invoke and execute in Rake

When writing tests for a Rake task at work, I came across invoke and execute as two different ways for calling a specific task. Most of the content I found online about the difference was fairly superficial: invoke can only get called once, execute can be called as many times as one wants.

There, are you happy? Move on.

Never being generally happy, I did not move on. I wanted to know how these similar-looking methods were executed differently. So, I consulted The Truth.

No, not Stack Overflow.

Consulting the source

When a Rake task object is created, lots of instance variables are also created. One of those is called @already_invoked and is initialized as false:

def initialize(task_name, app)
  ...
  @already_invoked = false
  ...
end

And in the function that actually does the invoke-ing, there’s this check and set:

return if @already_invoked
@already_invoked = true

Execution of the invocation stops here if true, and you’re left either grateful [that something didn’t run more than you wanted it to] or frustrated [that it did].

Also!

This seems to be overlooked in lots of the quick posts about the differences between invoke and execute: execute passes in the args as a hash, as expected. However, invoke takes the arguments and converts them from a hash to a TaskArgument.

def invoke(*args)
  task_args = TaskArguments.new(arg_names, args)

This fucked me up when I was trying to use args.with_defaults in my tests, but Jess figured out where my assumption was wrong. 🙌

Related: If you need to invoke a task multiple times, one could reenable a task manually by calling the appropriately-named reenable method, which resets the @already_invoked instance variable:

def reenable
  @already_invoked = false
end