strange fast_float (?) bug: > > On 2012-08-15, P Purkayastha <ppu...@gmail.com> wrote: >> On 08/15/2012 10:58 PM, Dima Pasechnik wrote: >>> On 2012-08-15, P Purkayastha<ppu...@gmail.com> wrote: >>>> On 08/15/2012 01:47 PM, Dima Pasechnik wrote: >>>>> The following produces a nice plot, all the way down to the horisontal >>>>> axis: >>>>> rr=2.5000000001 >>>>> y=9 -(9/rr)*x^2 >>>>> plot(sqrt(y) ,(-sqrt(rr),sqrt(rr))).show() >>>>> >>>>> but changing rr to rr=2.5 >>>>> makes the plot to stop approximately 0.25 units above >>>>> the horisontal axis. >>>>> >>>>> Is this a bug, or a feature? >>>>> Thanks, >>>>> Dima >>>>> >>>>> >>>> >>>> The error comes from quite an unusual stage. Doing set_verbose(1) shows >>>> us the real reason why it doesn't plot the end points. >>>> >>>> sage: set_verbose(1) >>>> sage: rrr=2.5; yy=9 -(9/rrr)*x**2; plot(sqrt(yy) ,(-sqrt(rrr),sqrt(rrr))) >>>> verbose 1 (2392: plot.py, generate_plot_points) >>>> Unable to compute f(-1.58113883008) (time = -0.044146) >>>> verbose 1 (2392: plot.py, generate_plot_points) >>>> Unable to compute f(1.58113883008) (time = -0.044146) >>>> >>>> This seems to stem from here: >>>> data[i] = (float(xi), float(f(xi))) >>>> if str(data[i][1]) in ['nan', 'NaN', 'inf', '-inf']: >>>> >>>> Note that f is a "fast_float" version of yy. This version of yy computes >>>> the value as nan. Computing yy(sqrt(rrr)) itself works just fine. >>>> >>> so this is a bug, I suppose... But where? >> >> Probably somewhere in fast_float or FastDoubleFunc. I can not say >> exactly where because I am unfamiliar with that code. >>
-- -- 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
