homepage

I was confused about the newline argument/attribute
in the misc. classes of the io module until I realize
there is a spelling issue between the docs and the
real implementation. (Py 2.6.5, Py2.7b2). Py 3.x not
tested.
>>> sys.version
2.7b2 (r27b2:81019, May 9 2010, 11:33:14) [MSC v.1500 32 bit (Intel)]
>>> sio = io.StringIO(u'abc')
>>> sio.encoding, type(sio.encoding)
(None, <type 'NoneType'>)
>>> sio.errors, type(sio.errors)
(None, <type 'NoneType'>)
>>> sio.newline, type(sio.newline)
Traceback (most recent call last):
 File "<psi last command>", line 1, in <module>
AttributeError: '_io.StringIO' object has no attribute 'newline'
>>> sio.newlines, type(sio.newlines)
(None, <type 'NoneType'>)
>>> 
>>> tio = io.TextIOWrapper(io.BytesIO())
>>> tio.buffer, type(tio.buffer)
(<_io.BytesIO object at 0x02E6E600>, <type '_io.BytesIO'>)
>>> tio.encoding, type(tio.encoding)
('cp1252', <type 'str'>)
>>> tio.errors, type(tio.errors)
('strict', <type 'str'>)
>>> tio.line_buffering, type(tio.line_buffering)
(False, <type 'bool'>)
>>> tio.newline, type(tio.newline)
Traceback (most recent call last):
 File "<psi last command>", line 1, in <module>
AttributeError: '_io.TextIOWrapper' object has no attribute 'newline'
>>> tio.newlines, type(tio.newlines)
(None, <type 'NoneType'>)
>>> sys.version
2.6.5 (r265:79096, Mar 19 2010, 21:48:26) [MSC v.1500 32 bit (Intel)]
>>> import io
>>> tio = io.TextIOWrapper(io.BytesIO())
>>> tio.encoding, type(tio.encoding)
('cp1252', <type 'str'>)
>>> tio.line_buffering, type(tio.line_buffering)
(False, <type 'bool'>)
>>> tio.errors, type(tio.errors)
(u'strict', <type 'unicode'>)
>>> tio.newline, type(tio.newline)
Traceback (most recent call last):
 File "<psi last command>", line 1, in <module>
AttributeError: 'TextIOWrapper' object has no attribute 'newline'
>>> tio.newlines, type(tio.newlines)
(None, <type 'NoneType'>)
>>> [e for e in dir(tio) if 'new' in e]
['__new__', 'newlines']
>>>

Thanks for the report. Could you check the docs for 3.1 and 3.x trunk (py3k) too?

This is as documented:
 http://docs.python.org/dev/3.0/library/io.html#io.TextIOBase.newlines
The keyword argument is named 'newline', the attribute is named 'newlines'. The attribute does not record what was passed to the newline argument, rather it records what newlines have been actually encountered.

2010年6月4日 R. David Murray <report@bugs.python.org>
>
> R. David Murray <rdmurray@bitdance.com> added the comment:
>
> This is as documented:
>
> http://docs.python.org/dev/3.0/library/io.html#io.TextIOBase.newlines
>
> The keyword argument is named 'newline', the attribute is named 'newlines'.
> The attribute does not record what was passed to the newline argument,
> rather it records what newlines have been actually encountered.
>
Ok, I see. I read, reread the doc prior posting, and, in my mind, it is not
obvious to understand this subtle difference from the doc.
Sorry for the noise.
Regards.

Could you think of a way to improve the docs on that point?

2010年6月4日 Éric Araujo <report@bugs.python.org>
>
> Éric Araujo <merwok@netwok.org> added the comment:
>
> Could you think of a way to improve the docs on that point?
>
> ----------
>
> Quick and dirty answer.
I have ~10 years experience with Python and it seems
to me the io module is technically excellent.
However, I have found it is not so obvious to
understand the usage of all these arguments,
errors, encoding, line_buffering, ... in the
class constructors and methods like io.open().
The doc describes what these arguments are,
their states, but not too much how to use
them.
As an exemple, I read some time ago on the c.l.p
mailing list, somebody complaining about the "encoding"
argument of the class TextIOWrapper. He defined an
"encoding='utf-8'" in the ctor and did not
understand why his "text" was not automagically
encoded. Answer: the encoding arg is only a kind
of information and does not do anything.
BTW, it was only when I read that post, I understand
a little bit more.
Regards.

This is off topic for this issue, but what you just reported makes no sense to me. TextIOWrapper calls the encoder corresponding to the specified encoding in its write method.
That said, it is true that the IO docs are currently a class reference and not a user reference, and this should be fixed. There may even already be an open issue for this.