Honestly, SQLAlchemy was the biggest design mistake we made in sourcehut, and we're working to remove it now. SQL tends to be the bottleneck, and ultimately we just want to write SQL queries which do the right thing, so that we can control for their performance better. Knowing the SQL you want and convincing SQLAlchemy to produce it is an annoying process which makes it hard to make performant queries. I would recommend that anyone starting a new project would avoid using an ORM and just write SQL queries instead.
One thing which is occasionally useful is a query building tool. For Go we use squirrel for this purpose. If you need unrelated parts of the code to work together to produce a single SQL query, it can help to have such a tool. This is a lot less than a full blown ORM, though, it's more like passing around a SQL AST in memory.
One thing which is occasionally useful is a query building tool. For Go we use squirrel for this purpose. If you need unrelated parts of the code to work together to produce a single SQL query, it can help to have such a tool. This is a lot less than a full blown ORM, though, it's more like passing around a SQL AST in memory.