11 Passing Checks
Our package is really starting to grow up on us!
But it’s time to test it - we do this with devtools::check()
This function runs all the general tests to see if our package can go onto CRAN. There are usually a few more checks to run, but it’s mostly 95% of the way there.
Essentially, devtools::check() does the following:
- updates documentation
- loads the package
- builds the package (essentially puts it into machine code)
- creates vignettes
- runs tests, if there are any
- runs examples
It does usually take a minute or so to do, so it’s not something we run all the time, but it’s a good idea to get into a habit of being somewhat regular about the checks.
Let’s run devtools::check() on learned
E creating vignettes (2.9s)
--- re-building ‘analysis-2014.qmd’ using html
processing file: analysis-2014.qmdWe pretty quickly get an error!
Let’s take a full look at the vignette error:
E creating vignettes (2.9s)
--- re-building ‘analysis-2014.qmd’ using html
processing file: analysis-2014.qmd
|
| | 0%
|
|.. | 5%
|
|.... | 10% [libraries]
|
|...... | 14%
|
|........ | 19% [functions]
|
|.......... | 24%
|
|............. | 29% [read-data]
Error: '/private/var/folders/9c/k3wqmhhx4qsb3fd66n4prhlw0000gq/T/RtmpBzprFu/Rbuild397478ba5ca0/learned/raw_education_2014.csv' does not exist.
Quitting from analysis-2014.qmd:64-75 [read-data]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<error/rlang_error>
Error:
! '/private/var/folders/9c/k3wqmhhx4qsb3fd66n4prhlw0000gq/T/RtmpBzprFu/Rbuild397478ba5ca0/learned/raw_education_2014.csv' does not exist.
---
Backtrace:
▆
1. ├─readr::read_csv(here("raw_education_2014.csv"))
2. │ └─vroom::vroom(...)
3. │ └─vroom:::vroom_(...)
4. └─vroom (local) `<fn>`("/private/var/folders/9c/k3wqmhhx4qsb3fd66n4prhlw0000gq/T/RtmpBzprFu/Rbuild397478ba5ca0/learned/raw_education_2014.csv")
5. └─vroom:::check_path(path)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~OK, so the issue is that we are reading in the CSV in a non standard way. But, now that we have the data built in to the package, we don’t need to worry about this.
In fact, looking at the vignette now, there’s a few things going on!
- What do you think we need to fix in this vignette to get it to run?
- Have a go at cleaning up the functions, the data read ins, and see if you can get
devtools::check()to pass building the vignettes
We resolve this at this commit.
11.1 Check can sometimes open a can of worms
Even as an experienced package developer, once I fixed the above, I still had quite a few things to fix. Let’s walk through these.
11.1.1 Introducing a new error
We also introduce a new error:
❯ checking package dependencies ... ERROR
VignetteBuilder package not declared: ‘quarto’
See section ‘The DESCRIPTION file’ in the ‘Writing R Extensions’
manual.Which I believe is fixed by adding quarto to Suggests - done in this commit.
This can also be done with:
use_package("quarto", type = "Suggests")Suggestsis a place where you put packages that are used in your vignettes, but not your package.
11.1.2 More errors, warnings, notes
Doing that, and running check() we now get a bit more of a can of worms of errors, warnings, and notes:
❯ checking re-building of vignette outputs ... ERROR
Error(s) in re-building vignettes:
--- re-building ‘analysis-2014.qmd’ using html
processing file: analysis-2014.qmd
Error in library(naniar) : there is no package called 'naniar'
Calls: .main ... withCallingHandlers -> withVisible -> eval -> eval -> library
Quitting from analysis-2014.qmd:35-39 [libraries]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<error/rlang_error>
Error in `library()`:
! there is no package called 'naniar'
---
Backtrace:
▆
1. └─base::library(naniar)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Execution halted
Error: processing vignette 'analysis-2014.qmd' failed with diagnostics:
✖ Error running quarto cli.
Caused by error:
! System command 'quarto' failed
--- failed re-building ‘analysis-2014.qmd’
SUMMARY: processing the following file failed:
‘analysis-2014.qmd’
Error: Vignette re-building failed.
Execution halted
❯ checking DESCRIPTION meta-information ... WARNING
Invalid license file pointers: LICENSE
❯ checking R code for possible problems ... NOTE
boxplot_study_state: no visible binding for global variable
‘prop_studying’
boxplot_study_state: no visible binding for global variable
‘state_territory’
clean_education_data: no visible binding for global variable
‘age_group’
clean_education_data: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘age_group’
summarise_prop_study: no visible binding for global variable
‘state_territory’
summarise_prop_study: no visible binding for global variable
‘prop_studying’
summarise_prop_study: no visible global function definition for
‘quantile’
summarise_prop_study: no visible global function definition for
‘median’
summarise_prop_study: no visible global function definition for ‘sd’
Undefined global functions or variables:
age_group median prop_studying quantile sd state_territory
Consider adding
importFrom("stats", "median", "quantile", "sd")
to your NAMESPACE file.
❯ checking for unstated dependencies in vignettes ... NOTE
'library' or 'require' call not declared from: ‘naniar’
1 error ✖ | 1 warning ✖ | 2 notes ✖This all looks a bit intense. It’s OK. We will get through.
Start by tackling the ERRORs.
11.1.3 Packages used in vignettes have to be declared in Suggests
This error
❯ checking re-building of vignette outputs ... ERROR
Error(s) in re-building vignettes:
--- re-building ‘analysis-2014.qmd’ using html
processing file: analysis-2014.qmd
Error in library(tidyverse) : there is no package called 'tidyverse'
Calls: .main ... withCallingHandlers -> withVisible -> eval -> eval -> library
Quitting from analysis-2014.qmd:35-41 [libraries]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<error/rlang_error>
Error in `library()`:
! there is no package called 'tidyverse'
---
Backtrace:
▆
1. └─base::library(tidyverse)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Execution halted
Error: processing vignette 'analysis-2014.qmd' failed with diagnostics:
✖ Error running quarto cli.
Caused by error:
! System command 'quarto' failed
--- failed re-building ‘analysis-2014.qmd’
SUMMARY: processing the following file failed:
‘analysis-2014.qmd’
Error: Vignette re-building failed.
Execution haltedWe actually don’t need some of these packages anymore - so let’s remove tidyverse, and here.
Then, doing that, we still get a similar error, but just for the “naniar” package.
We can resolve this by doing
use_package("naniar", type = "Suggests")This works!
That puts at at this commit
We now have this state
❯ checking DESCRIPTION meta-information ... WARNING
Invalid license file pointers: LICENSE
❯ checking R code for possible problems ... NOTE
boxplot_study_state: no visible binding for global variable
‘prop_studying’
boxplot_study_state: no visible binding for global variable
‘state_territory’
clean_education_data: no visible binding for global variable
‘age_group’
clean_education_data: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘age_group’
summarise_prop_study: no visible binding for global variable
‘state_territory’
summarise_prop_study: no visible binding for global variable
‘prop_studying’
summarise_prop_study: no visible global function definition for
‘quantile’
summarise_prop_study: no visible global function definition for
‘median’
summarise_prop_study: no visible global function definition for ‘sd’
Undefined global functions or variables:
age_group median prop_studying quantile sd state_territory
Consider adding
importFrom("stats", "median", "quantile", "sd")
to your NAMESPACE file.
0 errors ✔ | 1 warning ✖ | 1 note ✖
11.1.4 Addressing the warning
❯ checking DESCRIPTION meta-information ... WARNING
Invalid license file pointers: LICENSELet’s look at the DESCRIPTION file, and at the License section
It says:
License: MIT + file LICENSEWe didn’t initialise the License properly, so we run
use_mit_license()Which gives us
✔ Writing LICENSE.
✔ Writing LICENSE.md.
✔ Adding "^LICENSE\\.md$" to .Rbuildignore.checkpoint this commit
11.1.5 The NOTES
Running Check again, we are down to the NOTEs:
❯ checking R code for possible problems ... NOTE
boxplot_study_state: no visible binding for global variable
‘prop_studying’
boxplot_study_state: no visible binding for global variable
‘state_territory’
clean_education_data: no visible binding for global variable
‘age_group’
clean_education_data: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘prop_studying’
plot_study_age_state: no visible binding for global variable
‘age_group’
summarise_prop_study: no visible binding for global variable
‘state_territory’
summarise_prop_study: no visible binding for global variable
‘prop_studying’
summarise_prop_study: no visible global function definition for
‘quantile’
summarise_prop_study: no visible global function definition for
‘median’
summarise_prop_study: no visible global function definition for ‘sd’
Undefined global functions or variables:
age_group median prop_studying quantile sd state_territory
Consider adding
importFrom("stats", "median", "quantile", "sd")
to your NAMESPACE file.
OK, so there’s two things here:
- we forgot to namespace
median,quantile, andsd - We need to deal with “Undefined global functions or variables: age_group median prop_studying quantile sd state_territory”
The latter one is more painful, let’s deal with this.
Essentially, R is complaining because we use these variables inside dplyr code, and it doesn’t know where they come from. My preferred way of dealing with this is using a function called globalVariables. I usually do this inside a centralised R function in what is called a “package doc file”.
We can create this file with:
use_package_doc()Which tells us:
✔ Writing R/learned-package.R.
☐ Modify R/learned-package.R.
☐ Run devtools::document() to update package-level
documentation.And then inside that file I put age_group median prop_studying quantile sd state_territory. It ends up looking like this:
#' @keywords internal
"_PACKAGE"
## usethis namespace: start
## usethis namespace: end
NULL
globalVariables(c(
"age_group",
"median",
"prop_studying",
"quantile",
"sd",
"state_territory"
))So now, let’s run document(), and then run check() again.
And now we are clear! This actually solved the other problem…which is a bit mysterious, but that’s sometimes how it goes.
Here is the commit of this.