On Nov 2, 2007, at 10:48 AM, William Stein wrote:
> > On 11/2/07, Dan Christensen <[EMAIL PROTECTED]> wrote: >> numpy arrays are extremely flexible, with broadcasting, view >> semantics >> and in-place operations being the most important reason why. For >> example, if x is an array, then x[3:5] is a view of part of x, and >> I can adjust the entries in just that region with x[3:5] += [1,2]. >> This even works if x is multi-dimensional, with the [1,2] being >> broadcast to the right shape. >> >> At the same time, sage provides some fundamental types that can be >> useful as array elements. >> >> So it is completely natural to use numpy arrays of sage elements, >> and I >> have certainly done so many times. >> >> More generally, if the current implementation of in-place >> operations in >> sage causes problems in numpy, it is quite likely that there will >> also >> be problems with other python modules. So I regard this as a serious >> bug in sage. >> > > +1 After mulling this over for awhile, I feel exactly the same way > about everything said above. > > The inplace optimizations should be something that have to be > explicitly turned on, and they should be off by default. I know they > make certain things faster, but correctness by default is *always* a > much more important consideration with serious mathematical software > such as Sage. Always. > > I will very likely disable in-place optimization for sage-2.8.11, > until this issue gets properly discussed and resolved. :-(, but I have to concede to your logic. The line to change is 148 of coerce.pxi. Setting this value to 0 will turn them completely off. Other than numpy, (and the builtin libraries), do we use any other extension types? If there is a finite list of things for which it won't work, it would be possible to disable it just for those. - Robert --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---
