Escaping the Pointless: Embedding Pointful Array Programming in Python
summary
Ein is a new pointful array DSL embedded in Python that names indices and compiles into NumPy/JAX/TensorFlow calls; it implements an axial applicative compilation that converts comprehensions into labeled-axis arrays and elementwise broadcasts. The implementation enforces rectangular shapes via separate index typing, supports user-defined element classes and static typing, and hits comparable performance to existing systems within a small constant factor on benchmarks. Code and experiments are open source; the design bridges Dex-style pointful expressiveness with existing Python array runtimes.
AI-written summary · about
chapters (10)
- 0:00Embedding pointful array programming
- 1:46Limitations of Python arrays
- 3:27Pointful vs point-free explained
- 5:13Pointful array languages
- 6:37Practical pointful code example
- 8:23Features of embedded DSL
- 10:51Compiling to point-free
- 17:27Axial applicative scheme
- 19:12NumPy comparison and performance
- 20:58Conclusion and future work
related talks
38mA Ray of Hope: Array Programming for the 21st Century
02h 50mFunctional Array Funhouse Intensive - Part 3 of 5
58mFunctional Array Funhouse Intensive - Part 2 of 5
27mAPL-Inspired Techniques for Advanced NumPy
02h 24mMagical NumPy with JAX
01h 10mJAX: accelerated machine learning research via composable function transformations in Python
15mCompilation of Sparse Array Programming Models
43muarray - Efficient and Generic Array Computation