On 08.02.2012 13:36, Nick Coghlan wrote:
That isn't really the case with NumPy - it has a sizable developer base of its own, along with solid backing from Enthought.
I thing you got this the wrong way. Inclusion in the stdlib requires long-term support, it is not a way to ensure long-term support for projects that don't have it. (And NumPy is likely to be supported for a very long time.)
Incorporation into the standard library would be a *lot* of pain for minimal gain. If it helps, just consider SciPy Python's "stdlib++" if you're doing any kind of heavy number crunching with Python. There's a reason the PyPy folks were able to raise money to sponsor their NumPyPy compatibility effort - it's because the SciPy ecosystem is centred around NumPy, and NumPyPy promises to let developers enjoy the benefit's of PyPy without losing access to SciPy.
NumPy is not just for number-crunching. It is also a general memory abstraction, a mutable container for any kind of binary data, for any kind of bit and byte fiddling, reading and parsing binary data, memory mapping binary files, etc. Another use-cases are computer graphics and image processing. Sturla