Add function to create extendr package infrastructure

[clauswilke]

I have added a function use_extendr() to add an extendr scaffolding to an R package. There is currently no unit testing for this, it's a bit difficult to set up.

@Ilia-Kosenkov Could you try this on Windows?

[yutannihilation]

Do we add src/.gitignore?

*.o
*.so
*.dll
target
Cargo.lock

https://github.com/extendr/helloextendr/blob/main/src/.gitignore

[clauswilke]

I can add src/.gitignore. There's some discussion on whether Cargo.lock should or should not get version tracked, (don't track for Rust library crates, track for enduser applications) so maybe leave out of default .gitignore?

[yutannihilation]

Ah, I agree we should leave Cargo.lock, but I think taget should be ignored in any case.

[yutannihilation]

Also, these are not included in .gitignore if I create a package on RStudio. What do you mean by "default .gitignore"?

*.o
*.so
*.dll

[clauswilke]

What do you mean by "default .gitignore"?

I was looking here, since I'm copying helloextendr basically:
https://github.com/extendr/helloextendr/blob/main/src/.gitignore

[clauswilke]

.gitignore added.

[yutannihilation]

Thanks, looks good now. Let me try creating a package using this before approving.

[Ilia-Kosenkov]

@clauswilke,
If it is OK, please delay merging a little bit - I'd like to test a few things.

[andy-thomason]

Very nice. I fully intend to use this to generate some example packages
as soon as I am able.

[clauswilke]

@Ilia-Kosenkov Sure, no problem.

[Ilia-Kosenkov]

@clauswilke,
It works fine under Windows.
However, I suggest using to_toml to generate the skeleton Cargo.toml (see comments).

[Ilia-Kosenkov]

@clauswilke,
I was thinking about adding a test CI step that generates a dummy package, uses use_extendr() and then runs rcmdcheck on a simple test that calls hello_world(). If everything works as expected, the R CMD check will be successful.
Would you consider adding it here or should we discuss it in a separate issue/pr?

[clauswilke]

@Ilia-Kosenkov I think it's easier if I merge this and then you make a separate PR for the test.

[Ilia-Kosenkov]

Sure, go ahead.