Proposal: Using references to log messages in logging hooks

I was making some minor corrections in the logging docs when I happened upon hooks in our logging engine (what a nice surprise!). I noticed they were not documented, and as I tried to use them to make some examples, I got a bit frustrated by them. I'd like to discuss a possible change.

In Dancer2::Core::Role::Logger, we have:

around 'log' => sub {
    my ($orig, $self, @args) = @_;

    $self->execute_hook( 'engine.logger.before', $self, @args );
    $self->$orig( @args );
    $self->execute_hook( 'engine.logger.after', $self, @args );
};

In other places (such as Dancer2::Core::Role::Template), we pass around a reference so we can manipulate content in the hooks before continuing execution. This isn't possible in the logger hook, at least not trivially (unless I am missing something - which is completely likely!). So you'd have to push that functionality down into the logging adapter/logger engine, or create another wrapper method for log().

What are your thoughts about making @args a reference instead? This would allow developers to add additional information to log messages (such as a request ID or other metadata) without repeating code.

On the other hand, you can make a solid argument for not wanting to alter log messages. And maybe that was the original intent. I wasn't here for that. If that is the case, I can close this and you all can disregard.

Dancers, what are your thoughts here? Am I simply missing the point of what you're supposed to do with this hook?

Start by reading Dancer2::Core::Role::Logger, especially the around 'log' hook, and compare it with Dancer2::Core::Role::Template. Review how execute_hook receives @args and how logging adapters consume them. The issue does not define a settled behavior, so done would require agreement on whether hooks may modify log messages and a resulting implementation or documented decision.