policyengine-vectorization

Vectorization replaces scalar if-elif-else logic with NumPy array operations like where() and select() to process thousands of households simultaneously. This eliminates Python branching and applies conditions directly across array data, dramatically improving performance in microsimulations.

Use where() and select() when processing array data from multiple entities. These NumPy functions apply conditions across entire arrays without scalar branching, preventing performance bottlenecks that occur when looping through individual records in PolicyEngine microsimulations.

Yes. Boolean operations (&, |, ~) and clip() work on entire arrays simultaneously, replacing scalar conditionals. This vectorized approach maintains performance when applying bounds and lookups across thousands of household records.

Vectorized operations process entire arrays in a single operation, while scalar if-else requires iterating through each record individually. For PolicyEngine microsimulations handling thousands of households, vectorization eliminates loop overhead and leverages NumPy's optimized C-level implementations.

Replace each if-elif block with where() for binary conditions or select() for multiple branches, operating on boolean masks across your entire array. Then apply clip() for value bounds and vectorized lookups for table references instead of scalar dictionary access.