Structuring large Clojure codebases with Biff

9 months ago

Jacob O'Bryant is rewriting Yakread, a reading app, from scratch and open-sourcing it.
He is experimenting with new features for Biff, a Clojure web framework, to manage larger codebases.
Old Yakread had slow queries, like taking over 10 seconds to load subscription data.
Traditional denormalization was avoided by implementing materialized views for XTDB, stored in RocksDB.
Materialized views use pure 'denormalizer' functions to update data incrementally.
Testing is simplified with property-based testing and oracle functions to verify materialized views.
All application code in 'New Yakread' is pure, with no mocks needed in unit tests.
Functions are structured as state machines, separating pure logic from effect handlers.
Effect handlers are centralized and kept simple, improving testability and observability.
Unit tests compare function outputs against saved EDN files, making testing more convenient.
The codebase uses Pathom resolvers to separate model and view code, similar to OOP but with pure data.
Pathom resolvers handle database queries, derived data, and even UI fragments, keeping code modular.
GET request handlers define Pathom queries, simplifying backend logic and improving clarity.
The approach is similar to Fulcro but tailored for server-side rendering with htmx.
Yakread's rewrite is ongoing, with plans to integrate successful experiments into Biff.

Hasty Briefsbeta