Clojurians Log v2

I've been going through the podcast again, and as I was going through Sportify!, I found myself wondering: what, concretely, does it look like to separate out exception handling from io? Specifically, consider a situation where I want to execute a SQL query with next.jdbc/execute!, perhaps to store something in a db. I need to wrap that in a try/catch to handle failure, perhaps because a uniqueness constraint was violated. So, I could write a function that does all that, but now there's no way to test the catch logic (maybe it prepares a particular data structure that can be used to provide an appropriate HTML response) without involving the database. How do I take these apart elegantly? The only thing I can think of is abstracting the execute behind a function and passing either the next.jdbc/execute! call or some "mock" for testing that throws etc. This feels like exactly what I'd do in Java: mocking a repository or something. It also feels weird because that means jdbc stuff is represented in two places: the innermost level where the io actually happens, and the level above where the exception is handled. Maybe a way keeping the jdbc stuff contained is to keep the try/catch in the inner level and just indiscriminately transform exceptions to ex-data or something, which then gets handled higher up as a more "domain oriented" exception message? That doesn't let me test the exception logic where the execute happens, but since it's a very straightforward adapter, it doesn't matter; I can test the "real" handling by just faking the db part to return/throw the right ex-data. Thoughts? Examples of how to do this nicely?

I have lots of thoughts/suggestions but I'm at Conj so I'll follow up when I get back.