What happened?
These two tests break under the current development version of R which is in the process of being released to CRAN:
|
expected <- paste0( |
|
"Line 6 [library_call_linter] Move all library calls to the ", |
|
"top of the script.\n\nLine 8 [undesirable_operator_linter] Operator ", |
|
"`<<-` is undesirable. It\nassigns outside the current environment in a ", |
|
"way that can be hard to reason\nabout. Prefer fully-encapsulated ", |
|
"functions wherever possible, or, if\nnecessary, assign to a specific ", |
|
"environment with assign(). Recall that you\ncan create an environment ", |
|
"at the desired scope with new.env()." |
|
) |
|
"Line 3 [undesirable_operator_linter] Operator `<<-` is undesirable.", |
|
" It\nassigns outside the current environment in a way that can be hard", |
|
" to reason\nabout. Prefer fully-encapsulated functions wherever possible,", |
|
" or, if\nnecessary, assign to a specific environment with assign(). ", |
|
"Recall that you\ncan create an environment at the desired scope with", |
|
" new.env().\n\nLine 4 [undesirable_operator_linter] Operator `<<-` is", |
|
" undesirable. It\nassigns outside the current environment in a way that", |
|
" can be hard to reason\nabout. Prefer fully-encapsulated functions", |
|
" wherever possible, or, if\nnecessary, assign to a specific environment", |
|
" with assign(). Recall that you\ncan create an environment at the", |
|
" desired scope with new.env()." |
In general, I try and avoid relying on fragile things like "the exact wording of an error message produced by an upstream package".
For example, in {data.table}, we have this list of messages/message factories for test outputs that come directly from {base} R:
https://github.com/Rdatatable/data.table/blob/c29e313dac1de8b959f5e3438c33c0d00873be05/inst/tests/tests.Rraw#L176-L201
I will submit a quick-fix PR shortly, but it's worth thinking through a more robust way to structure these tests going forward.
Session Information
No response
Reproducible Example
No response
