On Saturday, May 26, 2012 9:45:39 PM UTC+2, Snark wrote:
>
> Le dimanche 20 mai, Julien Puydt a écrit:
> > sage -t --long -force_lib
> > devel/sage/sage/interfaces/maxima_abstract.py
> > **********************************************************************
> > File
> >
> "/home/jpuydt/sage-5.0/devel/sage-main/sage/interfaces/maxima_abstract.py",
> > line 1595: sage: float(maxima("1.7e+17")) Expected: 1.7e+17
> > Got:
> > 1.6999999999999997e+17
> > **********************************************************************
> >
> > This is an ECL bug, reported upstream :
> >
> http://sourceforge.net/tracker/?func=detail&atid=398053&aid=3495303&group_id=30035
>
>
> Sigh... upstream tested and found it wasn't their fault, so I guess
> there's something else somewhere :-/
>
Actually that number above is different from the one in the bug report,
which had been reduced to a call to FORMAT: (format t "~21e" 1.7e+17)
Perhaps Maxima is using some _other_ instruction for printing floating
point numbers? It also matters whether the floating point number is double
float or not.
It also matters what maxima() returns (a string?) and how python converts
it to floating point number: ECL uses a method to match printed
representation and floating point numbers which is NOT the C library, for
this has proven to be buggy (to say the least). Instead ECL uses a routine
found in SBCL which is from a paper (by Steele
http://grouper.ieee.org/groups/754/email/pdfq3pavhBfih.pdf) but this
printer is not always active: if you fix the number of decimal places, the
printer need not produce the most accurate representation.
For instance, if you type 1.7d+17 in the toplevel of a Common Lisp
implementation, the lisp will output a number 1.7d+17 However, if you use
(format t "~21v" 1.7d+17) you do not get this.
Ok, to make a long story short, the situation might be:
* Python relies on the C library to reconstruct a floating point number
representation which is accurate, but does not recover the right number (C
library does the wrong rounding).
* Maxima is using some other function for output of floating point numbers
and the bug report was thus inaccurate.
Juanjo
--
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
