Hi, On 20 Sep., 01:58, "Alex Ghitza" <[EMAIL PROTECTED]> wrote: > Hi folks, > > John Cremona and I have been working for the past two weeks on fixing the > way real precision gets passed from Sage to the Pari library and back. > Things were extremely broken, due mostly to a misunderstanding of how the > Pari library (as opposed to GP sessions) deals with precision for inexact > numbers. Precision arguments were either ignored or passed with > ridiculously huge values. Although the most important conversion-to-pari > functions in the Sage library did things properly, many others were either > plain wrong or inconsistent. > > We have been working on this at trac #4096, where there is now a source > patch and a doc patch for your (re)viewing pleasure. The patch is large, > but the bulk of it consists of fixing precision in doctests. The new > (correct) precision behavior is documented toward the top of the file > libs/pari/gen.pyx. In a nutshell, the default real precision in > interactions with the Pari library is now 53 bits, the same as the default > real precision in Sage. This is also the precision with which Pari library > objects get printed (i.e. 15 significant decimals). The patch implements > this across the Sage code, resulting in a consistent and better documented > behavior. Happy side-effects include the removal of some bugs related to > special functions, more clear and consistent code, and uniform behavior > (where possible) on both 32-bit and 64-bit machines. > > We would greatly appreciate it if someone could review this. I am happy to > answer any questions/concerns you might have about the changes. > > Best, > Alex
it's not the smallest of patches, but at a first and at a second sight it is not only very well done, e.g. with accurate doctests for the nice new functions, but also presented "on a silver plate". I still have to learn how to use Mercurial, how to apply a patch on a branch, and such; and it would take me probably more than one weekend. But if nobody else steps up, I'll volunteer to review. Of course, if somebody is already doing this, please tell me, to combine efforts (not blindly just double them), thanks! Cheers, gsw --~--~---------~--~----~------------~-------~--~----~ 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://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
