R cairo_pdf function does not respect plotting boundaries

Hi Paul,

Great, thank you for looking in to this, and I'm glad that you're able 
to reproduce it at your end too.

 From your reply, I'm happy that it seems like the fix may be fairly 
trivial, but I understand the necessity for caution.

If there's anything else I can do to help, please do let me know.

Thank you again,
Best,
Lee


On Monday, 4 March 2019, Paul Murrell <paul at stat.auckland.ac.nz 
<mailto:paul at stat.auckland.ac.nz>> wrote:

    Hi

    (cc'ed to r-devel where further discussion should probably take place)

    Thanks Lee.? I see that problem.

    There is a "+ 1" in the Cairo device code for setting the clipping
    region
    (https://github.com/wch/r-source/blob/ba600867f2a94e46cf9eb75dc8b37f12b08a4561/src/library/grDevices/src/cairo/cairoFns.c#L156
    <https://github.com/wch/r-source/blob/ba600867f2a94e46cf9eb75dc8b37f12b08a4561/src/library/grDevices/src/cairo/cairoFns.c#L156>)


    Remove the "+ 1" and the problem goes away (for your example at least).

    The comment on the line above that code suggests that the "+ 1" was
    modelled on the X11 device code, but X11 deals in integer pixels and
    Cairo (at the API level) does not, so it would seem that the "+ 1"
    is just unnecessary.

    However, I have a slight nagging worry that we have been here
    before, so I would like to do some more testing before committing
    that change.

    Paul

    On 1/03/19 8:13 AM, Lee Steven Kelvin wrote:

        Hello all,

        When producing a plot in R using the cairo_pdf device, the
        resultant plot does not respect the plotting boundaries. Lines
        and shaded regions will spill over the lower x-axis and the
        right-side y-axis (sides 1 and 4). I would like to know if it is
        possible to fix this behaviour when using 'cairo_pdf' in R?

        As an example, see the image at this web link:
        https://i.stack.imgur.com/0lfZd.png
        <https://i.stack.imgur.com/0lfZd.png>

        This image is a screenshot of a PDF file constructed using the
        following minimum working example code:

        cairo_pdf(file="test.pdf", width=0.5, height=0.5)
        par("mar"=c(0.25,0.25,0.25,0.25))
        plot(NA, xlim=c(0,1), ylim=c(0,1), axes=FALSE)
        polygon(x=c(-1,-1,2,2), y=c(-1,2,2,-1), density=5, col="green3",
        lwd=10)
        abline(h=0.25, col="red", lwd=5)
        abline(h=0.75, col="hotpink", lwd=5, lend=1)
        abline(v=0.25, col="blue", lwd=5)
        abline(v=0.75, col="cyan", lwd=5, lend=1)
        box()
        dev.off()

        Here I'm plotting a shaded region in green using 'polygon', with
        boundaries that lie outside the plot. I'm also drawing two sets
        of horizontal/vertical lines using 'abline'. The first in each
        pair uses standard rounded line caps, whilst the second in each
        pair uses butt line caps.

        As you can see, the shading lines and the default rounded-end
        ablines all extend beyond the plotting region along the lower
        and right-hand side axes. Only when using 'lend=1' am I able to
        contain the ablines to the plotting region. I know of no such
        fix for the shading lines however.

        I would naively expect the R plotting region to be respected,
        and for it to be impossible to plot outside of this region
        unless explicitly specified by the user.

        I have tested this on the other cairo devices (SVG and PS), and
        also reproduce the same behaviour, indicating that this is an
        issue with the cairo graphics API, or its implementation within R.

        This behaviour does not occur when using the standard R 'pdf'
        graphics device. I would switch to 'pdf' in general, however,
        'cairo_pdf' has several advantages over 'pdf', notably, reduced
        output file sizes on occasion and support for a larger array of
        UTF-8 characters, so ideally I would prefer to use cairo_pdf.

        I should note that I have also posted this message on
        StackOverflow at this web link:
        https://stackoverflow.com/questions/54892809/r-cairo-pdf-function-does-not-respect-plotting-boundaries
        <https://stackoverflow.com/questions/54892809/r-cairo-pdf-function-does-not-respect-plotting-boundaries>

        Thank you in advance for any insights into this issue.

        Sincerely,
        Lee Kelvin



        --
        Dr Lee Kelvin
        Department of Physics
        UC Davis
        One Shields Avenue
        Davis, CA 95616
        USA

        Ph: +1 (530) 752-1500
        Fax: +1 (530) 752-4717


         ? ? ? ? [[alternative HTML version deleted]]