Adding Alpha Masks to the Quartz Graphics Device

[hturner]

This is a continuation of work from R Project Sprint 2023 (#43) and will be working with Paul Murrell (@pmur002).

REQUIREMENTS: one or two people that can build R on macOS, apply the patch and runs some tests.

TODO

• Basic manual tests
• {gdiff} quartz() output for alpha masks with pdf() output (see comment below for why this is struck out)
• Compare with old (luminance) Quartz output (see suggested steps below)
• Run make check-all
  • On Linux
  • On macOS
• Commit with thanks to contributors!

[pmur002]

Code for manual test ...

library(grid)

ac <- circleGrob(r=.3, gp=gpar(col=NA, fill=rgb(0,0,0,.5)))
lc <- circleGrob(r=.3, gp=gpar(col=NA, fill=grey(.5)))
    
r1 <- rectGrob(width=.2, gp=gpar(fill=2))
r2 <- rectGrob(height=.2, gp=gpar(fill=3))

## No masks
grid.newpage()
grid.draw(r1)
grid.draw(r2)

## Alpha mask
grid.newpage()
grid.draw(r1)
pushViewport(viewport(mask=as.mask(ac, "alpha")))
grid.draw(r2)

## Luminance mask
grid.newpage()
grid.draw(r1)
pushViewport(viewport(mask=as.mask(lc, "luminance")))
grid.draw(r2)

[pmur002]

Attached expected output from manual test



.

[pmur002]

Code for mask tests in src/library/grid/tests/masks.R

[pmur002]

Steps for generating suite of test images for new alpha masks ...

1. Create directory, say quartzLuminance
2. Within that directory, run r-release , start quartz(type="png") device and run code in src/grid/tests/masks.R. That should generate multiple files, Rplot001.png, Rplot002.png, ... (these should be luminance masks, thanks to devMask() in masks.R)
3. Create directory, say quartzAlpha
4. Within that directory, run patched R , start quartz(type="png") device and run code in src/grid/tests/masks.R. That should generate multiple files, Rplot001.png, Rplot002.png, ... (these should be alpha masks, thanks to devMask() in masks.R)
5. Within parent directory, run something like ...

library(gdiff)
gdiffCompare("quartzLuminance", "quartzAlpha", "quartzDiff")

6. The code above should provide printout of identical/differing/unmatched PNGs. The quartzDiff directory should contain images showing any differences between non-identical PNGs

[pmur002]

Comparing quartz(type="pdf") with pdf() unfortunately does NOT produce identical images. Also, comparing png(type="quartz") with png(type="cairo") does NOT produce identical images. This is why the second step at the top is struck through. On the other hand these differences could be seen as bugs and future dev day issues themselves :)

[hturner]

Some minor changes to the above steps:

• Test code is in src/library/grid/tests/masks.R
• Need to use quartz(type = "png", bg = "white") otherwise the test plots overplot each other
• Need to remember to run dev.off() at the end, otherwise might miss last plot!

With these changes, the output is identical in the test and control cases.

[hturner]

make check-all runs without issue - I think this is ready to merge in !