`use_template()` no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

boshek · 2018-05-14T17:24:10Z

Closes #348 Allow user to decline to overwrite existing file, from use_template()

Modified write_over so that it produces a message rather than an error when can_overwrite() is FALSE
Revised relevant tests to test for a message rather than an error
Added a single entry in the NEWS file outlining this small change (possibly not big enough to merit a NEWS item?)

…tead of an error. Closes r-lib#348

boshek · 2018-05-14T18:31:52Z

R CMD check passing locally.

Appveyor error:

Error: package or namespace load failed for 'usethis':
.onLoad failed in loadNamespace() for 'usethis', details:
call: NULL
error: package 'backports' was installed by an R version with different internals; it needs to be reinstalled for use with this R version
Error: loading failed
Execution halted

possibly an old version of backports that hasn't updated yet?

travis error:
For some reason this is behaving correctly as the test is on R 3.1 that is failing:

usethis/tests/testthat/test-use-readme.R

Lines 14 to 15 in bb191e9

    
           # git2r::git2r::discover_repository() not working on R 3.1 (Travis) 
        
           skip_if(getRversion() < 3.2)

Possibly clearing caches on both would help?

jennybc · 2018-05-14T21:13:28Z

It looks like I'll be doing a small release of usethis soon (e.g. within the month), prompted by the backports update, so I will sort out these things once I dig in.

jennybc

Just one small change. Otherwise, looks good.

Next time you probably be better off to make the PR from a non-master branch in your fork.

It will be interesting to see if the CI situation has improved, when you make a new commit. Various things that have nothing to do with your PR have (hopefully) been sorted out. I pass tests fine locally with this.

jennybc · 2018-05-25T20:19:48Z

R/helpers.R

@@ -45,7 +45,7 @@ use_template <- function(template,
    use_build_ignore(save_as)
  }

-  if (open) {
+  if (open & new) {


This should be &&.

jennybc · 2018-05-25T20:42:45Z

Upon further thought, the failure on Travis re: use_readme_rmd() is real and needs to be dealt with.

Previously, we would have stopped inside use_template() if we can't overwrite an existing README.Rmd. But now we just message and go on to execute the next block, which calls uses_git() to decide whether to add a hook. And the failure on R 3.1 is alerting to this new behaviour.

In general, you need to do a survey of all existing uses of use_template() and see if subsequent code assumes that a new template file was created. How to deal with this? It depends on how many functions are affected. Can you characterize the scope of this problem?

boshek · 2018-05-28T05:18:22Z

Next time you probably be better off to make the PR from a non-master branch in your fork

Thanks for the tip.

Happy to characterize the scope of this. Hope to have some time near the end of this week to spend on this. Will that timing work?

jennybc · 2018-05-28T15:42:55Z

Happy to characterize the scope of this. Hope to have some time near the end of this week to spend on this. Will that timing work?

No worries, I might get to it sooner or not. But we have to sort it out before we can merge this. As you might see, I now think it's connected to another issue #178.

boshek · 2018-05-28T16:24:30Z

Here is a bare bones visualization that I'll leave here as a reference for myself (or whomever ultimately tackles this):

ateucher · 2018-05-30T19:56:12Z

I hope you don't mind me piping in here... With this PR, use_template() (via write_over()) returns FALSE with a message now instead of failing if can_overwrite is FALSE … but that behaviour existed already when the old file and the new file were identical. So is this just exposing something that was previously untested?

So the question is (specifically for use_readme_rmd()) would you want the hook to be added if the README.Rmd already existed and wasn't overwritten? The existing behaviour would add the hook if the file existed already but was identical to the template. Is the new case really any different?

My apologies if I'm off base or missed something obvious!

ateucher · 2018-05-30T22:58:51Z

This is some heavy yak-shaving, and possibly not very helpful, but here are the functions that call use_template() which have meaningful code that comes after the call to use_template().

Most of them are just todo()/code_block() messages, except for:

use_travis() and use_appveyor() call their respective *_activate() and use_*_badge() functions.
use_readme_rmd() we know about
use_rstudio() calls use_gitignore(), and use_build_ignore() if it's a package.
use_tidy_ci() and use_rmarkdown_template() call use_template() multiple times.

library(tidyverse)
library(igraph, quietly = TRUE, warn.conflicts = FALSE)
library(usethis)
# devtools::install_github("ropenscilabs/revtools")

find_code_after_use_template <- function(x) {
  x <- getFromNamespace(x, "usethis")
  fn_body <- body(x)
# find where use_template is called
  ut_block_no <- grep("use_template", fn_body)
  if (length(ut_block_no) > 1) {
    return("More than one call to use_template. Check this one out more carefully")
  }
# if call to use_template is the last line, or the line after the call is a return-type statement
  if (ut_block_no == length(fn_body) || grepl("^return|^invisible\\(", fn_body[ut_block_no + 1])) {
    return(character(0))
  }
# return the rest of the function body
  as.character(fn_body[seq(ut_block_no + 1, length(fn_body))])
}

# Get character vector of exported function names that call use_template
funs_that_call_use_template <- revtools:::create_package_igraph("~/dev/packages/usethis/") %>% 
  as_long_data_frame() %>% 
  filter(to_name == "use_template") %>% 
  pull(from_name)

map(funs_that_call_use_template, find_code_after_use_template) %>% 
  set_names(funs_that_call_use_template) %>% 
  Filter(length, .)
#> $use_travis
#> [1] "travis_activate(browse)" "use_travis_badge()"     
#> [3] "invisible(TRUE)"        
#> 
#> $use_appveyor
#> [1] "appveyor_activate(browse)" "use_appveyor_badge()"     
#> [3] "invisible(TRUE)"          
#> 
#> $use_code_of_conduct
#> [1] "todo(\"Don't forget to describe the code of conduct in your README.md:\")"                                                                                                                 
#> [2] "code_block(\"Please note that this project is released with a [Contributor Code of Conduct](CODE_OF_CONDUCT.md).\", \"By participating in this project you agree to abide by its terms.\")"
#> 
#> $use_pipe
#> [1] "todo(\"Run \", code(\"document()\"))"
#> 
#> $use_readme_rmd
#> [1] "if (uses_git()) {\n    use_git_hook(\"pre-commit\", render_template(\"readme-rmd-pre-commit.sh\"))\n}"
#> [2] "invisible(TRUE)"                                                                                      
#> 
#> $use_revdep
#> [1] "todo(\"Run checks with \", code(\"revdepcheck::revdep_check(num_workers = 4)\"))"
#> 
#> $use_rmarkdown_template
#> [1] "More than one call to use_template. Check this one out more carefully"
#> 
#> $use_rstudio
#> [1] "use_git_ignore(\".Rproj.user\")"                                                                     
#> [2] "if (is_package()) {\n    use_build_ignore(c(paste0(project_name(), \".Rproj\"), \".Rproj.user\"))\n}"
#> [3] "invisible(TRUE)"                                                                                     
#> 
#> $use_tidy_ci
#> [1] "More than one call to use_template. Check this one out more carefully"
#> 
#> $use_tidy_eval
#> [1] "todo(\"Run document()\")"
#> 
#> $use_tidy_coc
#> [1] "todo(\"Don't forget to describe the code of conduct in your README.md:\")"                                                                                                                         
#> [2] "code_block(\"Please note that this project is released with a [Contributor Code of Conduct](.github/CODE_OF_CONDUCT.md).\", \"By participating in this project you agree to abide by its terms.\")"

Created on 2018-05-30 by the reprex package (v0.2.0).

boshek · 2018-06-01T21:52:39Z

Using @ateucher's handy scoping:

use_travis and use_appveyor

A user with an existing *.yml file here can safely choose not to overwrite because neither the active and badge functions depend on use_template failing when an overwrite does not occur

use_readme_rmd

I think the question here identified by @ateucher is:

would you want the hook to be added if the README.Rmd already existed and wasn't overwritten
I think you would right?

I have skipped the readme_rmd test if R is less than 3.1 because it was trying to use git2r::::discover_repository() in the test now which fails on travis. Hopefully this fixes the build fail on 3.1.

use_rstudio

The main impact I see here is that calling create_project() or create_package() from a directory that has an existing .Rproj file will now ask you if you want to overwrite it.

This seems a little redundant so one could only call use_template() if an .Rproj is absent. Would this be desirable?

use_rmarkdown_template & use_tidy_ci.

This functions without any adverse downstream effects and actually provides greater control if one wants to replace the .yaml but not the rmarkdown template (or vice versa).

boshek · 2018-06-29T21:52:12Z

@jennybc Curious is it generally considered my responsibility to fix the merge conflicts that have arisen since the PR was originally submitted or is that usually left to the maintainer (you in this case)?

jennybc · 2018-06-29T23:14:20Z

You can leave it with me @boshek.

Also a test to see if I can push to boshek/master

jennybc · 2018-08-09T20:53:49Z

@boshek I'm working on this now and will try to finish things up in your PR. This means I'm going to be pushing to the branch you made it from (and have already done so). In future, you probably want to work in a non-master branch for PRs, but we'll just carry on with this.

Are OK to just let me take over master branch of your fork for a little while?

boshek · 2018-08-09T21:11:35Z

@jennybc Fine by me.

I could, alternatively, close everything down and re-submit a clean PR if that were the path of least resistance at this point.
Happy to further contribute as well if any tasks can be farmed out

# Conflicts: # NEWS.md # R/helpers.R # R/write.R # tests/testthat/test-use-readme.R # tests/testthat/test-use-template.R

jennybc · 2018-08-12T17:51:18Z

@boshek I think this is done. I'm just starting a vacation would like to wrap up this release before I do a complete mental check out.

Will you sanity check this again, both by taking a look at the code I pushed and by confirming it still behaves the way you want? I'll be working through my other pre-release checklist items in parallel.

Thanks!

jennybc · 2018-08-12T17:51:40Z

@ateucher Review by you, instead or in addition, is also welcome!

boshek · 2018-08-13T03:24:23Z

@jennybc Looks great to me on both counts. Nothing odourous as far as I can see and the behaviour is exactly what I was after.

Note: It isn't totally clear to me how this, this and this commit aren't in the boshek-master branch but are still part of the PR here. I assume this reflect my lack of understanding but am flagging this just in case.

jennybc · 2018-08-13T06:31:45Z

Thanks for taking a look!

Note: It isn't totally clear to me how this, this and this commit aren't in the boshek-master branch but are still part of the PR here. I assume this reflect my lack of understanding but am flagging this just in case.

That's a branch in this repo that I created by mistake, in some sense. I will delete it. The branch that matters for the PR is the master branch in your fork, which does have these commits.

a negative user response to overwriting a file produces a message ins…

75a4f41

…tead of an error. Closes r-lib#348

ateucher mentioned this pull request May 14, 2018

Tidy up various tryCatch statements when upstream fixes get implemented bcgov/bcgovr#39

Open

2 tasks

boshek added 2 commits May 14, 2018 10:55

only modify if open = TRUE and the file is actually created

5c7ae8d

modify test_that names to reflect warnings

2f53744

jennybc reviewed May 25, 2018

View reviewed changes

jennybc mentioned this pull request May 25, 2018

Consider "skip" option when overwriting files #178

Closed

use longer form comparison

a458e93

skip R 3.1 when testing message for use_readme overwrite

5716b1c

Tweak docs

ace9e5a

Also a test to see if I can push to boshek/master

jennybc added 6 commits August 9, 2018 15:42

Swap an if-else

65afc00

Merge branch 'master' into boshek-master

7fea54b

# Conflicts: # NEWS.md # R/helpers.R # R/write.R # tests/testthat/test-use-readme.R # tests/testthat/test-use-template.R

These functions have moved

1765b4f

Use done() instead of message()

94e65d3

Add expectations for file or dir in the project

8d1be01

Finish adjusting to new behaviour or write_over() and use_template()

9d47752

jennybc merged commit 3046fde into r-lib:master Aug 13, 2018

ateucher mentioned this pull request Aug 14, 2018

v1.0.1 - Fix bugs from internal usethis changes (and others) bcgov/bcgovr#46

Merged

`use_template()` no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

`use_template()` no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

boshek commented May 14, 2018 •

edited by jennybc

boshek commented May 14, 2018

jennybc commented May 14, 2018

jennybc left a comment •

edited

jennybc May 25, 2018

jennybc commented May 25, 2018 •

edited

boshek commented May 28, 2018

jennybc commented May 28, 2018

boshek commented May 28, 2018 •

edited

ateucher commented May 30, 2018 •

edited

ateucher commented May 30, 2018 •

edited

boshek commented Jun 1, 2018

boshek commented Jun 29, 2018

jennybc commented Jun 29, 2018

jennybc commented Aug 9, 2018

boshek commented Aug 9, 2018

jennybc commented Aug 12, 2018

jennybc commented Aug 12, 2018

boshek commented Aug 13, 2018

jennybc commented Aug 13, 2018

use_template() no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

use_template() no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

Conversation

boshek commented May 14, 2018 • edited by jennybc

boshek commented May 14, 2018

jennybc commented May 14, 2018

jennybc left a comment • edited

Choose a reason for hiding this comment

jennybc May 25, 2018

Choose a reason for hiding this comment

jennybc commented May 25, 2018 • edited

boshek commented May 28, 2018

jennybc commented May 28, 2018

boshek commented May 28, 2018 • edited

ateucher commented May 30, 2018 • edited

ateucher commented May 30, 2018 • edited

boshek commented Jun 1, 2018

use_travis and use_appveyor

use_readme_rmd

use_rstudio

use_rmarkdown_template & use_tidy_ci.

boshek commented Jun 29, 2018

jennybc commented Jun 29, 2018

jennybc commented Aug 9, 2018

boshek commented Aug 9, 2018

jennybc commented Aug 12, 2018

jennybc commented Aug 12, 2018

boshek commented Aug 13, 2018

jennybc commented Aug 13, 2018

`use_template()` no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

`use_template()` no longer errors when a user chooses not to overwrite an existing file and instead produces a message #350

boshek commented May 14, 2018 •

edited by jennybc

jennybc left a comment •

edited

jennybc commented May 25, 2018 •

edited

boshek commented May 28, 2018 •

edited

ateucher commented May 30, 2018 •

edited

ateucher commented May 30, 2018 •

edited