v1.4 feature freeze + release timeline

Sorry for dropping the ball on moving things forward.

I would propose a feature freeze this weekend and the goal of getting
the first RC out by the end of the month. We should probably plan a
hangout sometime in between now and then if there is stuff we want to
work out at a lower latency.

There are 62 open issues tagged as 1.4.0, but there are a number of
issues which have both an issue and a PR so it is more like ~40 open
issues. At least some of these will probably be punted.

The major features that are outstanding and need work/discussion:
- grid lines in polar plots (#1441, #1730, #1847, #2203
- Qt5 (#2471) do we want to squeeze this in? ipython has it working
so I don't think it will be too bad (despite all the furry in that
issue).
- turning down the garbage collection calls (#3045)
- #2986 axes order for event handling

Bugs this still need patches:
- auto-scale + sharex/sharey + aspect heuristics don't always work
(#1325, #2492). I could be convinced to punt this as it is 'fixed' by
just setting the limits by hand + the forcing parameters.
- #2856 / #2900 issues with saving animations
- #3065 font priority test failure
- #2698 deal with empty lines in text + box
+ #3039 tornado should not be an install dependency
+ #3042 revert changes to default style in boxplot
  - #2941 / #3012 calling cla() in `set_ticks_position`
  - #2945 issues with hist, patch in issue
  - #2977 pyqt string validation (patch in issue)

Bug PRs that need minor work:
  - #1981 / #2662 mark-every upgrade
  - #2150 / #2942 fixing issues with auto-limit + 0 height bar plots
  - #2394 / #2951 issues in mpl_toolkits.axes_grid1.anchored_artists
  - #2475 / #2843 contour labels (release critical)
  - #3000 / #3017 multi-line in pgf

Bug PRs that need major work:
- #1012, animation related PR from jdh.
- #2366 / #2738 errorbar + linestyle (this is one of mine, Eric does
not like it)
- #2828 / #2927 issue with mutable objects, ==, and hash

Documentation:
  - #2350 annotation arrow example, code in issue needs PR
  - #2742 how to do a re-base (mine, needs work)
  - #2796 offset on default formatter documentation
  - #2603 suggested edit in issue
  - #2908 404 in docs
  - #2918 re-write contribution guide lines to match reality
  - #2929 doc about clip-on
  - #2999 install documentation
  - #3004 remove all references to nxutils
  - #3035 documentation in Polygon object
  - #3055 maybe add warning to get_window_extent

PRs that need review:
- #2288 / #3053
- #2602 / #3015 fix stem plots
- #2822 (documentation clarification)
- #2952 / #2518 turn off clipping on all pie charts

feature PRs that need minor work:
- #2855 adds flag to spy, pep8 issue
- #2541 / 2874 fixes mouse over text in polar mode (rebase)
- #2961 3D quiver
- #2996 violin plots

No idea what to do with
- #2025, tkagg + osx framework build stuff

Maybe resolved:
- issues with the screen not matching the saved images (#1441, #2935,
#2831, and #2889) might all be fixed by #3054

I think I am going to start just doing merges with CHANGELOG conflicts
on my local computer and pushing to matplotlib/master as the amount of
re-bases we end up needing is getting absurd.

Tom

···

--
Thomas Caswell
tcaswell@...149...

Sorry for dropping the ball on moving things forward.

On the contrary, you have been doing a great job on this, and you deserve a lot of thanks from all of us.

I would propose a feature freeze this weekend and the goal of getting
the first RC out by the end of the month. We should probably plan a
hangout sometime in between now and then if there is stuff we want to
work out at a lower latency.

Feature freeze: good.

There are 62 open issues tagged as 1.4.0, but there are a number of
issues which have both an issue and a PR so it is more like ~40 open
issues. At least some of these will probably be punted.

Agreed!
It's a daunting list.

The major features that are outstanding and need work/discussion:
  - grid lines in polar plots (#1441, #1730, #1847, #2203
  - Qt5 (#2471) do we want to squeeze this in? ipython has it working
so I don't think it will be too bad (despite all the furry in that
issue).
  - turning down the garbage collection calls (#3045)
  - #2986 axes order for event handling

Bugs this still need patches:
  - auto-scale + sharex/sharey + aspect heuristics don't always work
(#1325, #2492). I could be convinced to punt this as it is 'fixed' by
just setting the limits by hand + the forcing parameters.
  - #2856 / #2900 issues with saving animations
  - #3065 font priority test failure
  - #2698 deal with empty lines in text + box
  + #3039 tornado should not be an install dependency
  + #3042 revert changes to default style in boxplot
   - #2941 / #3012 calling cla() in `set_ticks_position`
   - #2945 issues with hist, patch in issue
   - #2977 pyqt string validation (patch in issue)

Bug PRs that need minor work:
   - #1981 / #2662 mark-every upgrade
   - #2150 / #2942 fixing issues with auto-limit + 0 height bar plots
   - #2394 / #2951 issues in mpl_toolkits.axes_grid1.anchored_artists
   - #2475 / #2843 contour labels (release critical)
   - #3000 / #3017 multi-line in pgf

Bug PRs that need major work:
  - #1012, animation related PR from jdh.
  - #2366 / #2738 errorbar + linestyle (this is one of mine, Eric does
not like it)
  - #2828 / #2927 issue with mutable objects, ==, and hash

Documentation:
   - #2350 annotation arrow example, code in issue needs PR
   - #2742 how to do a re-base (mine, needs work)
   - #2796 offset on default formatter documentation
   - #2603 suggested edit in issue
   - #2908 404 in docs
   - #2918 re-write contribution guide lines to match reality
   - #2929 doc about clip-on
   - #2999 install documentation
   - #3004 remove all references to nxutils
   - #3035 documentation in Polygon object
   - #3055 maybe add warning to get_window_extent

PRs that need review:
  - #2288 / #3053
  - #2602 / #3015 fix stem plots
  - #2822 (documentation clarification)
  - #2952 / #2518 turn off clipping on all pie charts

feature PRs that need minor work:
  - #2855 adds flag to spy, pep8 issue
  - #2541 / 2874 fixes mouse over text in polar mode (rebase)
  - #2961 3D quiver
  - #2996 violin plots

No idea what to do with
  - #2025, tkagg + osx framework build stuff

Maybe resolved:
  - issues with the screen not matching the saved images (#1441, #2935,
#2831, and #2889) might all be fixed by #3054

I think I am going to start just doing merges with CHANGELOG conflicts
on my local computer and pushing to matplotlib/master as the amount of
re-bases we end up needing is getting absurd.

Maybe this problem could be reduced by having a directory for CHANGELOG chunks, one file per PR, with the PR number as the filename. Then at release time they could be concatenated, edited, appended to the real CHANGELOG, and deleted.

Another approach would be to have the CHANGELOG entry as an identifiable chunk in the commit message. Then they would be extracted and dealt with at release time, as above.

Eric

···

On 2014/05/15 5:16 PM, Thomas Caswell wrote:

Tom

I like this idea a lot. PR number as name is a little awkward as you
don't know the PR# till you submit the PR. As long as they all go in
the same folder and we use a quasi-unique scheme
(YYYY-MM-DD-[initials]_[n].txt) to name them it should be fine

···

On Fri, May 16, 2014 at 4:17 AM, Eric Firing <efiring@...229...> wrote:

Maybe this problem could be reduced by having a directory for CHANGELOG
chunks, one file per PR, with the PR number as the filename. Then at
release time they could be concatenated, edited, appended to the real
CHANGELOG, and deleted.

--
Thomas Caswell
tcaswell@...149...

Thanks for doing all this work, 1.4 is more than overdue!

I justed wanted to chime in on Qt 5: Put it in and mark
the support as experimental. So we can fix it in minor
releases, also without exposure bugs won't be found.

Till

Sorry for dropping the ball on moving things forward.

On the contrary, you have been doing a great job on this, and you
deserve a lot of thanks from all of us.

Yes -- I can't thank you enough. I've been feeling bad about getting being so overstretched on non-matplotlib things these past few months (in related news, STScI is hiring!) but I know it's in capable hands, even if I don't have time to dig in and help like I used to.

I would propose a feature freeze this weekend and the goal of getting
the first RC out by the end of the month. We should probably plan a
hangout sometime in between now and then if there is stuff we want to
work out at a lower latency.

Feature freeze: good.

There are 62 open issues tagged as 1.4.0, but there are a number of
issues which have both an issue and a PR so it is more like ~40 open
issues. At least some of these will probably be punted.

Agreed!
It's a daunting list.

The major features that are outstanding and need work/discussion:
   - grid lines in polar plots (#1441, #1730, #1847, #2203
   - Qt5 (#2471) do we want to squeeze this in? ipython has it working
so I don't think it will be too bad (despite all the furry in that
issue).
   - turning down the garbage collection calls (#3045)
   - #2986 axes order for event handling

Bugs this still need patches:
   - auto-scale + sharex/sharey + aspect heuristics don't always work
(#1325, #2492). I could be convinced to punt this as it is 'fixed' by
just setting the limits by hand + the forcing parameters.
   - #2856 / #2900 issues with saving animations
   - #3065 font priority test failure
   - #2698 deal with empty lines in text + box
   + #3039 tornado should not be an install dependency
   + #3042 revert changes to default style in boxplot
    - #2941 / #3012 calling cla() in `set_ticks_position`
    - #2945 issues with hist, patch in issue
    - #2977 pyqt string validation (patch in issue)

Bug PRs that need minor work:
    - #1981 / #2662 mark-every upgrade
    - #2150 / #2942 fixing issues with auto-limit + 0 height bar plots
    - #2394 / #2951 issues in mpl_toolkits.axes_grid1.anchored_artists
    - #2475 / #2843 contour labels (release critical)
    - #3000 / #3017 multi-line in pgf

Bug PRs that need major work:
   - #1012, animation related PR from jdh.
   - #2366 / #2738 errorbar + linestyle (this is one of mine, Eric does
not like it)
   - #2828 / #2927 issue with mutable objects, ==, and hash

Documentation:
    - #2350 annotation arrow example, code in issue needs PR
    - #2742 how to do a re-base (mine, needs work)
    - #2796 offset on default formatter documentation
    - #2603 suggested edit in issue
    - #2908 404 in docs
    - #2918 re-write contribution guide lines to match reality
    - #2929 doc about clip-on
    - #2999 install documentation
    - #3004 remove all references to nxutils
    - #3035 documentation in Polygon object
    - #3055 maybe add warning to get_window_extent

PRs that need review:
   - #2288 / #3053
   - #2602 / #3015 fix stem plots
   - #2822 (documentation clarification)
   - #2952 / #2518 turn off clipping on all pie charts

feature PRs that need minor work:
   - #2855 adds flag to spy, pep8 issue
   - #2541 / 2874 fixes mouse over text in polar mode (rebase)
   - #2961 3D quiver
   - #2996 violin plots

No idea what to do with
   - #2025, tkagg + osx framework build stuff

Maybe resolved:
   - issues with the screen not matching the saved images (#1441, #2935,
#2831, and #2889) might all be fixed by #3054

I think I am going to start just doing merges with CHANGELOG conflicts
on my local computer and pushing to matplotlib/master as the amount of
re-bases we end up needing is getting absurd.

Maybe this problem could be reduced by having a directory for CHANGELOG
chunks, one file per PR, with the PR number as the filename. Then at
release time they could be concatenated, edited, appended to the real
CHANGELOG, and deleted.

Another approach would be to have the CHANGELOG entry as an identifiable
chunk in the commit message. Then they would be extracted and dealt
with at release time, as above.

Another solution for CHANGELOG entries that we use in the astropy project is to have section headers in the CHANGELOG file upfront (put in at the beginning of each release cycle). Then merge conflicts tend to be localized to one subsystem and are less frequent. But that's only a partial solution. Eric's suggestions also make a lot of sense.

Mike

···

On 05/16/2014 04:17 AM, Eric Firing wrote:

On 2014/05/15 5:16 PM, Thomas Caswell wrote:

Eric

Tom

------------------------------------------------------------------------------
"Accelerate Dev Cycles with Automated Cross-Browser Testing - For FREE
Instantly run your Selenium tests across 300+ browser/OS combos.
Get unparalleled scalability from the best Selenium testing platform available
Simple to use. Nothing to install. Get started now for free."
http://p.sf.net/sfu/SauceLabs
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
matplotlib-devel List Signup and Options

--
Michael Droettboom
Science Software Branch
Space Telescope Science Institute