Skip to content

Fire-Dragon-DoL/sonda

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Sonda

Sonda is a telemetry library for Elixir, providing configurable sinks for recording signals.

By default, Sonda is configured to record signals in memory and can be inspected through the provided utility functions.

{:ok, pid} = Sonda.start_link()
Sonda.record(pid, :a_signal, %{"some" -> "data"})
Sonda.recorded_once?(pid, &match?({:a_signal, _, _}, &1)) # => true
Sonda.record(pid, :another_signal, 123)
Sonda.record(pid, :signal_only)
Sonda.records(pid)
# [
#    {:a_signal, ~N[2020-04-22 23:07:05.905776], %{"some" -> "data"}},
#    {:another_signal, ~N[2020-04-22 23:07:06.905776], 123}},
#    {:signal_only, ~N[2020-04-22 23:07:07.905776], nil}
# ]

Installation

The package can be installed by adding sonda to your list of dependencies in mix.exs:

def deps do
  [
    {:sonda, "~> 0.1.0"}
  ]
end

Docs can be found at https://hexdocs.pm/sonda.

Basic Usage

The default configuration of Sonda provides all the utility functions for recording signals in memory and then inspecting the output:

Start

{:ok, pid} = Sonda.start_link()

Record signals

Sonda.record(pid, :a_signal, %{"some" -> "data"})
Sonda.record(pid, :another_signal, 123)
Sonda.record(pid, :signal_only)

Inspect

Was any message ever recorded with this shape?

Sonda.recorded?(pid, &match?({:a_signal, _, _}, &1)) # => true

Was any message ever recorded with this shape only one time?

Sonda.recorded_once?(pid, &match?({:a_signal, _, _}, &1)) # => true

List all the recorded messages starting from the oldest

Sonda.records(pid)
# [
#    {:a_signal, ~N[2020-04-22 23:07:05.905776], %{"some" -> "data"}},
#    {:another_signal, ~N[2020-04-22 23:07:06.905776], 123}},
#    {:signal_only, ~N[2020-04-22 23:07:07.905776], nil}
# ]

List all the recorded messages starting from the oldest and matching the provided function

Sonda.records(pid, fn {_, _, data} -> data != 123 end)
# [
#    {:a_signal, ~N[2020-04-22 23:07:05.905776], %{"some" -> "data"}},
#    {:signal_only, ~N[2020-04-22 23:07:07.905776], nil}
# ]

Gets the first recorded message matching the provided function, only if there is only one. An error is returned in all the other cases

Sonda.one_record(pid, fn {_, _, data} -> data == 123 end)
# {:ok, {:another_signal, ~N[2020-04-22 23:07:06.905776], 123}}}

Gets the first recorded message matching the provided function, only if there is only one. An error is returned in all the other cases

Sonda.one_record(pid, fn {_, _, data} -> data == 123 end)
# {:ok, {:another_signal, ~N[2020-04-22 23:07:06.905776], 123}}}

Is this signal allowed to be recorded?

Sonda.record_signal?(pid, :a_signal) # => true

See the Filtering Signals section for more information about why a signal might be recorded or not. This output of this function is false for any signal that's not accepted.

Filtering Signals

It's possible to ignore some signals from being recorded. It's sufficient to start Sonda with the following configuration:

{:ok, pid} = Sonda.start_link(signals: [:a_signal, :another_signal])

The output of the function Sonda.record_signal?/2 is true only for the signals :a_signal and :another_signal, like in the following example:

{:ok, pid} = Sonda.start_link(signals: [:a_signal, :another_signal])

Sonda.record_signal?(pid, :a_signal) # => true
Sonda.record_signal?(pid, :another_signal) # => true
Sonda.record_signal?(pid, :not_recorded) # => false

When the input of the functions record/2 and record/3 is a signal for which the output of Sonda.record_signal?/2 is false, no operation is performed. The following example shows that :not_recorded signal is ignored:

{:ok, pid} = Sonda.start_link(signals: [:a_signal, :another_signal])

Sonda.record(pid, :a_signal, %{"some" -> "data"})
Sonda.record(pid, :not_recorded)
Sonda.record(pid, :another_signal, 123)

Sonda.records(pid)
# [
#    {:a_signal, ~N[2020-04-22 23:07:05.905776], %{"some" -> "data"}},
#    {:another_signal, ~N[2020-04-22 23:07:06.905776], 123}}
# ]

Notice that the default Sonda configuration when no parameters are passed to start_link is equivalent to passing the :any option to signals, like in the following example:

{:ok, pid} = Sonda.start_link(signals: :any)
# Equivalent to
{:ok, pid} = Sonda.start_link()

Common Usage

ExUnit

One of the most common purpose of Sonda is to inspect the usage of closures or the behavior of GenServers. The following example shows the closure being invoked twice by inspecting the signals recorded through Sonda:

defmodule SomeTest do
  use ExUnit.Case

  test "Enum.map/2 is invoked once per list element" do
    sonda = start_supervised!(Sonda)
    elements = [1, 2]

    elements = Enum.map(elements, fn element ->
      Sonda.record(sonda, :map, element)
      element * 2
    end)
    records = Sonda.records(sonda)

    assert elements == [2, 4]
    assert length(records) == 2
  end
end

Advanced Usage

Sonda Agent

When Sonda is started using Sonda.start_link/*, a Sonda.Agent is started behind the scenes, holding the state of a sink. By default, a Sonda.Sink.Proxy is used, which is a proxy to multiple sinks. The default configuration provided to Sonda.Sink.Proxy holds only one sink: Sonda.Sink.Memory, which provides the functionality of appending and inspecting messages.

The Sonda module

The Sonda module delegates all the work to a special Sonda.Agent called Sonda.Agent.Default. Sonda utility functions can be used as long as the configuration of the Sonda.Agent uses a Sonda.Sink.Proxy with the first sink being a Sonda.Sink.Memory.

This is not a limitation, feel free to create your own utility module and use that rather than the Sonda module.

The Sink protocol

Sonda doesn't make assumptions on what you want to do with the signals recorded. It's sufficient to implement the Sonda.Sink protocol to record signals in the data structure of your choice. The sink data structure is stored in Sonda.Agent, so the data is kept as long as the process is running, without the need of having to implement a GenServer.

The sink protocol requires the following function to be implemented:

record(sink, signal, timestamp, data)

  • sink the data structure implementing Sonda.Sink
  • signal an atom
  • timestamp a NaiveDateTime
  • data any type of data

This function should handle the message in any way you see fit. Sonda uses Sonda.Sink.Memory behind the scenes to provide the functionality of, appending messages and inspecting them.

Sonda.Sink.Proxy

Sonda provides out of the box support for using multiple sinks. When starting Sonda, the option to provide an alternative list of sinks can be provided like in the following example:

{:ok, pid} = Sonda.start_link(sinks: [a_sink, another_sink])

Notice that if the first sink is not a Sonda.Sink.Memory, all inspection functions on the Sonda module will not operate correctly.

The following configuration preserves the default functionality of Sonda as well as expanding it with other sinks you see fit (for example, recording messages to a database):

memory_sink = Sonda.Sink.Memory.configure(signals: :any)
other_sink = %OtherSink{} # This is the sink implemented by you
{:ok, pid} = Sonda.start_link(sinks: [memory_sink, other_sink])

Custom timestamps

By default, Sonda uses NaiveDateTime.utc_now/0 to determine the timestamp of a recorded message. This behavior can be altered by providing the option :clock_now when starting Sonda like in the following example:

{:ok, pid} = Sonda.start_link(clock_now: fn -> DateTime.utc_now() end)
# Timestamps will now be of type DateTime

Notice that Sonda is just passing down this option to Sonda.Agent, which is the module actually supporting the :clock_now option.

Using a custom Sonda.Agent

By using a custom Sonda.Agent and a custom sink implementing Sonda.Sink protocol, it's possible to greatly alter the behavior of Sonda, giving access to the lowest levels of machinery available in the library.

Sonda.Agent is a GenServer that holds a Sonda.Agent.clock function to provide timestamps and a Sonda.Sink data structure.

Only two functions are provided:

record(server, signal, data)

  • server is the identifier for the Agent, usually a PID
  • signal is an atom for the message
  • data is any type of data that is attached to the message

This function is used to call record on the sink stored inside this agent.

get_sink(server, sink_block)

  • server is the identifier for the Agent, usually a PID
  • sink_block is a function which accepts one argument: the sink stored in the agent. It can return any type of data, which in turn will be returned by get_sink/2

This function is used to access the sink, to perform any reading operation.

Thanks

This library is heavily influenced by telemetry, the Eventide library for ruby instrumentation.

About

Telemetry recording for testing in Elixir

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages