defmodule Tesla.Middleware.Logger.Formatter do
@moduledoc false
# Heavily based on Elixir's Logger.Formatter
# https://github.com/elixir-lang/elixir/blob/v1.6.4/lib/logger/lib/logger/formatter.ex
@default_format "$method $url -> $status ($time ms)"
@keys ~w(method url status time query)
@type format :: [atom | binary]
@spec compile(binary | nil) :: format
@spec compile(pattern) :: pattern when pattern: function | {module, fun :: atom}
def compile(nil), do: compile(@default_format)
def compile(fun) when is_function(fun), do: fun
def compile({mod, fun}) when is_atom(mod) and is_atom(fun), do: {mod, fun}
def compile(binary) do
~r/(?<h>)\$[a-z]+(?<t>)/
|> Regex.split(binary, on: [:h, :t], trim: true)
|> Enum.map(&compile_key/1)
end
defp compile_key("$" <> key) when key in @keys, do: String.to_atom(key)
defp compile_key("$" <> key), do: raise(ArgumentError, "$#{key} is an invalid format pattern.")
defp compile_key(part), do: part
@spec format(
Tesla.Env.t(),
Tesla.Env.result(),
integer,
format | function | {module, atom}
) :: IO.chardata()
def format(request, response, time, fun) when is_function(fun) do
apply(fun, [request, response, time])
end
def format(request, response, time, {mod, fun}) do
apply(mod, fun, [request, response, time])
end
def format(request, response, time, format) do
Enum.map(format, &output(&1, request, response, time))
end
defp output(:query, env, _, _), do: env.query |> Tesla.encode_query()
defp output(:method, env, _, _), do: env.method |> to_string() |> String.upcase()
defp output(:url, env, _, _), do: env.url
defp output(:status, _, {:ok, env}, _), do: to_string(env.status)
defp output(:status, _, {:error, reason}, _), do: "error: " <> inspect(reason)
defp output(:time, _, _, time), do: :io_lib.format("~.3f", [time / 1000])
defp output(binary, _, _, _), do: binary
end
defmodule Tesla.Middleware.Logger do
@moduledoc ~S"""
Log requests using Elixir's Logger.
With the default settings it logs request method, URL, response status, and time taken in milliseconds.
## Examples
```elixir
defmodule MyClient do
use Tesla
plug Tesla.Middleware.Logger
end
```
## Options
- `:log_level` - custom function for calculating log level (see below)
- `:filter_headers` - sanitizes sensitive headers before logging in debug mode (see below)
- `:debug` - show detailed request/response logging
- `:format` - custom string template or function for log message (see below)
## Custom log format
The default log format is `"$method $url -> $status ($time ms)"`
which shows in logs like:
```
2018-03-25 18:32:40.397 [info] GET https://bitebot.io -> 200 (88.074 ms)
```
It can be changed globally with config:
```elixir
config :tesla, Tesla.Middleware.Logger, format: "$method $url ====> $status / time=$time"
```
Or you can customize this setting by providing your own `format` function:
```elixir
defmodule MyClient do
use Tesla
plug Tesla.Middleware.Logger, format: &my_format/3
def my_format(request, response, time) do
"request=#{inspect(request)} response=#{inspect(response)} time=#{time}\n"
end
end
```
## Custom log levels
By default, the following log levels will be used:
- `:error` - for errors, 5xx and 4xx responses
- `:warn` or `:warning` - for 3xx responses
- `:info` - for 2xx responses
You can customize this setting by providing your own `log_level/1` function:
```elixir
defmodule MyClient do
use Tesla
plug Tesla.Middleware.Logger, log_level: &my_log_level/1
def my_log_level(env) do
case env.status do
404 -> :info
_ -> :default
end
end
end
```
## Logger Debug output
When the Elixir Logger log level is set to `:debug`
Tesla Logger will show full request & response.
If you want to disable detailed request/response logging
but keep the `:debug` log level (i.e. in development)
you can set `debug: false` in your config:
```
# config/dev.local.exs
config :tesla, Tesla.Middleware.Logger, debug: false
```
Note that the logging configuration is evaluated at compile time,
so Tesla must be recompiled for the configuration to take effect:
```
mix deps.clean --build tesla
mix deps.compile tesla
```
In order to be able to set `:debug` at runtime we can
pass it as a option to the middleware at runtime.
```elixir
def client do
middleware = [
# ...
{Tesla.Middleware.Logger, debug: false}
]
Tesla.client(middleware)
end
```
### Filter headers
To sanitize sensitive headers such as `authorization` in
debug logs, add them to the `:filter_headers` option.
`:filter_headers` expects a list of header names as strings.
```
# config/dev.local.exs
config :tesla, Tesla.Middleware.Logger,
filter_headers: ["authorization"]
```
"""
@behaviour Tesla.Middleware
alias Tesla.Middleware.Logger.Formatter
@config Application.compile_env(:tesla, __MODULE__, [])
@format Formatter.compile(@config[:format])
@type log_level :: :info | :warn | :warning | :error
if Version.compare(System.version(), "1.11.0") == :lt do
@warning_level :warn
else
@warning_level :warning
end
require Logger
@impl Tesla.Middleware
def call(env, next, opts) do
{time, response} = :timer.tc(Tesla, :run, [env, next])
config = Keyword.merge(@config, opts)
optional_runtime_format = Keyword.get(config, :format)
format =
if optional_runtime_format, do: Formatter.compile(optional_runtime_format), else: @format
level = log_level(response, config)
Logger.log(level, fn -> Formatter.format(env, response, time, format) end)
if Keyword.get(config, :debug, true) do
Logger.debug(fn -> debug(env, response, config) end)
end
response
end
defp log_level({:error, _}, _), do: :error
defp log_level({:ok, env}, config) do
case Keyword.get(config, :log_level) do
nil ->
default_log_level(env)
fun when is_function(fun) ->
case fun.(env) do
:default -> default_log_level(env)
warning when warning in [:warn, :warning] -> @warning_level
level -> level
end
warning when warning in [:warn, :warning] ->
@warning_level
atom when is_atom(atom) ->
atom
end
end
@spec default_log_level(Tesla.Env.t()) :: log_level
def default_log_level(env) do
cond do
env.status >= 400 -> :error
env.status >= 300 -> @warning_level
true -> :info
end
end
@debug_no_query "(no query)"
@debug_no_headers "(no headers)"
@debug_no_body "(no body)"
@debug_stream "[Elixir.Stream]"
defp debug(request, {:ok, response}, config) do
[
"\n>>> REQUEST >>>\n",
debug_query(request.query),
?\n,
debug_headers(request.headers, config),
?\n,
debug_body(request.body),
?\n,
"\n<<< RESPONSE <<<\n",
debug_headers(response.headers, config),
?\n,
debug_body(response.body)
]
end
defp debug(request, {:error, error}, config) do
[
"\n>>> REQUEST >>>\n",
debug_query(request.query),
?\n,
debug_headers(request.headers, config),
?\n,
debug_body(request.body),
?\n,
"\n<<< RESPONSE ERROR <<<\n",
inspect(error)
]
end
defp debug_query([]), do: @debug_no_query
defp debug_query(query) do
query
|> Enum.flat_map(&Tesla.encode_pair/1)
|> Enum.map(fn {k, v} -> ["Query: ", to_string(k), ": ", to_string(v), ?\n] end)
end
defp debug_headers([], _config), do: @debug_no_headers
defp debug_headers(headers, config) do
filtered = Keyword.get(config, :filter_headers, [])
Enum.map(headers, fn {k, v} ->
v = if k in filtered, do: "[FILTERED]", else: v
[k, ": ", v, ?\n]
end)
end
defp debug_body(nil), do: @debug_no_body
defp debug_body([]), do: @debug_no_body
defp debug_body(%Stream{}), do: @debug_stream
defp debug_body(stream) when is_function(stream), do: @debug_stream
defp debug_body(%Tesla.Multipart{} = mp) do
[
"[Tesla.Multipart]\n",
"boundary: ",
mp.boundary,
?\n,
"content_type_params: ",
inspect(mp.content_type_params),
?\n
| Enum.map(mp.parts, &[inspect(&1), ?\n])
]
end
defp debug_body(data) when is_binary(data) or is_list(data), do: data
defp debug_body(term), do: inspect(term)
end