Issue 928881: Default encoding harmful (original) (raw)

Issue928881

Created on 2004-04-03 16:00 by tgoeller, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Messages (3)
msg20422 - (view)	Author: Toni Goeller (tgoeller)	Date: 2004-04-03 16:00
Although I am a newbie to Python and seem to be the only one in the world who is unhappy about it (at least I didn't find open bugs against it): Setting the default encoding to 'ascii' and the error handling to 'strict' is the worst choice to make. The following rules to ease the creation of stable programs are violated: Rule 1: str() should *NEVER* raise exceptions. It is meant to give you a readable representation of your data. Period. It is ok to return "You fool! Why do you have non-ascii characters in your Unicode string when ascii is your default encoding?", but it is *NOT* ok to return an exception! Rule 2: Avoid data driven exceptions, if they can be avoided. In most cases, it is absolutely not what the programmer wants to get an exception for half of the possible byte values (and this is the less tested half in English-speaking countries). The default should enable the less experienced programmer to write acceptably stable programs, non-default settings should enable the expert to fine-tune program behaviour. This gets worse for Python's choice to make it extremely hard to set the default encoding after the sitecustomize.py file was run (if it is run at all). No other programming language I know is as hard to use in this respect as Python. I know this is a defined feature, not a bug, but it is a bad and dangerous design decision nonetheless.
msg20423 - (view)	Author: Hyeshik Chang (hyeshik.chang) *	Date: 2004-04-03 16:53
Logged In: YES user_id=55188 > str() should *NEVER* raise exceptions. It is meant to > give you a readable representation of your data. If you want just string representations, use repr(). str() have never guaranteed not raising exceptions. > It is ok to return "You fool! Why do you have non-ascii > characters in your Unicode string when ascii is your > default encoding?", but it is *NOT* ok to return an > exception! What remains when except exceptions? > This gets worse for Python's choice to make it > extremely hard to set the default encoding after the > sitecustomize.py file was run (if it is run at all). I agree that sitecustomize.py and "if 0:" blocks that controls encodings stuff are somewhat newbie-unfriendly. Can't you say any good idea to improve this situation?
msg20424 - (view)	Author: Marc-Andre Lemburg (lemburg) *	Date: 2004-04-03 17:13
Logged In: YES user_id=38388 There is a simple solutions to both of your problems: always use Unicode for text data and use proper encodings when doing text data I/O. The intention of the strict Python defaults is to teach programmers to right correct programs rather than get away with bad design. Closing this as won't fix.