Incomparable abominations (was: python-dev Summary)

[Andrew Koenig]

David> I completely DO NOT BUY Guido's point in the thread.  The
David> failure of The argument is sometimes made that it is a
David> programming mistake to try to compare items of incomensurable
David> types.

Actually, the problem is somewhat worse than that, because at present,
comparison between long integers and floating-point numbers is done by
converting the integer to floating-point and comparing the converted
result.  One problem with this approach is that the conversion might
fail.  So, for example, [0.0,1L<<10000].sort() typically causes a
run-time error.

More subtle problems can occur because of loss of information in
conversions.  For example:

>>> x = 10000000000000000000000000000000000
>>> y = x+1
>>> z = float(x)
>>> x == z
1
>>> y == z
1
>>> x == y
0

See SF bug report #513866 for more detail.

-- 
Andrew Koenig, ark at research.att.com, http://www.research.att.com/info/ark