rformassspectrometry / Spectra

Low level infrastructure to handle MS spectra
https://rformassspectrometry.github.io/Spectra/
38 stars 25 forks source link

Fix is.vector() #318

Open lgatto opened 8 months ago

lgatto commented 8 months ago

This line uses is.vector() to check if a variable is an atomic vector, not only a vector (including a list):

> is.vector(1:5)
[1] TRUE
> is.vector(list())
[1] TRUE
> is.vector(1:5, "numeric")
[1] TRUE
> is.vector(list(), "list")
[1] TRUE

The solution here is the also specify a mode:

> is.vector(1:5, "numeric")
[1] TRUE
> is.vector(list(), "list")
[1] TRUE
> is.vector(1:5, "list")
[1] FALSE

I would suggest to use

in such cases.

For that specific code chunk, there's no final else. We could probably test for

  1. is.list(.)
  2. inherits(., "List")
  3. assume the rest are atomic vectors.

@jorainer @sgibb - what do you think?

jorainer commented 8 months ago

Is there a problem with the original code or why do you suggest to change this?

lgatto commented 8 months ago

I have not observed an issue with the original code.

I just wanted to highlight that is.vector() doesn't test if we have an atomic vector. It doesn't matter here, given that we check if it's a list first, so it implicitly test for an atomic vector.

And while looking at the code, I thought that it could be re-written with a final else. But it doesn't have to be changed; I don't think there's a but - at least in the anticipated use-cases :-)

I thought that additional pairs of (critical and expert) eyes should look at it.

jorainer commented 8 months ago

I'm OK with the original code - it looks clean/clear and easy to understand. But I'm also OK with changing it is you think it is necessary? Generally I would prefer is.list and similar instead of their negation !is.list - but that's maybe just my poor brain that has sometimes problems thinking around negations, double negations etc.