Unicode thing that causes a traceback in 2.6 and 2.7 but not in 2.5,and only when writing to a pipe,

[Dan Stromberg]

What is this about? It's another n~ thing, but this time in 2.x.
All I'm doing is printing a str containing a character > 127.

It works fine in 2.5, to a terminal or to a pipe. In 2.6 and 2.7, it
fails when writing to a pipe but works fine writing to my terminal -
an mrxvt.\

I really kind of want octets to just be octets, since I'm on a Linux
system - it probably should be up to the user and their related
software to decide how those octets are interpreted. I'm assuming
that the bytes() workarounds I'm using in 3.x aren't going to work in
2.x - it looks like bytes() is just an alias for str() in 2.6 and 2.7.

BTW, the print(line) in the traceback looks like Python 3 syntax, and
it is I guess, but it's also Python 2 compatible. In 3.x, it's a
function with a single argument, in 2.x it's a statement with a
parenthesized expression.

Thanks!

benchbox-dstromberg:~/src/home-svn/readline0/trunk i686-pc-linux-gnu
17932 - above cmd done 2010 Wed Dec 01 09:40 PM

$ find /usr/share/jpilot -print0 | /usr/local/cpython-2.5/bin/python
../test-readline0 2>&1 | cat
/usr/share/jpilot
/usr/share/jpilot/CalendarDB-PDat.pdb
/usr/share/jpilot/MañanaDB.pdb
/usr/share/jpilot/jpilotrc.green
/usr/share/jpilot/jpilotrc.purple
/usr/share/jpilot/ContactsDB-PAdd.pdb
/usr/share/jpilot/MemoDB.pdb
/usr/share/jpilot/jpilotrc.blue
/usr/share/jpilot/ExpenseDB.pdb
/usr/share/jpilot/jpilotrc.default
/usr/share/jpilot/AddressDB.pdb
/usr/share/jpilot/ToDoDB.pdb
/usr/share/jpilot/TasksDB-PTod.pdb
/usr/share/jpilot/DatebookDB.pdb
/usr/share/jpilot/Memo32DB.pdb
/usr/share/jpilot/MemosDB-PMem.pdb
/usr/share/jpilot/jpilotrc.steel
benchbox-dstromberg:~/src/home-svn/readline0/trunk i686-pc-linux-gnu
17932 - above cmd done 2010 Wed Dec 01 09:40 PM

$ find /usr/share/jpilot -print0 | /usr/local/cpython-2.6/bin/python
../test-readline0 2>&1 | cat
/usr/share/jpilot
/usr/share/jpilot/CalendarDB-PDat.pdb
Traceback (most recent call last):
File "./test-readline0", line 22, in <module>
print(line)
UnicodeEncodeError: 'ascii' codec can't encode character u'\xf1' in
position 20: ordinal not in range(128)
benchbox-dstromberg:~/src/home-svn/readline0/trunk i686-pc-linux-gnu
17932 - above cmd done 2010 Wed Dec 01 09:40 PM

$ find /usr/share/jpilot -print0 | /usr/local/cpython-2.6/bin/python
../test-readline0 2>&1
/usr/share/jpilot
/usr/share/jpilot/CalendarDB-PDat.pdb
/usr/share/jpilot/MañanaDB.pdb
/usr/share/jpilot/jpilotrc.green
/usr/share/jpilot/jpilotrc.purple
/usr/share/jpilot/ContactsDB-PAdd.pdb
/usr/share/jpilot/MemoDB.pdb
/usr/share/jpilot/jpilotrc.blue
/usr/share/jpilot/ExpenseDB.pdb
/usr/share/jpilot/jpilotrc.default
/usr/share/jpilot/AddressDB.pdb
/usr/share/jpilot/ToDoDB.pdb
/usr/share/jpilot/TasksDB-PTod.pdb
/usr/share/jpilot/DatebookDB.pdb
/usr/share/jpilot/Memo32DB.pdb
/usr/share/jpilot/MemosDB-PMem.pdb
/usr/share/jpilot/jpilotrc.steel
benchbox-dstromberg:~/src/home-svn/readline0/trunk i686-pc-linux-gnu
17932 - above cmd done 2010 Wed Dec 01 09:41 PM

$

 

[Piet van Oostrum]

What is this about? It's another n~ thing, but this time in 2.x.
All I'm doing is printing a str containing a character > 127.

It works fine in 2.5, to a terminal or to a pipe. In 2.6 and 2.7, it
fails when writing to a pipe but works fine writing to my terminal -
an mrxvt.\

I really kind of want octets to just be octets, since I'm on a Linux
system - it probably should be up to the user and their related
software to decide how those octets are interpreted. I'm assuming
that the bytes() workarounds I'm using in 3.x aren't going to work in
2.x - it looks like bytes() is just an alias for str() in 2.6 and 2.7.

Strings are indeed byte strings (or octet strings if you prefer) in Python 2.x.

Traceback (most recent call last):
File "./test-readline0", line 22, in <module>
print(line)
UnicodeEncodeError: 'ascii' codec can't encode character u'\xf1' in
position 20: ordinal not in range(128)

This message shows that line is not a byte string but a unicode string.
If you want to output a unicode string it has to be converted to bytes
first, and Python must know which encoding to use. For output to the
terminal it can usually get that information from the system, but not
for a pipe. You will have to supply it.

By the way, if you just want to work with octets, why is the variable
line in unicode? If you keep everything in byte strings your problem may
disappear. In Python 3 you have to do this explicitely as the default is unicode.