Add qhull_warnings() function

[Bisaloo]

Fix #38.

To do:

• Add documentation for qhull_warnings()
• Add tests

[codecov-commenter]

Codecov Report

Merging #42 (c5c5b0c) into main (638a469) will decrease coverage by 4.49%.
The diff coverage is 38.88%.

@@            Coverage Diff             @@
##             main      #42      +/-   ##
==========================================
- Coverage   97.26%   92.76%   -4.50%     
==========================================
  Files          12       13       +1     
  Lines         219      235      +16     
==========================================
+ Hits          213      218       +5     
- Misses          6       17      +11     

Impacted Files	Coverage Δ
R/utils.R	0.00% <0.00%> (ø)
R/zzz.R	0.00% <ø> (ø)
R/fd_fric.R	81.81% <53.84%> (-18.19%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 638a469...c5c5b0c. Read the comment docs.

[Rekyt]

Here is a list of conditions where we expect to have a qhull warning:

1. When there are not enough points (n. point < n. dimensions)
2. When there are too many dimensions (n. dimensions > 16)
3. When they are duplicate points (qhull is expected to throw an error in this case see https://github.com/cran/geometry/blob/master/tests/testthat/test-convhulln.R#L83-L85)
4. When the volume has a specific configuration making the convex hull not computable (we need a dataset to reproduce this!)

Within fd_fric() and fd_chull() we're accounting for 2. and 3.

[Bisaloo]

We are handling your first point as well in fd_chull():

https://github.com/Bisaloo/fundiversity/blob/863dc7303fec667060f8a7f051611cf07e117f74/R/fd_chull.R#L16-L23

[Bisaloo]

What about putting this issue on hold until someone comes to us with a problematic dataset?

[Rekyt]

What about putting this issue on hold until someone comes to us with a problematic dataset?

Probably the best to do unless we get a clear example. We can revisit anytime we encounter an issue with a dataset.

[Bisaloo]

I'm trying to revisit this issue and I'm not convinced this is the right approach since we would have to duplicate code in fd_fric(), fd_fric_intersect(), fdiv().

Ideally, this should be in fd_chull(). At the same, it seems somewhat inevitable that some part stays in fd_fric() unless we want to print "There are warnings()" all the time.

I need to take more time to investigate how base R manages to deal with this.