lim_args was updating xlim and ylim even if specified

[mclements]

Nice package -- great to see base R graphics getting some attention.

This is a very minor bug fix. The following code shows that xlim and ylim were being ignored when ymin/etc are specified:

library(tinyplot)
x=y=0:1
conf.low=y-0.5
conf.high=y+0.5
par(mfrow=1:2)
plt(y~x,ymin=conf.low,ymax=conf.high,type="ribbon")
plt(y~x,ymin=conf.low,ymax=conf.high,type="ribbon",xlim=c(0,0.5),ylim=c(0,0.5))

The code changes lim_args() to check whether xlim or ylim are NULL.

[vincentarelbundock]

Thanks for taking a look at this!

Looks like it breaks some plots. In particular, it cuts the confidence intervals.

To run the test suite and check this yourself, you can comment out the "Linux" line in inst/tinysnapshot/helpers.R. Then,

pkgload::load_all()
tinytest::run_test_dir()

[mclements]

Interesting - I want to cut the confidence intervals, particularly when the intervals are very wide. Semantically, I suggest that xlim and ylim should be followed when specified (as per base::plot()). Do you disagree?

[vincentarelbundock]

Yes, I agree that if the user explicitly sets xlim or ylim, the plot should set axes accordingly. I think your change is a necessary one.

However, I think that when ylim is not specified, the plot limits should be set to include the whole interval. This is not the case in your branch.

One of the examples I pasted above does not specify ylim, and yet cuts the axes:

https://github.com/grantmcdermott/tinyplot/blob/main/inst/tinytest/test-ribbon.R#L10

[mclements]

I have further simplified the calculations for xlim and ylim.

I am getting errors for test-legend-gradient.R (6/7) and test-rect.R (1/2) due to different versions of libpng -- which may be due to how the test cases were generated.

Can you re-run the tests, please?

[vincentarelbundock]

Thanks a lot for this @mclements ! It's a great and useful bug fix.

I updated a few snapshots (no visual difference at all) and everything passes. Merging now.

Cheers,

[grantmcdermott]

Thanks for the submission @mclements and for handling @vincentarelbundock!

I'll have a bit more breathing room towards the end of the week once I get some of these work deliverables over the line.