While we are here, I would like people’s thoughts on the best way to configure the backend during the runtime.

LoggerBackends offers configure/2, which can be called like this:

LoggerBackends.configure(LoggerBackends.SQL,
   option1: value1,
   option2: value2,
   ...
)

and will send a :configure event to the backend that it can capture. The reference implementation for LoggerBackend.Console is here. This call will only succeed after a backend has already been added with LoggerBackend.add(LoggerBackends.SQL) (as someone needs to listen to these events).

However, you definitely want to set the path to a SQLite log database before adding the backend to perform all the initialization in the backend’s init function.

Something like this would be nice

LoggerBackends.add(LoggerBackends.SQL,
   path: "./my.sqlite", 
   repo: nil
)

However, LoggerBackends.add/2 only processes the :flush option and ignores everything else.

So my current solution is to expose my own configure/1 to do the runtime config:

LoggerBackends.SQL.configure(
  path: "./test1.sqlite",
  repo: nil
)
LoggerBackends.add(LoggerBackends.SQL)

The previous option with LoggerBackend.add/2 is IMHO cleaner, but requires changing the existing contract.

Thoughts?