[Scipy-tickets] [SciPy] #1014: UnivariateSpline objects should work on an empty array?
SciPy Trac
scipy-tickets@scipy....
Sat Jul 17 08:50:04 CDT 2010
#1014: UnivariateSpline objects should work on an empty array?
--------------------------------+-------------------------------------------
Reporter: lebigot | Owner: somebody
Type: defect | Status: reopened
Priority: normal | Milestone: Unscheduled
Component: scipy.interpolate | Version: 0.7.0
Resolution: | Keywords: UnivariateSpline empty array
--------------------------------+-------------------------------------------
Changes (by lebigot):
* status: closed => reopened
* resolution: invalid =>
Comment:
I guess I was not clear enough: the problem is not about
UniveriateSpline(), but about the resulting spline not being able to take
an empty array.
In fact, the created spline behaves apparently like most (if not all)
mathematical functions: it can take a single float, or an array of floats.
I would contend that NumPy users are very much used to the fact that all
these functions also yield [] when applied to an empty list that the
resulting spline should behave in the same way. Furthermore, Python's map
behaves in the same way, as does list comprehension.
So, what does matter is not the few functions that may have an exceptional
behavior with respect to empty lists, but that heavily used NumPy or
Python functions return an empty list when given an empty list of
arguments.
Simplicity and regularity of behavior is in fact much more convenient that
unexpected, exceptional behavior: it really would be great that the spline
created by UnivariateSpline behave in the usual way when given an empty
list of arguments. I strongly suggest that this "normal" feature be
implemented for the spline returned by UnivariateSpline, so that users
don't have their code raise an exception on corner cases, which is
particularly painful when it happens after a long calculation. This
change would definitely make SciPy more convenient and simple to use.
--
Ticket URL: <http://projects.scipy.org/scipy/ticket/1014#comment:3>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.
More information about the Scipy-tickets
mailing list