Le 14/03/2011 14:12, David Kirkby a écrit :
On 14 March 2011 13:04, David Kirkby<david.kir...@onetel.net> wrote:
On 13 March 2011 15:34, Julien PUYDT<julien.pu...@laposte.net> wrote:
Hi,
among the few failing tests with my ARM built, two are because of accuracy
reasons :
File "/home/jpuydt/sage-4.6.2/devel/sage/sage/functions/other.py", line 497:
sage: gamma1(float(6))
Expected:
120.0
Got:
119.99999999999997
This is a bit of a pain, as the number you get is less than expected.
I don't believe there's anything magic about using the number 6 -
virtually any positive integer would work. I'd try to find one such
the number is either exact, or *above* the nearest integer. Gamma(n)
should be factorial (n-1) - at least for positive integers. (See the
next example for how to fix it if you get one where the result is just
over an integer value).
File "/home/jpuydt/sage-4.6.2/devel/sage/sage/symbolic/expression.pyx", line
6067:
sage: SR(10.0r).gamma()
Expected:
362880.0
Got:
362880.00000000047
That's easy to change. If the test was changed to look for 362880.000000000...
then it would pass since it will ignore the "47". The "..." means to
ignore anything beyond there.
Actually I'm wrong here, as "362880.000000000..." would fail if the
result was 362880.0, as it would expect to string of zeros, which are
not present. Changing the test to "362880.0..." would work, but I'm
not sure that is such a good idea, as it would allow a very high
relative error. (362880.099999999 would pass).
Perhaps you can find value of n, such that gamma(n) gives an exact
integer result. If that happens on other CPUs too, then I suggest the
argument to the doctest is changed.
The solution of finding a "good" integer is fragile : it will break
anytime wind will change direction.
Actually, the best solution is the one I gave : handle fuzzy results!
You have to keep in mind that floats can't seriously be used for
equality -- there is a margin of error (and the margin moves according
to the floating comma which gave its name to the type).
Are doctests pure string checks? Can't they allow things like "(result -
expected) < 0.0000001" ?
Snark on #sage-devel
--
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
