Recent Breaking Changes in s.label?

[neellab-umd]

I have been working with the MEM vignette that includes several plots made with s.label. All of a sudden those plots are throwing errors. s.label was working as expected as of yesterday (12/19/2023) before I updated to the new version of RStudio ('Ocean Storm" SHA-256: 96A34972 Version: 2023.12.0+369 Released: 2023-12-15). I am using
R version 4.3.2 (2023-10-31 ucrt) -- "Eye Holes" on Windows: x86_64-w64-mingw32/x64 (64-bit)

I get the issues using code from your s.label.R script in your test folder when I run R in the new RStudio.

x0 <- runif(50, -2, 2)
y0 <- runif(50, -2, 2)
z <- x0 ^ 2 + y0 ^ 2
dfxy1 <- data.frame(x0, y0)
g1 <- s.label(dfxy1, label = as.character(z < 1), paxes.draw = TRUE, axis.text = list(col = "grey"))

Error in s.label(dfxy1, label = as.character(z < 1), paxes.draw = TRUE, :
unused arguments (paxes.draw = TRUE, axis.text = list(col = "grey"))

But it runs as expected if I use R outside of RStudio. So it seems that changes in RStudio did something to break s.label.

[thioulouse]

Package adegraphics is not loaded. Trying again after using:
library(ade4)
library(adegraphics)

[neellab-umd]

I had both of those packages loaded. I checked again - twice - and still get the same error.

[thioulouse]

This is strange, I just checked with Rstudio Version 2023.12.0+369 (2023.12.0+369) and R version 4.3.2 (2023-10-31) -- "Eye Holes" and it works fine. Did you take care to load adegraphics AFTER ade4 ?
If you load ade4 after adegraphics, then adregraphics graph functions are overwritten and you get the mentioned error.

[neellab-umd]

Thanks. The script that uses s.label loads ade4 before adegraphics so I have never seen this behavior and didn't find it via google searches. I must have loaded adegraphics in another script before loading ade4 and then adegraphics again in the offending script.

Sorry to bother you! I try my best to troubleshoot before submitting issues but did not see the source of this one.

[thioulouse]

No worries !
This is explained in the help page of functions library() and require() : Both functions check and update the list of currently attached packages and do not reload a namespace which is already loaded.
So if you do successively:
library(adegraphics)
library(ade4)
library(adegraphics)
then you end up using ade4 graph functions instead of adegraphics ones, which may be confusing.
One way around this (apart from starting a new R session) is to detach() the adegraphics package and then load it again:
detach(package:adegraphics, unload = TRUE)
library(adegraphics)

[francisvolh]

Hello, sorry to comment on a closed issue, but maybe my problem does not require to open a new one.

I am aiming to work without calling library, but using the call of the package in the function such as "adespatial::dbmem()".

I was trying to work with ade4, adespatial, and adegraphics for the first time, and bumped into something that sounds like this "required" vs "library" issue of some sorts.

So basically, is there now way to "NOT" require/load libraries to work with adegraphics?

I tried the example code, without loading/requiring libraries like this,
and nothing worked:

x0 <- runif(50, -2, 2)
y0 <- runif(50, -2, 2)
z <- x0 ^ 2 + y0 ^ 2
g1 <- adegraphics::s.label(data.frame(x0, y0), label = as.character(z < 1), paxes.draw = TRUE, 
              axis.text = list(col = "grey"))

Any comments would be greatly appreciate, thanks!

[thioulouse]

Sorry, no way as far as I know. You need to load ade4 and adegraphics (and adespatial too, I think) explicitly or many functions will not work properly.

[sdray]

I can not remember exactly what happens with adegraphics and your example. The original spirit of ade4/adegraphics is to not store data in objects but keep only their names and evaluate/load them when needed (for graphics, etc). This choice was done 20/10 years ago for ade4/adegraphics and now this causes problems with new practices about managment of package environment. Changing this approach (storing names of data objects instead of data objects) would require to fully rethink and implement our packages and this would causes heavy objects (e.g., in adegraphics, data should be stored in each graphic object) and thus would require a deep evaluation of these potential changes. To be honest, I am not sure that we have enough time to work on this aspect soon.

[francisvolh]

Thanks for replying quickly! I understand now, and it makes sense if this was done 10-20 years ago. I think I noticed the same thing with adehabitatHR (same family of packages?) that the library call with the function didn't work. Time to work on long surviving packages is surely an issue, so no worries! thanks for answering my question. No sure if you would want to close the issue? Or leave it open to remember/work on this eventually. I will leave it up to you. Thanks again!