| msg209625 - (view) |
Author: Serhiy Storchaka (serhiy.storchaka) *  |
Date: 2014-01-29 09:16 |
| >>> import io, _pyio >>> io.StringIO('a\nb\r\nc\rd', newline=None) <_io.StringIO object at 0xb707c734> >>> io.StringIO('a\nb\r\nc\rd', newline=None).getvalue() 'a\nb\nc\nd' >>> _pyio.StringIO('a\nb\r\nc\rd', newline=None).getvalue() 'a\nb\r\nc\rd' >>> s = io.StringIO(newline=None); s.write('a\nb\r\nc\rd'); s.getvalue() 8 'a\nb\nc\nd' >>> s = _pyio.StringIO(newline=None); s.write('a\nb\r\nc\rd'); s.getvalue() 8 'a\nb\r\nc\rd' |
|
|
| msg210019 - (view) |
Author: Antoine Pitrou (pitrou) * |
Date: 2014-02-02 21:39 |
| It's a bug in _pyio.StringIO.getvalue(). Attaching patch. |
|
|
| msg210028 - (view) |
Author: Serhiy Storchaka (serhiy.storchaka) * |
Date: 2014-02-02 22:02 |
| I rather think that it's a bug in _io.StringIO.getvalue(). |
|
|
| msg210031 - (view) |
Author: Antoine Pitrou (pitrou) * |
Date: 2014-02-02 22:06 |
| With the patch, getvalue() is consistent with read() and other methods. Without the patch, _pyio.StringIO.getvalue() returns a different value from _pyio.StringIO.read() *and* from _io.StringIO.getvalue(). Changing _pyio.StringIO.getvalue() is the path of least resistance here. |
|
|
| msg210033 - (view) |
Author: Antoine Pitrou (pitrou) * |
Date: 2014-02-02 22:09 |
| In other words, the bug is that _pyio.StringIO.getvalue() doesn't do any newline conversion; the patch fixes that. |
|
|
| msg210041 - (view) |
Author: Serhiy Storchaka (serhiy.storchaka) * |
Date: 2014-02-02 22:31 |
| But how then other values of newline except '\n' can be useful? The problem is that newline converting is applied twice, in write() and in read(). When constructor uses write() and getvalue() returns same value as read(), we have no chance to get newlines encoded or decoded only once. Current results for newline != '\n' looks meanless to me. Here is my half-baked patch. It fixes only _pyio.StringIO, but _io.StringIO are still partially broken. The patch also contains new tests. |
|
|
| msg210043 - (view) |
Author: Antoine Pitrou (pitrou) * |
Date: 2014-02-02 22:36 |
| > Current results for newline != '\n' looks meanless to me. They don't look meaningless to me, e.g.: >>> io.StringIO("a\r\nc\rd", newline=None).getvalue() 'a\nc\nd' >>> sio = io.StringIO(newline=None); sio.write("a\r\nc\rd"); sio.getvalue() 6 'a\nc\nd' There may be other cases where they make less sense, but that's a separate issue. |
|
|
| msg210044 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2014-02-02 22:40 |
| New changeset 99168e7d4a3d by Antoine Pitrou in branch '3.3': Issue #20435: Fix _pyio.StringIO.getvalue() to take into account newline translation settings. http://hg.python.org/cpython/rev/99168e7d4a3d New changeset aadcc71a4967 by Antoine Pitrou in branch 'default': Issue #20435: Fix _pyio.StringIO.getvalue() to take into account newline translation settings. http://hg.python.org/cpython/rev/aadcc71a4967 |
|
|
| msg210045 - (view) |
Author: Roundup Robot (python-dev) |
Date: 2014-02-02 22:42 |
| New changeset 3e61d8e06ef7 by Antoine Pitrou in branch '2.7': Issue #20435: Fix _pyio.StringIO.getvalue() to take into account newline translation settings. http://hg.python.org/cpython/rev/3e61d8e06ef7 |
|
|
| msg210047 - (view) |
Author: Antoine Pitrou (pitrou) * |
Date: 2014-02-02 22:42 |
| I've fixed the reported bug. If other problems need fixing, better open a new issue :-) |
|
|
| msg210079 - (view) |
Author: Serhiy Storchaka (serhiy.storchaka) * |
Date: 2014-02-03 06:39 |
| Thank you for fixing this bug. Maybe I reported it too hastily. |
|
|