[R-pkg-devel] ‘Package required and available but unsuitable version’ and notes about examples execution time

[Duncan Murdoch]

On 21/03/2021 7:18 a.m., Gianmarco Alberti wrote:
> Hello,
> 
> I am in the process of testing a new version of my GmAMisc package before submitting it to CRAN.
> 
> I have checked it using:
> (a) devtools::check_win_devel()
> https://win-builder.r-project.org/T9yi5bawc203
> 
> (b) devtools::check_win_release()
> https://win-builder.r-project.org/VACDopQ9QBMi
> 
> (c) devtools::check_win_oldrelease()
> https://win-builder.r-project.org/yoIK3iE7IZqk
> 
> Now,
> (a) returns 2 notes about functions’ examples whose execution time is larger than 10 sec
> This didn’t happen when testing earlier versions of the same package. I did not make any major change to those functions, excluding updating some chunks of code to adapt them to the new structure of the ‘spatstat' package. Most of my functions whose execution time is larger than 10 sec are feature Monte Carlo-based resampling; I have tried to reduce the number of randomised iterations (e.g., from 199 to 99) but no significant changes were produced when it comes to the execution time.

Did the examples actually become slower?  You can time them locally: 
you won't get the same timings as Win-builder, but relative timings 
should be roughly proportional.

If they did become slower, then you really should modify the examples. 
This might mean using an unrealistically small number of iterations; 
that's better than not running them at all, because at least it will 
show users how your functions work.

Or maybe Win-builder was under heavy load, and that's why the timings 
are slower.  You could take a chance and submit with a note about this.

> 
> (b) returns the same two notes as (a)
> 
> (c) returns 1 error:
> Package required and available but unsuitable version: ‘spatstat.linnet'

How are you stating your dependency on this package?  Your current 
version only mentions "spatstat (>= 1.56-0)".  CRAN has spatstat.linnet 
2.0-0, but perhaps Win-builder hadn't updated its old release library to 
that version when you ran your test.

Duncan Murdoch

> 
> I am wondering:
> (1) are the 2 notes returned by (a) and (b) a relevant obstacle when it comes to getting my new version on CRAN? I am a little concerned because I do not know how to reduce the execution time further.
> 
> (2) why the difference between (a)-(b) and (c) when it comes to the error related to spatstat.linnet? Is the error returned by (c) fatal in anticipation of submitting the package to CRAN? If it is, how can I tackle that?
> 
> 
> Thank you in advance for any help on the above issues.