Hi! I just came across your article. Cool stuff! Thanks for sharing it!

I noticed there, however, a misuse of the new function.

Essentially, di-polysemy should not be using the new function, nor trying to construct a Di value at all. Instead, the various runXxx functions in your library should take a Di value as a parameter.

You can think of the Di value as file handle or MVar that will be accessed from multiple places in the application, possibly from different threads, with and without polysemy. If you construct a new Di value each time diToIO runs, then there's no guarantee that the log messages that are finally rendered to the outside world will happen in an absolute order, and if you attempt to log things from multiple threads you'll most definitely see log messages stepping on each other on stderr or similar.

new is expected to be used just once, in main, as soon as the application starts, and the resulting Di value should be given as argument to the various runXxx functions, and to the logging functions such as info, notice, etc. It's perfectly safe to use this one Di value, as is, from multiple threads.

As a convenience, to keep things simple in end-user applications (that is, in applications that depend on di-polysemy), you'll probably want to call Di.new from the di package, which is a bit higher-level than Di.Core.new.