Allow logging Operation#wtf?

[phillipoertel]

Hey there,

our developers love #wtf? and leave it in even after feature development is complete, to reproduce and debug on staging/prod what was going on, in case something fails (it happens ;-D).

Problem is, the #wtf output messes up the output of tests, particularly rspec when run in --format documentation mode, and the #wtf output can't be turned off easily at the moment.

So my request is:

• allow configuring where the output of #wtf? goes -- to STDOUT by default (backwards compatibility), to a logger (i.e. allow assigning a logger that the wtf output gets sent to), or somewhere else (silencing it by setting the config to nil/false)
• I think the log level should be debug or at most info
• (optional) if running in Rails context, use it's logger by default

One thing I'm not so sure about is how to preserve the colorized, "ASCII-art" output of wtf in a log message. As is, will be quite hard to read and piece together if sent to a logger. In the simplest case, just make sure all of output ends up in a single log message, including the ASCII colorization codes and newlines. Then it will be fairly easy to copy out and restore. Serializing such a tree structure to a (JSON) string may be another option; but that's possible overkill and will likely be harder to read anyways. Maybe you additional ideas. What wouldn't work well is to send each line of output as a separate message, it will be hard to piece together again.

I've attached a screenshot of a typical #wtf? output for reference:

And here's the monkeypatch we currently use to silence #wtf?, as it may be helpful to others. It works in the context of a Rails application using the dotenv gem for configuration.

In RAILS_ROOT/.env:

[...]
# turn on to suppress trailblazer's wtf? output (in specs, for example)
SILENCE_TRAILBLAZER_WTF=true

In RAILS_ROOT/config/initializers/trailblazer.rb

# monkeypatch to silence the trailblazer wtf? output
module Trailblazer::Developer
  module Wtf
    module_function

    def puts(string)
      (ENV["SILENCE_TRAILBLAZER_WTF"] == "true") ? nil : super
    end
  end
end

[Drowze]

+1 to this!

While I love #wtf?, I'd prefer if it was easier to programatically print its output based on some custom logic (say - only print it when a test fails)

I'll probably resort to some monkey patch for now, but a more flexible API would definitely be preferable 😄

[apotonick]

@Drowze Thanks for your input! 🍻 We actually want to solve this problem cleanly by changing the way operations are (or should be) invoked. With endpoints, the operation is basically not called directly anymore:

# just an example, where TRB is an imaginary endpoint constant
TRB._(Memo::Operation::Create, params: {here  we go})

In most cases, this could be hidden, either behind Controller#run or, in a test, behind an #assert_run abstraction or your own helpers. While this might feel clumsy, it's actually much much cleaner than monkeypatching all the things.

Any ideas or objections about this?