The 'push ifs up and fors down' principle is a programming heuristic that suggests moving conditional logic (if statements) upward in a function while pushing iterative loops (for operations) downward. This approach aims to enhance code clarity and performance by centralizing branching logic and utilizing bulk operations.
The principle is discussed in the Tiger Style document of TigerBeetle and has been elaborated upon by programmer matklad in a blog post. Matklad emphasizes that by pushing conditionals up, a function can become more efficient. For instance, instead of handling an Option<Walrus> internally, the caller should manage the None case, allowing the function to accept a plain Walrus. This adjustment narrows the function's input state space and acts as a filtering mechanism.
Conversely, pushing loops down involves deferring loops until after data filtering or reduction, which minimizes unnecessary computations. Instead of calling frobnicate(walrus) in a loop, a batch function like frobnicate_batch(walruses) should be used, allowing the inner loop to run without branching, making it a candidate for vectorization.
The principle also applies to database query optimization. In SQL, it is recommended to perform projections and selections early in the query execution to reduce the dataset's width and filter out irrelevant data. This is akin to pushing conditionals up in programming.
In functional programming and category theory, the principle can be illustrated through the concept of predicates and subsets. A predicate can define a subset of elements that satisfy certain conditions, which can be represented in code by using types that eliminate the need for conditional checks.
Overall, the 'push ifs up and fors down' principle serves as a guideline for optimizing code structure and performance, advocating for early filtering and efficient batch processing.