Static Analysis for Ruby with Jake Zimmerman

What It Covers

Jake Zimmerman explains Sorbet, Stripe's static type checker for Ruby written in C++, covering its architecture using control flow graphs, performance optimizations to handle 15-20GB codebases, and upcoming features for shapes, tuples, and ecosystem support.

Key Questions Answered

• •Control Flow Graph Architecture: Sorbet compiles Ruby's abstract syntax tree into a unified control flow graph representation, allowing control flow sensitive typing to be implemented once rather than separately for if, unless, while, rescue, and break nodes, simplifying the codebase significantly.
• •Memory Optimization Strategy: Stripe's codebase currently requires 15-20GB in memory for type checking. The team is rearchitecting Sorbet to type check code subsets sequentially rather than loading 100% of the codebase simultaneously, preventing memory issues as the codebase doubles in size.
• •C++ Language Choice: The team selected C++ over Ruby, Rust, or Go in 2017 because three founding members had extensive C++ experience, and fine-grained allocation control was critical for performance. The mypy team explicitly advised against implementing in Python due to performance tuning difficulties.
• •Interface Usage Pattern: Developers underutilize abstract methods and interfaces, assuming they only matter with multiple implementations. Interfaces provide valuable data hiding and implementation encapsulation even with single implementations, exposing only necessary public methods while hiding internal fields and helper functions.