如何创建嵌套在 git 目录中的 R 包

Question

I used to be able to build my R packages within my github repo, but the devtools functions have been deprecated and replaced with new functions that won't allow me to do this anymore.

I was previously using devtools::setup('NewPackage', check=FALSE) in my directory /home/User/NewPackage/ , to create my R package /home/User/NewPackage/NewPackage

The new alternative gives the following error

usethis::create_package('NewPackage')
#New project 'NewPackage' is nested inside an existing project '/home/User/NewPackage/'
#This is rarely a good idea. Do you wish to create anyway?
#1: Negative
#2: I agree
#3: No way

I had automated my package creation, but this new function breaks the code in all of my packages.

I have posted this as a bug https://github.com/r-lib/usethis/issues/553

But I am still looking for a work around or any way to automatically submit I agree ?

Hidden Files

There are also hidden files, such as .Rd2pdf12059 , in my /home/User/NewPackage/ directory

Sloppy Fix

Until this is fixed, I have just copied the old devtools source code and removed what I didn't like.

Answer 1

Not the cleanest way of doing things, but you could use testthat::with_mock to trick check_not_nested :

testthat::with_mock(
  check_not_nested = function(path, name) return(),
  usethis::create_package('NewPackage2'),
  .env = "usethis"
)

Answer 2

As the testthat::with_mock function is now deprecated, and the functions from mockr or mockery cannot replace functions in other packages, I have found the following to work.

Note that check_not_nested is now named challenge_nested_project in the usethis package (version 2.0.1)

  challenge_nested_project <- function(path, name) return()
  
  rlang::env_unlock(env = asNamespace('usethis'))
  rlang::env_binding_unlock(env = asNamespace('usethis'))
  assign('challenge_nested_project', challenge_nested_project, envir = asNamespace('usethis'))
  rlang::env_binding_lock(env = asNamespace('usethis'))
  rlang::env_lock(asNamespace('usethis'))

This code does not temporarily replace the function until the package is unloaded and loaded again, and the rlang functions are listed as experimental, so handle with care and results might change in the future.