mhampton wrote:
There has been some previous discussion about this on sage-devel, I
can't find exactly the thread I remember but here's a somewhat related
one:
http://groups.google.com/group/sage-devel/browse_thread/thread/b91c51672ae0f475/
Thank you.
Personally I think it makes sense to put the most effort into getting
sage to 100% coverage.
I agree with you there.
The randomised testing happens to be one area where I could perhaps personally
contribute something useful, whereas I don't feel I can with the doc tests - at
least not many of them.
I did wonder if the get_memory_usage() should perhaps take the output and
compare it to what is reasonable. i.e. if Sage starts and uses more than 300 MB
RAM, then something is probably wrong. Conversely if it uses less than 30 MB
RAM, something is probably wrong. (Those numbers might change over time of course).
Whenever possible when writing doctests the
results should be checked. Perhaps in some test blocks it could be
remarked that the result has been verified by another system.
Yes agree. I mentioned this the other day when a doc test failed on Solaris, but
the reason the expected value was given in the doc test was not justified. It
seemed to be based on what someone got as a result. When I computed the result
in Mathematica using high precision, the result I actually got was closer to the
true value that the "expected" value in the doc test.
I tried to address that somewhat patch to correct a numerical noise issue, where
I stated
* The high precision value of 'e'
* The nearest rounded number
* The nearest number corresponding to the IEEE standard.
* The number I actually got on my SPARC.
http://trac.sagemath.org/sage_trac/attachment/ticket/8374/8374-numerical-noise.2.patch
(Both http://trac.sagemath.org/sage_trac/ticket/8374 and
http://trac.sagemath.org/sage_trac/ticket/8375, which are very similar patches,
are still awaiting review.
There should be some justification deciding what is the expected result, not
simply putting a number in the doc test, since that is what someone's computer
happens to give.
-Marshall
Dave
--
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
