Fix gtk3agg alpha channel. #10297
Merged
Fix gtk3agg alpha channel. #10297
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Is there a test for this? |
|
Basically run the example code yourself and observe the background color of the figure as explained above... (not sure it can be automated in any reasonable way) |
This was referenced Jan 23, 2018
QuLogic
reviewed
Feb 4, 2018
| if sys.byteorder == "little": | ||
| alpha = buf[..., -1:] | ||
| if alpha.min() != 0xff: | ||
| buf = (np.dstack([buf[..., :3] * (alpha / 0xff), alpha]) |
There was a problem hiding this comment.
Do you need to make a copy? Isn't buf an independent thing, so can you do an in-place multiply?
There was a problem hiding this comment.
was a bit more tricky to write correctly but fixed it
9d62bb0 to
fd9da34
Compare
|
When I try out your example, I get an error: Traceback (most recent call last):
File "/home/elliott/code/matplotlib/lib/matplotlib/backends/backend_gtk3agg.py", line 76, in on_draw_event
buf.data, cairo.FORMAT_ARGB32, width, height)
File "/usr/lib/python3.6/site-packages/cairocffi/surfaces.py", line 660, in create_for_data
return cls(format, width, height, data, stride)
File "/usr/lib/python3.6/site-packages/cairocffi/surfaces.py", line 649, in __init__
% (length, stride * height))
ValueError: Got a 640 bytes buffer, needs at least 1228800. |
fd9da34 to
04e22f3
Compare
|
Ah, that's because I use pycairo (which actually works with py3 now...) rather than cairocffi. Ravel()ing the buffer before passing it to create_for_data fixes the problem and works for both pycairo and cairocffi, pushed that. |
QuLogic
approved these changes
Feb 9, 2018
Should we remove the warning then? |
|
Yup. Have another PR in the works for that. |
|
rebased |
The gtk3agg backend works by drawing the ARGB32 buffer (from Agg) onto a
cairo context (passed by gtk3). However, cairo wants a *premultiplied*
ARGB32 buffer (i.e., where 100% blue with 50% transparency is
represented by (r=0%, g=0%, b=50%, a=50%) instead of (r=0%, g=0%,
b=100%, a=50%), which we didn't do before.
This is only apparent if the entire buffer contains some transparency,
e.g. if the figure background is transparent. Consider e.g. under
gtk3agg:
from pylab import *
rcParams["figure.facecolor"] = (0, 0, 0, 0)
gca()
show()
Without the patch, the area surrounding the axes is white (because of
the misinterpretation of premultiplied alpha). With the patch, it is
(correctly) gray, which is the background color of the gtk widget.
(Note that when running the example under qt5agg or tkagg, the situation
is complicated by the fact that these backends themselves set the widget
background color to white rather than gray.)
As a comparison point, qt5agg builds the QImage using Format_ARGB32,
instead of Format_ARB32_Premultiplied; i.e. Qt provides its own
conversion from non-premultiplied to premultiplied.
The premultiplication step involves allocating a full new buffer, so
check whether there is actually any transparency before doing it.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
The gtk3agg backend works by drawing the ARGB32 buffer (from Agg) onto a
cairo context (passed by gtk3). However, cairo wants a premultiplied
ARGB32 buffer (i.e., where 100% blue with 50% transparency is
represented by (r=0%, g=0%, b=50%, a=50%) instead of (r=0%, g=0%,
b=100%, a=50%), which we didn't do before.
This is only apparent if the entire buffer contains some transparency,
e.g. if the figure background is transparent. Consider e.g. under
gtk3agg:
Without the patch, the area surrounding the axes is white (because of
the misinterpretation of premultiplied alpha). With the patch, it is
(correctly) gray, which is the background color of the gtk widget.
(Note that when running the example under qt5agg or tkagg, the situation
is complicated by the fact that these backends themselves set the widget
background color to white rather than gray.)
As a comparison point, qt5agg builds the QImage using Format_ARGB32,
instead of Format_ARB32_Premultiplied; i.e. Qt provides its own
conversion from non-premultiplied to premultiplied.
The premultiplication step involves allocating a full new buffer, so
check whether there is actually any transparency before doing it.
PR Summary
PR Checklist