Request: Consider making phyloseq an optional dependency

[funnell]

Hello!

Since phyloseq is a bit of a heavy dependency, and isn't really required for running FLORAL, and probably won't be used by many users of FLORAL, installation could be made simpler and faster by not requiring phyloseq for installation. This would require checking for phyloseq before running any of the phyloseq related code.

[tengfei-emory]

In a previous update we allowed users to convert a phyloseq object to FLORAL-compatible input data. That function (see R/utils.R) explicitly imported phyloseq.

Is there a way to keep that function while not list phyloseq as a dependency? Copying @nickp60.

[nickp60]

I've used the pattern described here successfully before; then phyloseq could be dropped from the DESCRIPTION

[tengfei-emory]

I think we can try it. Hopefully CRAN will take it without compliants.

[tengfei-emory]

It didn't quite work out due to examples and test code for the util function. I will need to find additional ways to get around the check.

[funnell]

@tengfei-emory I think the idea would be to put phyloseq into suggests, and then you can guard the example code to avoid errors:

• https://r-pkgs.org/dependencies-in-practice.html#sec-dependencies-in-suggests-in-tests
• https://r-pkgs.org/dependencies-in-practice.html#sec-dependencies-in-suggests-in-examples-and-vignettes

[tengfei-emory]

This is not yet addressed by the latest pull request. Sorry for the confusion.