First-Order Laziness

on YouTube ↗27m

" summary

Proposes replacing thunk closures with lazy constructors that are defunctionalized first‑order values, implemented in Koka; evaluation happens on pattern match via a new lazy match and results use in‑place memorization plus optional in‑place reuse when reference count is one. Benchmarks (heap/queue examples) show traditional laziness can cost up to 5× in Koka, while lazy constructors cut allocations and make append‑heavy streams much faster; debugging and printing improve, but lazy match can expose memoization and break referential transparency.

AI-written summary · about

" chapters (17)

1. 0:00First-order laziness intro
2. 2:49Higher-order laziness problems
3. 4:15Defunctionalizing laziness
4. 7:25Lazy constructors in practice
5. 9:29Benefits of lazy constructors
6. 11:56Compiling in Koka
7. 13:18In-place memorization
8. 15:25Reuse vs memorization
9. 17:13Benchmarking performance
10. 19:01Key ideas and conclusion
11. 19:32Q&A: Whole-program vs user control
12. 20:37Q&A: Modularity costs
13. 21:40Q&A: Lazy fields in OOP
14. 22:44Q&A: Code data connections
15. 23:45Q&A: Defining lazy functions
16. 24:53Q&A: Impact on semantics
17. 26:21Q&A: Non-termination in constructors