[Python-Dev] Should bytearray(b'a')[0] should 1-char or number?

[Amaury Forgeot d'Arc]

On Fri, May 2, 2008 at 11:47 AM, ocean wrote:
> # I tried to post this to bug tracker, but couldn't, so posted here...
>
>  r62095 says
>
>  >Fix and enable a skipped test:
>  >with python 2.6, enumerating bytes yields 1-char strings, not numbers.
>  >
>  >Don't merge this into the py3k branch.
>
>  This is true for bytes, but not for bytearray.
>
>  >>> bytearray(b'a')[0]
>  97
>  [28493 refs]
>  >>> b'a'[0]
>  'a'
>
>  And this causes error on my environment like this.
>
>  ======================================================================
>  FAIL: testDecoder (__main__.StatefulIncrementalDecoderTest)
>  ----------------------------------------------------------------------
>  Traceback (most recent call last):
>   File "test_io.py", line 629, in testDecoder
>     self.assertEquals(d.decode(input, eof), output)
>  AssertionError: u'o.i.a.b.c.d.' != u'abcd.'
>
>  But strange, I cannot see this error on python.org buildbot.

The buildbots don't fail, because StatefulIncrementalDecoderTest is
not listed in the run_unittest() call :-(.

I found the cause of the problem: in io.py, TextIOWrapper.tell() calls
decode() with a bytearray, without any reason it seems.
I will take care of this tonight, if bugs.python.org is available.

-- 
Amaury Forgeot d'Arc