Hi Florent, On 22 Mai, 10:37, Florent Hivert <florent.hiv...@univ-rouen.fr> wrote: > I had a very similar problem with the test suite error messages which sometimg > slowed sown the suite but several order of magnitude. I solved this issue in > #8742 by creating a class for lazy format string. I'm not sure this will solve > your problem but if it does, maybe it is not a good idea to duplicate a > feature. > >... > > In my usecases, I had some error message containing some " %s "%(bla) where > the conversion from bla to its string representation take some time. So my > idea was to create a class LazyFormat inheriting from str to create an object > sur that the % operator is only evaluated when __repr__ is called. I still > create the instance of LazyFormat which take a little time. I'm not sure how > this fits your usecase.
It certainly sounds interesting! However, there are two problems: The error message is slightly different for extension classes and other classes. raise_attribute_error and AttributeErrorMessage do a case distinction that seems difficult to mimmick with lazy format strings. Moreover, according to some tests that I just posted on the ticket, they are simply too slow. I wrote a Cython function that repeated raising and catching an attribute error n times. - raise_attribute_error in unpatched sage needs about 20 seconds for n=10^7 iterations. - raising an error with an AttributeErrorMessage as argument needs about 3 seconds for n=10^7 iterations. - raising an error with a lazy format string as argument needs about 16 seconds for only n=10^6 iterations. Cheers, Simon -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
