GeezerVison

The World Wide Web, restored to 800 pixels of glory.

GeezerVison is showing github.com — Faster image comparison decorator by mdboom · Pull Request #98 · matplotlib/matplotlib

Conversation

mdboom commented Apr 27, 2011

Copy link

Copy Markdown

Member

This is a riff on the ideas in the comments on the pull requests by Paul @ivanov to reduce memory consumption when running the regression tests. See:

#96

and

#56

This is different from the other two in that it:

Calls the test code to generate the figure object only once, and then calls savefig three times on the same figure -- this results in an ~ 12% speedup.
Automatically closes all figures and restores rcParams and the unit registry after every test. The latter prevents some nasty side effect problems presented by some tests. This also revealed a memory leak in pyplot.close('all'). This saves massive amounts of memory vs. the current master, but it roughly equivalent to Paul's approach in pull request 96.
Runs Inkscape in "shell" mode so that a single inkscape process is fired up and re-used to convert each SVG to PNG. This results in an additional ~ 32% speedup.

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

All reactions

mdboom added 5 commits

April 26, 2011 16:29


          Updates to the image comparison testing decorator:

27d4c0c

  1) Run the test itself that generates figure(s) only once, then
     savefig it to each output format.  This appears to have an approx
     2.75x speedup on run time.

  2) Cleanup better after each test:

       Close all figures
       Reset rcParams to their defaults
       Reset the unit registry to defaults

This requires the following changes to the tests themselves:

  1) The tests no longer need to call 'savefig' -- just create figures
     and the decorator will find them.

  2) Tests that don't use the image_comparison decorator should use the
     new cleanup decorator to make sure that cleanup happens.


          Fix memory leak in pyplot.close('all')

354dee8


          Run inkscape for SVG conversion as a server -- prevents starting inks…

a074823

…cape up multiple times.


          Do N SVG->PNG conversions with inkscape before shutting down the inks…

9086ef1

…cape server and starting a new one to prevent memory leaks in Inkscape from becoming problematic.


          Reduce the number of dictionary copies around each test. Call the mai…

84bd649

…n test setup function to reset rcParams state.

mdboom commented Apr 27, 2011

Copy link

Copy Markdown

Member Author

This is moved over from here:

#97

@WeatherGod, @ivanov and @mdboom were participating over there.

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

ivanov commented Apr 28, 2011

Copy link

Copy Markdown

Member

hmm, i'm getting some errors in python2.5:

Failure: AttributeError ('function' object has no attribute 'im_class') ... ERROR matplotlib.tests.test_backend_svg.test_visibility.test ... ok matplotlib.tests.test_basic.test_simple ... ok matplotlib.tests.test_basic.test_simple_knownfail ... KNOWNFAIL: Test known to fail matplotlib.tests.test_basic.test_override_builtins ... ok matplotlib.tests.test_cbook.Test_delete_masked_points.test_bad_first_arg ... ok matplotlib.tests.test_cbook.Test_delete_masked_points.test_datetime ... ok matplotlib.tests.test_cbook.Test_delete_masked_points.test_rgba ... ok matplotlib.tests.test_cbook.Test_delete_masked_points.test_string_seq ... ok matplotlib.tests.test_cbook.test_is_string_like ... ok matplotlib.tests.test_cbook.test_restrict_dict ... ok matplotlib.tests.test_mlab.test_colinear_pca ... ok matplotlib.tests.test_mlab.test_recarray_csv_roundtrip ... FAIL matplotlib.tests.test_mlab.test_rec2csv_bad_shape ... ok matplotlib.tests.test_mlab.test_prctile ... ok matplotlib.tests.test_transforms.test_Affine2D_from_values ... ok Failure: AttributeError ('function' object has no attribute 'im_class') ... ERROR matplotlib.tests.test_axes.test_arc_ellipse.test ... ok matplotlib.tests.test_axes.test_arc_ellipse.test ... ok matplotlib.tests.test_axes.test_arc_ellipse.test ...

and then stalls and never finishes. Is this a nose issue? (I'm on nose 1.0.0)

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

mdboom commented Apr 28, 2011

Copy link

Copy Markdown

Member Author

Ah. I'm having issues with nose 1.0 as well. Investigating...

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.


          Fix new test decorator to work with nose-1.0.0 in addition to nose-0.…

ea16278

…11.3

mdboom commented Apr 28, 2011

Copy link

Copy Markdown

Member Author

This latest commit seems to fix the nose-1.0.0 issue on Python 2.7. @ivanov -- do you mind kicking the tires with Python 2.5? Python 3.x will have its own set of issues, I suspect, but I'll defer that to when I merge this into the py3 branch.

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.


          obscure way of getting to the function, but works in 2.5

be6af2f

see http://bugs.python.org/issue5982 for the source

ivanov commented Apr 28, 2011

Copy link

Copy Markdown

Member

@mdboom - I'm getting a bunch of ERROR: Failure: AttributeError ('staticmethod' object has no attribute '__func__') messages, but once I got around that with ivanov@be6af2f - it still stalls on the third matplotlib.tests.test_axes.test_arc_ellipse.test ...

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

mdboom commented Apr 28, 2011

Copy link

Copy Markdown

Member Author

It's probably hitting some sort of deadlock with the Inkscape server. I've got 0.48.1. But it also could be a difference in the behavior of the "subprocess" module in Python 2.5 vs. 2.7, or some sort of platform difference. If you revert lib/matplotlib/testing/compare.py to what is in master, does the stalling go away? If so, maybe we should defer the Inkscape server part of this pull request until we better understand how to do it in a robust way.

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.


          reverting testing/compare.py back to master

5e50b90

ivanov commented Apr 28, 2011

Copy link

Copy Markdown

Member

mdboom, on 2011-04-28 05:37, wrote:

If you revert lib/matplotlib/testing/compare.py to what is in master, does the stalling go away?

That did the trick, Mike! I still have a couple of failures on
that box, but I also get them on master, so I'll have to look
into that later.

I still needed to use the get(1) business, instead of
func - see https://github.com/ivanov/matplotlib/commits/faster-image-comp

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

mdboom added 3 commits

April 28, 2011 12:46


          Fix indentation.

cc644a8


          Merge remote-tracking branch 'ivanov/faster-image-comp' into faster_i…

43ffb29

…mage_comparison_decorator

Fix for Python 2.5 compatibility.


          Revert back to using inkscape once for each SVG->PNG conversion, rath…

bffb15d

…er than running it as a server, which mysteriously causes a deadlock on some systems.

ivanov commented Apr 28, 2011

Copy link

Copy Markdown

Member

looks good now, Mike, thanks!

All reactions

Sorry, something went wrong.

Uh oh!

There was an error while loading. Please reload this page.

mdboom added a commit that referenced this pull request

May 2, 2011


          Merge pull request #98 from mdboom/faster_image_comparison_decorator.

fdf5fae

Faster image comparison decorator

mdboom merged commit fdf5fae into matplotlib:master

May 2, 2011

richbwood mentioned this pull request

Dec 19, 2012

test_pcolormesh hangs #1609

Closed

anntzer mentioned this pull request

Mar 8, 2017

Investigate alternative svg renderers for the test suite #8242

Closed

Reviewers

No reviews

Uh oh!

There was an error while loading. Please reload this page.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Uh oh!

There was an error while loading. Please reload this page.

3 participants

Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.