Your 2790 patch seems to have completely wiped out the handling of marker colors in the simple case "N.plot(N.random.random(1000),'r.')", as noted by Stefan Van der Walt in a message yesterday to the matplotlib users list. I presume you are monitoring that list, and saw his message and my reply in which I noted that I was reverting the changes to axes.py. I looked at the changes to axes.py, and it appeared that you had simply forgotten the marker color handling; otherwise, the changes appeared to be code rearrangements, not API changes that would interact with other files, so I thought that reverting that file alone would work--that is, restore basic functionality to the svn version. My original thought was to wait and let you fix it, and evidently I should have, but I decided it would be better not to leave svn without the marker color functionality in the interim, and I had no way of knowing whether you would be able to get to it right away. In any case, I made the change, verified that it did restore marker color handling, committed it to svn, and left it at that.
Of course I was too hasty, and neglected to look at the patch as a whole; now I see that you were trying to shift the setting of marker colors out of axes.py and into lines.py. I still don't know why Stefan's simple example was broken, and I don't have time now to understand your patch properly, though I wish I did. I certainly don't object to the idea of trying to make the code cleaner and easier to follow.
So: please either restore your patch with additional changes to fix the problem Stefan found, or revert all of it cleanly until this can be sorted out.
If you think there might be incompatibilities introduced by your patch, then it would be good to know what they are.
Sorry to have made the situation more confusing than necessary.
Norbert Nemec wrote:
Hi Eric, hi everybody,
I just noticed that you, Eric, have committed r2799 to SVN which partly
reverts my commit r2790.
I understand that, obviously, the changes that I introduced are more
controversial than I would have expected, so we should try to reach an
Was your objection generally about my patch or just about some detail? I
know that the patch might have broken a bit of compatibility, but since
the old logic was inconsistent and hard to understand, I figured that a
cleaner and simpler logic would justify a some possible break of
If you disagree with some change in principle, I would like to know your
objection. If there simply was a bug introduced by the change, I
appologize and hope I can fix it.
In any case, I have to point out that the current state of the SVN code
after the partial reversion of my commit is not self-consistent: My
commit changed several files besides axes.py - only reverting this one
file is bound to cause problems.