical isolation between concurrent tests.


The Ecto SQL Sandbox Usage Guide included in the docs is a Livebook with an interactive demonstration of all the sandbox features.

  {:req, "~> 0.3.0"},
  {:req_sandbox, "~> 0.1.0"}

req = "http://localhost:4000" |> ReqSandbox.attach()!(req, url: "/api/posts", json: %{"post" => %{"msg" => "Hello, world!"}}).body
# => %{"data" => %{"id" => 2, "msg" => "Hello, world!"}}

# => "BeamMetadata (g2gCZAACdjF0AAAAA2QABW93bmVyWGQAInZ2ZXMzM2o1LWxpdmVib29...)"


If you are writing HTTP clients in Elixir to communicate with Phoenix/Plug + Ecto services, then this package is for you! Sandboxing client requests is an effective way to provide developers with an integration environment that can be operated concurrently while maintaining logical isolation between development sessions.


im surprised this doesn’t just add $callers into the header, that way the http request also has access to mox, and whatever else in the ecosystem gets that kind of concurrency support, in stead of just ecto

There’s no long running pid to set $callers to for the phoenix sandbox. It’s meant to work for e2e tests for tools, which might not even run on the beam, which checkout and checkin db connections statelessly as far as the beam is concerned.

1 Like

Yeah ReqSandbox is only useful if you have an Elixir client communicating with a Phoenix/Plug server over HTTP. The client initiates the sandbox session and is responsible for including the given metadata on each request.

I decided to make ReqSandbox because anecdotally it’s a pain to account for the header value all the way thru your client application, especially if your app fires off HTTP requests in an ad-hoc fashion rather than passing around a client/request struct. Further, if you are using the sandbox to support integration tests it’s rather important to ensure that all of your requests are properly sandboxed, otherwise you run the risk of tainting your integration database with un-sandboxed operations.

v0.1.2 (2022-12-22)

  • Removes POST body content before making the sandbox request.

  • Adds missing content-length header on the sandbox request.

  • Adds ReqSandbox.token/0 to fetch the current token if it exists.

1 Like