@tacaswell would you mind checking this?

This looks reasonable, but had a few comments. Good use of a context manager.

Could you please also add a test as well?

Could you compare this approach to #18932 and why you think this is a better path @hukaidong ?

This looks reasonable, but had a few comments. Good use of a context manager.

Could you please also add a test as well?

Thanks, this does make sense. I'll append a test soon.

Could you compare this approach to #18932 and why you think this is a better path @hukaidong ?

I see quite amount of refraction tries to reduce the amount of method provided in artists class, which is not relevant to this topic. So I put them apart. For the work refractoring do_3d_projection, it saves space from storing offset with a cost re-computing projection transform twice. One in do_3d_projection in order to compute the over-all z order in the Axes#draw method. The other in Collection#Draw to get the projection. So, I'd like to say storing an extra offset and swap during drawing time will be cheaper.

@tacaswell I'm not sure if this is the best way to test the issue but I've added a test case there. Also a null guard is set in context manager so we won't have to worry about AttributeError.

This looks reasonable, but had a few comments. Good use of a context manager.
Could you please also add a test as well?

Thanks, this does make sense. I'll append a test soon.

And also, in case, I've checked main branch and new test case failed there as excepted.

@tacaswell Looks I'm not able to do the merge. Who should I call for the merging?

@hukaidong just FYI, we do require two approving reviews from core contributors before merging (for changes to library code) which is why @tacaswell did not previously merge this when he first approved it

@hukaidong just FYI, we do require two approving reviews from core contributors before merging (for changes to library code) which is why @tacaswell did not previously merge this when he first approved it

Awesome! Thanks for letting me know.