Skip to content

jclem/logfmt-elixir

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Logfmt

Build Status Module Version Hex Docs Total Download License Last Updated

Logfmt is a module for encoding and decoding logfmt-style log lines.

Installation

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

def deps do
  [
    {:logfmt, "~> 3.3"}
  ]
end

Usages

Decode log lines into maps:

iex> Logfmt.decode "foo=bar"
%{"foo" => "bar"}

Encode Dict implementation values into log lines:

iex> Logfmt.encode [foo: "bar"]
"foo=bar"

iex> Logfmt.encode %{foo: "bar"}
"foo=bar"

Custom types can encoded by implementing the ValueEncoder protocol for it.

For example to encode DateTime and NaiveDateTime and implementation could look like this:

defimpl Logfmt.ValueEncoder, for: NaiveDateTime do
  def encode(naive_date_time), do: NaiveDateTime.to_iso8601(naive_date_time)
end

defimpl Logfmt.ValueEncoder, for: DateTime do
  def encode(date_time), do: DateTime.to_iso8601(date_time)
end

Type Coercion

When decoding a log line, Logfmt will coerce some strings into booleans and numbers:

iex> Logfmt.decode "foo=true"
%{"foo" => true}

iex> Logfmt.decode "foo=-1.2e9"
%{"foo" => -1.2e9}

In the future, this may be optional, or more robust. For example, it might make sense for "foo=true" to decode into %{"foo" => true}, but ~s(foo="true") to decode into %{"foo" => "true"}.

Another option might be to allow the user to provide a formatting map to the decode function, which expects coercion functions as values:

iex> "foo=1 bar=2" |> Logfmt.decode %{
...> foo: &Logfmt.TypeCoercion.parse_integer/1
...> }
%{"foo" => 1, "bar" => "2"}

Why decode into maps?

Originally, this library both decoded and encoded maps. However, this was problematic because key ordering in maps is not guaranteed. A developer wants to be able to ensure that their log output will have identical ordering for multiple calls for the sake of readability.

To solve this, the second version encoded and decoded Keyword lists only. Of course, this is also problematic because decoding log lines into Keyword lists involves converting user strings into non-garbage-collected atoms.

Now, this module decodes into maps only (with string keys) and encodes any Dict implementation type. This is a fair compromise, because ordering upon decoding a Logfmt line is not important, and keeping only the last value for a duplicate key in a log line is fair, as well.