Fix for empty collection check in axes.add_collection #1497

Merged
merged 4 commits into from Feb 17, 2013

5 participants

@akhmerov

Fixes #1490

@ivanov
Matplotlib Developers member

looks good, @akhmerov, thanks for the fix. Could you add a test for this, so we don't run into it again - something like the original report in #1490?

@akhmerov

Done. As I mentioned, this is not a full fix of #1490, and another issue has to be opened to ensure that datalimits of empty collections are calculated properly.

@NelleV NelleV commented on the diff Nov 14, 2012
lib/matplotlib/axes.py
@@ -1491,9 +1491,8 @@ def add_collection(self, collection, autolim=True):
if collection.get_clip_path() is None:
collection.set_clip_path(self.patch)
- if autolim:
- if collection._paths and len(collection._paths):
- self.update_datalim(collection.get_datalim(self.transData))
+ if autolim and collection._paths and len(collection._offsets):
@NelleV
NelleV added a line comment Nov 14, 2012

Tiny nitpick on something you haven't written: we should not test for an emtpy list using len(list) but just list.
An empty list evaluates as false in python.

Maybe you can replace:

if autolim and collection._paths and len(collection._offsets):

by

if autolim and collection._paths and collection._offsets:
@akhmerov
akhmerov added a line comment Nov 14, 2012

As I wrote in a comment earlier, collection._offsets is a numpy array, not a list. This means that bool(np.array([0])) == False, and bool(np.array([0, 0])) raises an error.

@NelleV
NelleV added a line comment Nov 14, 2012

My mistake.

I'd then use the shape method instead of the length attribute, but that's too much nitpicking :)

Thanks for clarifying.

@akhmerov
akhmerov added a line comment Nov 14, 2012

Behavior of len is well-defined and documented for numpy arrays, so I still believe it's the better way due to improved readability.

@WeatherGod
Matplotlib Developers member
WeatherGod added a line comment Nov 14, 2012

I agree with @akhmerov on this point. I personally don't like the idiom of the empty list evaluating to false partly because of this reason. I work with numpy arrays so much. My second reason is that an iterator to an empty list evaluates to True, and so could cause a lot of confusion when coding in py3k with iterators being so prevelent there.

@akhmerov
akhmerov added a line comment Nov 15, 2012

@WeatherGod Well, iterators don't necessarily have length, so some assumptions about collection._paths must be made, and it seems to be a list in all cases that occur so far. I think that without going beyond the scope of this pull request, there's no further improvement to the check. What I find a bit ugly is that both _paths and _offsets are internal to collection, and should preferably not be used from outside of collections.py .

Is there anything else to be done with this pull request?

@WeatherGod
Matplotlib Developers member
WeatherGod added a line comment Nov 15, 2012

The more and more I look at this PR, the more I can't bring myself to accept it. I would rather fix the underlying problem.

@akhmerov
akhmerov added a line comment Nov 15, 2012

This PR doesn't introduce yet another special condition, it rather corrects the condition which already existed before (I even suspect the original condition was a typo, since the second clause was pointless). In this sense it doesn't hurt anything. Even if path.get_path_collection_extents would be working properly, this condition could be reasonable to keep.

@WeatherGod
Matplotlib Developers member
WeatherGod added a line comment Dec 3, 2012

The problem that can now occur is if collection._offsets is None. Previously, the check on collection._paths would be sufficient to prevent an exception from being thrown when doing a len() on None. Now that protection is gone.

@akhmerov
akhmerov added a line comment Dec 3, 2012

Line 1440 of collections.py in my fork of matplotlib uses the same assumption:

if len(self._offsets):
    xs = self.convert_xunits(self._offsets[:0])

Additionally, searching for all occurences of _offsets, I cannot confirm that they can ever assume a None value, instead _offsets are always an array.

@pelson
Matplotlib Developers member
pelson added a line comment Sep 23, 2013

@WeatherGod / @akhmerov - I've submitted a PR which moves this line to the Collection class in v1.3.1: #2444

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
@dmcdougall
Matplotlib Developers member

What's the status on this? Is it good to go? The tests pass but I'm aware @WeatherGod has some concerns. How do others feel?

@efiring
Matplotlib Developers member

I don't see any reason not to merge it.

@WeatherGod
Matplotlib Developers member

I would be willing to back off if we make a new issue to investigate the root reason for the bug that this is papering over. We just need to figure out a way to still be able to test for that bug even with this fix.

@akhmerov

This is also my suggestion (I didn't fix the underlying issue due to lacking time/cpp skill). Testing the underlying issue would be easy: one just needs to check output on path.get_path_collection_extents when supplied an empty collection, as I described in #1490.

@dmcdougall
Matplotlib Developers member

Ok sounds like, as @WeatherGod said, if we open a separate issue for the underlying problem we are pretty happy to merge this PR.

@WeatherGod Would you mind opening a new issue?

@WeatherGod
Matplotlib Developers member
@dmcdougall
Matplotlib Developers member

I think that's a fine idea.

@akhmerov

A known failure test would require knowing what is the correct return value for path.get_path_collection_extents if the collection is empty. I don't think this is set anywhere as of now.

@akhmerov

Since creating a KnownFailure test would require deciding first, what should happen if the test was not to fail, and since no such action was taken, I have created #1735 to not lose track of the bug in _path.get_path_collection_extents. I think this PR can be now merged and #1490 closed.

@mdboom mdboom commented on an outdated diff Feb 4, 2013
lib/matplotlib/tests/test_axes.py
@@ -54,6 +54,18 @@ def test_formatter_ticker():
ax.set_xlabel( "x-label 005" )
ax.autoscale_view()
+def test_add_collection():
@mdboom
Matplotlib Developers member
mdboom added a line comment Feb 4, 2013

This needs to have the @cleanup decorator to the figures get cleared after the test.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
@pelson pelson commented on the diff Feb 14, 2013
lib/matplotlib/tests/test_axes.py
@@ -54,6 +54,19 @@ def test_formatter_ticker():
ax.set_xlabel( "x-label 005" )
ax.autoscale_view()
+@cleanup
+def test_add_collection():
@pelson
Matplotlib Developers member
pelson added a line comment Feb 14, 2013

I think this test needs a comment re what the purpose of the test is.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
@efiring efiring merged commit 99d1a2f into matplotlib:master Feb 17, 2013

1 check failed

Details default The Travis build could not complete due to an error
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment