http://docs.python.org/tut/node5.html states, as an example, round(_, 2)
should round to the second decimal place.  makes sense.  when i
attempted it with different numbers, round() returned what looks like a
binary-coded decimal error (see the jpg enclosed), and did not return
just two digits, but 15 digits.  the same error was produced for
round(_, 3).  when the number was changed to 1 digit, the return result
rounded from another direction.

i'd guess this isn't a feature :)

Not really a feature, but not a bug. See
http://www.python.org/doc/faq/general/#why-are-floating-point-calculations-so-inaccurate
for an explanation.

interesting, and expected as the BCD calculation.  however, the round function should be able to handle that correctly, dealing with the number in a string fashion as opposed a mathematical one, to display the requested decimal places.  slower, to be sure, but hella accurate.

that's what the bug was about.  i understand the issues with BCD.  i don't understand how a function to display a number can get the display so wrong, even when the BCD number itself "fails" to the 15th place.  round to 2 places, you should still display 14.78, not 14.7799999999999999x

just my take.  thanks for getting back to me.

00 caren
http://www.parkgallery.org
george davis creek, north fork

round() does not return a string, but a floating point number, the best
approximation to the "real" result that is achievable with the limited
precision of binary floating point.

That result is then formatted to a string for display by the interactive
interpreter using the repr() function, which formats a float with (I
think) 17 decimal digits.

So, the point here is that repr() produces "too many" digits for your
liking ;) You can use str() to cut off at fewer digits, which will
produce the "correct" string in most cases.

travelgirl: the 'decimal' module is designed to deal with this. Note
that there's a difference between BCD (which can represent decimal
numbers exactly) and binary (which can't, in general).