Roadmap¶
lazyarray is a small, stable library that has been in production use as a component of PyNN for over a decade. It is essentially in maintenance mode: the public API is unlikely to change substantially, and most development effort is spent keeping the package compatible with new releases of its dependencies.
The current priorities are:
Track new versions of Python, NumPy and SciPy. Each release adds testing for the latest stable Python, and drops Python versions that have reached end-of-life. The CI matrix exercises three pinned NumPy/SciPy sets (oldest supported, middle, newest).
Improve interoperability with NumPy. A longer-term goal is to integrate more cleanly with the modern NumPy dispatch protocols (
__array_function__and__array_ufunc__) so that more third-party functions acceptlarrayinstances transparently. Currently, only a subset of NumPy ufuncs is supported.Add type hints. The public API is small enough that fully annotating it is feasible. Running
mypyin CI would catch regressions in signatures and make integration with downstream type-checked code easier.Maintenance-only stance for the public API. New features will be considered when they are well-motivated by a downstream user (most often PyNN), but the project is not actively seeking feature growth.
Suggestions and pull requests in any of these areas are welcome — see Contributing to lazyarray.