w5851/julia-pro

Julia Pro

When to apply

• Implementing or refactoring Julia scientific backend code.
• Diagnosing performance issues (allocations, type instability, slow hot paths).
• Designing extensible APIs with multiple dispatch.
• Improving package quality (tests, docs, reproducibility).

Core workflow

1. Clarify numerical goal, invariants, and expected precision/tolerance.
2. Make data contracts explicit at boundaries (input/output types and units).
3. Keep kernels type-stable and allocation-light.
4. Validate correctness before optimization.
5. Profile hot paths, then optimize only measured bottlenecks.

Implementation rules

• Prefer concrete container element types in hot loops.
• Avoid global mutable state in compute kernels.
• Isolate IO/config parsing from numerical kernels.
• Use function barriers when dealing with abstract inputs.
• Avoid type piracy; extend methods in your own namespace.
• Keep APIs composable and predictable.

Performance checklist

• Run allocation checks for representative workloads.
• Confirm return type stability for key kernels.
• Benchmark before/after each optimization change.
• Guard against regressions with deterministic smoke cases.

Testing expectations

• Add/maintain unit tests for numerical correctness.
• Include edge-case tests for domain boundaries.
• Use tolerance-aware assertions for floating-point outputs.
• Keep fixtures minimal and reproducible.

Response format

• Summary of approach
• Proposed/implemented code changes
• Verification method and benchmark notes
• Remaining risks and next actions