On 9/29/13 1:30 PM, Terry Reedy wrote:
On 9/29/2013 6:53 AM, Ned Batchelder wrote:
This is the nature of Unicode pain in Python 2 (Python 3 has a different
kind!). This may help you understand what's going on:
http://nedbatchelder.com/text/unipain.html
This is really excellent and I bookmarked it.
There is one minor error: "the conversion from int to float can't fail,"
>>> float(10**1000)
Traceback (most recent call last):
File "<pyshell#0>", line 1, in <module>
float(10**1000)
OverflowError: long int too large to convert to float
Even when it succeeds, it can fail in the sense of losing information.
>>> int(float(12345678901234567890))
12345678901234567168
>>> float(int(1.55))
1.0
This is somewhat analogous to a combination of errors='ignore' and
errors='replace' (with random garbage).
I think the presentation would be strengthened with the correction, as
it shows that the problems of conversion are *not* unique to bytes and
unicode.
Thanks, these are excellent points.
--Ned.
--
https://mail.python.org/mailman/listinfo/python-list
