[R-pkg-devel] CRAN submission struggle

[Ivan Krylov]

On Thu, 28 Dec 2023 18:00:37 +0200
Christiaan Pieterse <pietie.cjp.1908 using gmail.com> wrote:

> I only get 3 notes (see below), and if I run it in PositCloud, it
> crashes or yields the same 1 ERROR and 2 NOTES result as before. Why
> might this be? 

Does the PositCloud check crash with "Killed" (most likely out of RAM)
or with a different error message?

> Is it a problem or is it fine if I continue working in RStudio since
> I cannot increase the specs in PositCloud because I'm working on a
> research group account?

If your local R CMD check works, it should be fine. Rough
specifications for the machines running CRAN checks can be found at
<https://cran.r-project.org/web/checks/check_flavors.html>. We ought to
test our packages on the weakest hardware that could plausibly be used
to run our code, but that's not always easy to do. I know I don't
always dig out my old Intel Atom ultraportable to run the checks myself.

> The second is the runtime that is too long:
> * checking examples ... [43s] NOTE
> Examples with CPU (user + system) or elapsed time > 5s
>       user system elapsed
> IOPS 10.06   3.35   35.04

Similar NOTEs can be seen about the use of multi-threading, but here
the "elapsed" (real, as measured by a clock) time exceeds the "user"
(CPU time spent inside applications) + "system" (CPU time spent inside
the operating system kernel) time, so the code uses less than 100% of
one CPU core on average, which fits comfortably in the 200% allowed by
the CRAN policy for examples and tests.

Unfortunately, 35 seconds is still too much.

> How can I reduce this time? I'm not sure how to reduce the size of my
> ExampleTradeData without the check giving errors when running the
> example.

How does the algorithm work? I've seen it fail due to Proximities being
a 0x0 matrix. Can you work backwards from
economiccomplexity::proximity() returning a 0x0 matrix to derive the
requirements that IOPS places on the dataset? It may help to experiment
with sample.int() to subset the rows and see which combinations work.
Perhaps you can reduce the dataset to two countries and two products?

Have you tried profiling? You can profile your code for both speed and
memory use, and replacing less performant idioms with those using less
CPU time and memory may solve both the CPU time problem and the
OOM-crash problem:
https://cran.r-project.org/doc/manuals/R-exts.html#Tidying-and-profiling-R-code
 
> The third note I am unsure what it means:
> * checking for detritus in the temp directory ... NOTE
> Found the following files/directories:
>   'lastMiKTeXException'

Have you installed the inconsolata MiKTeX package?
https://cran.r-project.org/doc/manuals/R-admin.html#LaTeX-on-Windows
Try running R CMD Rd2pdf on your package directory: maybe MiKTeX will
pop up an interactive dialog to let you install any remaining missing
dependencies. If not, there should be a lastMiKTeXException file for
you to read.

-- 
Best regards,
Ivan