Brian Curtin cur...@acm.org added the comment:
The previously mentioned comments about backwards incompatibility with the
number of items in the sequence are now a problem, since structseq now inherits
from tuple. It seems that n_in_sequence gets ignored and we have a 9 item tuple.
--
Marc-Andre Lemburg m...@egenix.com added the comment:
Brian Curtin wrote:
Brian Curtin cur...@acm.org added the comment:
The previously mentioned comments about backwards incompatibility with the
number of items in the sequence are now a problem, since structseq now
inherits from
Brian Curtin cur...@acm.org added the comment:
Yep, setting this back to closed.
--
status: open - closed
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue7766
___
Eric Smith e...@trueblade.com added the comment:
The more I think about this, the more concerned I am about changing the number
of elements in the tuple. That's the change that broke platform.py. Maybe we
should add a parameter named something like level, defaulting to 0.
0 = existing
Marc-Andre Lemburg m...@egenix.com added the comment:
Eric Smith wrote:
Eric Smith e...@trueblade.com added the comment:
The more I think about this, the more concerned I am about changing the
number of elements in the tuple. That's the change that broke platform.py.
Maybe we should
Eric Smith e...@trueblade.com added the comment:
Great idea, Marc-Andre. I agree that's the better approach.
It looks like PyStructSequence supports this, by setting n_in_sequence to a
value smaller then the number of PyStructSequence_Fields. A quick look doesn't
show any uses of this in the
Marc-Andre Lemburg m...@egenix.com added the comment:
Eric Smith wrote:
Eric Smith e...@trueblade.com added the comment:
Great idea, Marc-Andre. I agree that's the better approach.
It looks like PyStructSequence supports this, by setting n_in_sequence to a
value smaller then the
Eric Smith e...@trueblade.com added the comment:
Here's a patch that implement's Marc-Andre's suggestion. The docstring and
documentation need work. I still need to verify that this isn't a misuse of
PyStructSequence, but the tests pass on Windows. I need to verify a few other
platforms to
Brian Curtin cur...@acm.org added the comment:
Thanks a lot for taking a look at this, Eric and Marc-Andre. Apologies for the
few mistakes in there, I jumped the gun and submitted that last patch before
having run the full test suite again.
Good catch on the missing #ifdef. I will also run
Eric Smith e...@trueblade.com added the comment:
I can confirm the most recent patch (ex2) doesn't break anything on MacOS or
Linux.
It's clear that structseq.c is designed to be used this way, with more named
members than unnamed members (and vice-versa, actually). See n_members vs.
Eric Smith e...@trueblade.com added the comment:
Here's the final version of the patch. After some testing on various platforms
I'll commit it.
--
Added file: http://bugs.python.org/file16020/winver_as_structseq_ex4.diff
___
Python tracker
Eric Smith e...@trueblade.com added the comment:
Committed in trunk r77763, in py3k r77765.
--
assignee: - eric.smith
resolution: - accepted
stage: patch review - committed/rejected
status: open - closed
___
Python tracker rep...@bugs.python.org
Eric Smith e...@trueblade.com added the comment:
Here's an updated patch. I fixed some docstrings, modified it to work with the
most recent assertIsInstance changes, and added #ifdef for Windows.
There are a number of test failures still, I think all of them relating to
errors in platform.py
New submission from Brian Curtin cur...@acm.org:
I always find myself wishing sys.getwindowsversion() utilized the named tuple
concept, so here it is against trunk. sys.version_info was also changed in this
manner for 2.7.
Because it is a PyStructSeq/named tuple, it is still accessible like a
Eric Smith e...@trueblade.com added the comment:
I like this. I've visually reviewed the patch, but haven't tested it yet. I'm
willing to commit this.
Could you add to the tests to assert that .major is equal to [0], etc.?
Also, the documentation says that element [4] is text, but you've
Brian Curtin cur...@acm.org added the comment:
Good point about OSVERSIONINFOEX. I've actually wanted some of that info as
well, and according to MSDN the minimum supported client to get that structure
is Windows 2000 - same as OSVERSIONINFO.
Attached is a patch updated with your comments
16 matches
Mail list logo