Packages in Imports but not in NAMESPACE

[almeidasilvaf]

I often review packages that list one or many packages in Imports, but not in the NAMESPACE. In extreme cases, >90% of the packages listed in Imports are not the NAMESPACE.

IMO, these packages should fail with WARNINGs or ERRORs. Could we add a check for that? Essentially, BiocCheck would have to check if each entry in Imports has at least one match in the NAMESPACE file.

Best,
Fabricio

[hpages]

AFAICT R CMD check already reports this as a NOTE:

...
 checking dependencies in R code ... NOTE
Namespaces in Imports field not imported from:
  ‘SummarizedExperiment’ ‘stats4’
  All declared Imports should be used.
* checking S3 generic/method consistency ... OK
...

Declaring dependencies that are not used seems serious enough to diserve a WARNING so I wonder why R core decided to make this a NOTE only. Maybe there's a good reason for that so it would be good to investigate.

In any case maybe it would be better to try and get this in R CMD check rather than in BiocCheck since it's more like a general R package thing than a specific Bioconductor package thing. That would also avoid duplicating the effort that has already been put in R CMD check to detect this situation.