Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
image (PR#11493)
9 messages · jms2003 at med.cornell.edu, Henrik Bengtsson, Simon Urbanek +3 more
On Tue, May 20, 2008 at 5:05 PM, <jms2003 at med.cornell.edu> wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
<sarcasm> ...and so are bug reports without reproducible examples. </sarcasm> More seriously, for the hardworking R core people [I'm not one] spending their time going through bug reports (and people on R-devel list), please imagine yourself in their position and you realize quickly that you can help them out with by giving a reproducible example that they can cut'n'paste into their R session. Also, reporting the output of sessionInfo() is considered good etiquette. In your case you might even be able to submit "before and after" PNGs/PDFs illustrating the difference. Cheers Henrik
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device. The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti- aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium). Cheers, Simon
Sorry for lack of clarity but I didn't find away to upload images.
I am running Mac OS 10.5.2, R 2.7.0
The problem arrises when using anything that depends upon image()
using the Quartz() device. This sounds very much like what you are
describing with the background showing through (most obvious with tiny
boxes in the example, n=100000). The problem does not occur when I use
an x11 device. I did not have the problem using quartz as a screen
device prior to upgrading to 2.7.0. (Although I did have some odd
behavior when I saved images from the quartz device as postscript or
related file formats.)
Do you know of a workaround? I have tried setting quartz(antialias=F)
but still have the problem.
tempF<-function(n) {
im<- matrix(0,nrow=n,ncol=5)
for (i in 1:5) {
im[,i] <- seq(1,n)
}
image(im , col = topo.colors(100))
}
tempF(10)
tempF(1000)
tempF(100000)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Summary.001.png
Type: image/png
Size: 30555 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-devel/attachments/20080521/b1408526/attachment.png>
-------------- next part --------------
On May 21, 2008, at 10:22 AM, Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device. The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti-aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium). Cheers, Simon
Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura
Version: 2.7.0
OS: Mac 10.5
Submission from: (NULL) (140.251.50.94)
Since updating to 2.7.0 all plots that use image() (heatmap, etc...)
now draw
visible boxes around each rectangle in the plot. When there are many
rectangles
the surrounding color becomes dominant over the rectangle color and
the overall
image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device.
I see lines at the borders of the grid used by image on my Mac (Tiger, R 2.7.0 Patched (2008-05-20 r45743)). They are most visible in the last of the examples plotted by example(image), the one that starts image(x, y, volcano, col = terrain.colors(100), axes = FALSE). It opens a Quartz device. How do you do a snapshot on a Mac? I see online that it's Cmd-Shift-4, and I get the snapshot click, but I don't know where the picture ended up.
The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti- aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium).
I don't know if what I'm seeing is new or not; I've only got one R version installed. Duncan Murdoch
Cheers, Simon
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
(Edited to add link to sample picture)
Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura
Version: 2.7.0
OS: Mac 10.5
Submission from: (NULL) (140.251.50.94)
Since updating to 2.7.0 all plots that use image() (heatmap, etc...)
now draw
visible boxes around each rectangle in the plot. When there are many
rectangles
the surrounding color becomes dominant over the rectangle color and
the overall
image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device.
I see lines at the borders of the grid used by image on my Mac (Tiger, R 2.7.0 Patched (2008-05-20 r45743)). They are most visible in the last of the examples plotted by example(image), the one that starts image(x, y, volcano, col = terrain.colors(100), axes = FALSE). It opens a Quartz device. How do you do a snapshot on a Mac? I see online that it's Cmd-Shift-4, and I get the snapshot click, but I don't know where the picture ended up. AHA! It goes to the desktop. Okay, a sample picture is available at http://www.stats.uwo.ca/faculty/murdoch/temp/grid.png Not as bad as Joseph was describing, but not nearly as good as Windows produces ;-).
The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti- aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium).
I don't know if what I'm seeing is new or not; I've only got one R version installed. Duncan Murdoch
Cheers, Simon
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
The effect Duncan's picture shows is typical of using anti-aliasing for rectangles and polygons. The cairo-based devices have it turned off for filled regions, as it seems to have no advantage for R uses of such regions. (You also see it with some on-screen renderers of postscript or pdf versions of this plot, e.g. those based on ghostscript.) However, this does not match the original report, so I think we need a reproducible example (with screenshot) from Joseph Scandura to make any further progress.
On Wed, 21 May 2008, Duncan Murdoch wrote:
(Edited to add link to sample picture) Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device.
I see lines at the borders of the grid used by image on my Mac (Tiger, R 2.7.0 Patched (2008-05-20 r45743)). They are most visible in the last of the examples plotted by example(image), the one that starts image(x, y, volcano, col = terrain.colors(100), axes = FALSE). It opens a Quartz device. How do you do a snapshot on a Mac? I see online that it's Cmd-Shift-4, and I get the snapshot click, but I don't know where the picture ended up. AHA! It goes to the desktop. Okay, a sample picture is available at http://www.stats.uwo.ca/faculty/murdoch/temp/grid.png Not as bad as Joseph was describing, but not nearly as good as Windows produces ;-).
The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti- aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium).
I don't know if what I'm seeing is new or not; I've only got one R version installed. Duncan Murdoch
Cheers, Simon
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
______________________________________________ R-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595
5 days later
Sorry for lack of clarity in my original message but I'm new to this
list and I couldn't find away to upload images.
I am running Mac OS 10.5.2, R 2.7.0
The problem arrises when using anything that depends upon image()
using the Quartz() device. This sounds very much like what you are
describing with the background showing through (most obvious with tiny
boxes in the example, n=100000). The problem does not occur when I use
an x11 device. It sounds like this is an old problem without a good
solution. What puzzles me is that I did not have the problem using
quartz as a screen device prior to upgrading to 2.7.0.
Do you know of a workaround? I have tried setting quartz(antialias=F)
but still have the problem.
tempF<-function(n) {
im<- matrix(0,nrow=n,ncol=5)
for (i in 1:5) {
im[,i] <- seq(1,n)
}
image(im , col = topo.colors(100))
}
tempF(10)
tempF(1000)
tempF(100000)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Summary.001.png
Type: image/png
Size: 30555 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-devel/attachments/20080527/c4acaebd/attachment.png>
-------------- next part --------------
On May 21, 2008, at 10:22 AM, Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device. The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti-aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium). Cheers, Simon
Joseph, please try a more recent R, I have addressed the issue in R-devel/ R-2.7-patched after your report. Cheers, Simon
On May 27, 2008, at 6:07 PM, Joseph Scandura wrote:
Sorry for lack of clarity in my original message but I'm new to this
list and I couldn't find away to upload images.
I am running Mac OS 10.5.2, R 2.7.0
The problem arrises when using anything that depends upon image()
using the Quartz() device. This sounds very much like what you are
describing with the background showing through (most obvious with
tiny boxes in the example, n=100000). The problem does not occur
when I use an x11 device. It sounds like this is an old problem
without a good solution. What puzzles me is that I did not have the
problem using quartz as a screen device prior to upgrading to 2.7.0.
Do you know of a workaround? I have tried setting
quartz(antialias=F) but still have the problem.
tempF<-function(n) {
im<- matrix(0,nrow=n,ncol=5)
for (i in 1:5) {
im[,i] <- seq(1,n)
}
image(im , col = topo.colors(100))
}
tempF(10)
tempF(1000)
tempF(100000)
<Summary.001.png>
On May 21, 2008, at 10:22 AM, Simon Urbanek wrote:
On May 20, 2008, at 8:05 PM, jms2003 at med.cornell.edu wrote:
Full_Name: Joseph Scandura Version: 2.7.0 OS: Mac 10.5 Submission from: (NULL) (140.251.50.94) Since updating to 2.7.0 all plots that use image() (heatmap, etc...) now draw visible boxes around each rectangle in the plot. When there are many rectangles the surrounding color becomes dominant over the rectangle color and the overall image is borderline useless.
Can you, please, specify exactly which graphics device you are using and possibly a snapshot of the problem? I don't see any additional boxes being drawn on any device. The only issue I'm aware of are anti-aliasing effects around the edges of adjacent rectangles which don't fall on the pixel boundary (if anti-aliasing device is used). Depending on the subpixel location of the edge, the background color may shine through very slightly. It's not what you describe, but it's closest to what I can imagine you could mean. However, AFAICS this has not been changed recently and is a rendering artifact which is hard to get rid of in the current setup as devices are resolution-independent (the only cure I'm aware of [short of disabling anti-aliasing] is to distort the original plot such that rectangles are aligned with the pixels of the output medium). Cheers, Simon