A Pipeable Logger

The Elixir Logger is pretty good. You can easily log anything with it, just call one of debug, info, warn or error. For example:

1
Logger.info("something happened")

Turns into:

12:34:56.789 [info]  something happened

Very nice. However, there are cases where you may want to, say, change some data structure, like update a map or a list, and then log the transition, without breaking the pipe. Example:

1
2
3
4
5
6
7
8
def my_function do
  list = [1, 2, 3]

  list
  |> Logger.debug("before insert: #{inspect list}")
  |> Enum.into([0])
  |> Logger.debug("after insert: #{inspect list}")
end

This doesn’t work for many reasons. First, we can’t refer to list that way. If we do, we will always be logging [1, 2, 3], because Elixir’s data structures are immutable. Second, Logger.* functions return the :ok atom, which means you can’t use them in a pipe—unless that is what you want to return.

The solution to both issues is actually pretty straightforward: use a lambda! A lambda is just an anonymous function. We can define it and call it right away. So the code above becomes:

1
2
3
4
5
6
7
8
9
10
11
12
def my_function do
  [1, 2, 3]
  |> (fn list ->
    Logger.debug("before insert: #{inspect list}")
    list
  end).()
  |> Enum.into([0])
  |> (fn list ->
    Logger.debug("after insert: #{inspect list}")
    list
  end).()
end

If we call this function, we get:

12:34:56.789 [debug] before insert: [1, 2, 3]

12:34:56.823 [debug] after insert: [0, 1, 2, 3]

Great, exactly what we want! Except the syntax is horrible. But fear not, we can improve on it. How about we make a wrapper?

1
2
3
4
5
6
7
8
9
10
11
12
13
defmodule PipeableLogger do
  require Logger

  def debug(data, msg) do
    Logger.debug(msg)
    data
  end

  # def warn, do: ...
  # def error, do: ...
  # def info, do: ...

end

Let’s rewrite our function once again:

1
2
3
4
5
6
def my_function do
  [1, 2, 3]
  |> (&PipeableLogger.debug(&1, "before insert: #{inspect &1}")).()
  |> Enum.into([0])
  |> (&PipeableLogger.debug(&1, "after insert: #{inspect &1}")).()
end

Still not pretty though, as we still needed to wrap the function in a lambda. If we want to build a proper Logger wrapper, there are at least two different cases we may want to handle:

  1. Logging a simple message (without any data);
  2. Logging the data we receive from the pipe, maybe also with a message.

Here’s the improved version of PipeableLogger:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
defmodule PipeableLogger do
  require Logger

  def debug(data, msg \\ "", metadata \\ [])
  def debug(data, msg, metadata) when msg == "", do: Logger.debug(data, metadata)
  def debug(data, msg, metadata) do
    Logger.debug(msg, metadata)
    data
  end

  # def warn, do: ...
  # def error, do: ...
  # def info, do: ...

end

Let’s use it:

1
2
3
4
5
6
def my_function do
  [1, 2, 3]
  |> PipeableLogger.debug("before insert")
  |> Enum.into([0])
  |> PipeableLogger.debug("after insert")
end

Much, much simpler! The only problem now is, we’re logging just a message. What if we want to log the data? It’s a lambda all over again.

Here’s the final version I came up with:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
defmodule PipeableLogger do
  require Logger

  def debug(data, msg \\ "", metadata \\ [])
  def debug(data, msg, metadata) when msg == "", do: Logger.debug(data, metadata)
  def debug(data, msg, metadata) when is_binary(data) do
    Logger.debug(msg <> data, metadata)
    data
  end
  def debug(data, msg, metadata) do
    Logger.debug(msg <> inspect(data), metadata)
    data
  end

  # def warn, do: ...
  # def error, do: ...
  # def info, do: ...

end

The assumption is that we always want to concatenate the data with the message, which is fair enough I think. Let’s see it in action:

1
2
3
4
5
6
def my_function do
  [1, 2, 3]
  |> PipeableLogger.debug("before insert: ")
  |> Enum.into([0])
  |> PipeableLogger.debug("after insert: ")
end
1
2
3
4
5
6
iex> my_function()

12:34:56.789 [debug] before insert: [1, 2, 3]

12:34:56.789 [debug] after insert: [0, 1, 2, 3]
[0, 1, 2, 3]

Now we can log the data with a message, all in a pipe and without a lambda! Nice!


Summing up, I’m not convinced a Logger wrapper is the right way. This kinda goes against the blog post, but to be fair I think Elixir people tend to use pipes way too much (I’m guilty as well). So I wouldn’t probably wrap Logger in any project.

It’s also worth noting that Logger supports the concept of metadata, which basically means you can already attach any data you want. For example, if you put this in your config.exs:

1
2
config :logger, :console,
  metadata: [:my_list]

You can then call Logger like this:

1
2
3
4
5
6
7
iex(1)> require Logger
Logger

iex(2)> Logger.info "Work done", my_list: inspect [1, 2, 3]

12:34:56.789 my_list=[1, 2, 3] [info]  Work done
:ok

Point is, you don’t need a wrapper if all you want is concatenate some data in the log message. You do need a wrapper though (or a lambda) if you want to use Logger in a pipe.

So how about this instead?

1
2
3
4
5
6
def my_function do
  list = [1, 2, 3]
  Logger.debug("before insert: #{inspect list}")
  new_list = Enum.into(list, [0])
  Logger.debug("after insert: #{inspect new_list}")
end

Simple is better. It’s fine to break that pipe every once in a while!

Comments