Any `# nolint: ` statements stops all linting on line

[RasmusSkytte]

I have noticed what I expect is unintended behaviour.

Consider the example below

Starting with an empty folder, we include create file "test.R") with the following content:

N <- 2

If we run lintr::lint_dir(), we get the following lints:

Line 1 [object_name_linter] Variable and function name style should match snake_case or symbols

If we then alter the file to read:

N <- 2  # nolint: some_linter

And then we then run lintr::lint_dir(), we now get no lints at all despite not supressing the object_name_linter.

Looking into the issue, it seems to stem form the default setting for "exclude_linter"
which takes the value "^[[:space:]]*:[[:space:]]*(?<linters>(?:(?:[^,.])+[[:space:]]*,[[:space:]]*)*(?:[^,.])+)\."

If I now add a .lintr.R file to my project with the following content:

exclude_linter <- r"{"^[[:space:]]*:[[:space:]]*((?:(?:[^,.])+[[:space:]]*,[[:space:]]*)*(?:[^,.])+)\.?"}"

And run the linter again, we correctly get lints for our altered file

Line 1 [object_name_linter] Variable and function name style should match snake_case or symbols

This regex expression is different in two ways from the default:

1. It ends with \.? instead of \.
2. The naming of the capture group was dropped.
   The latter change is only because it will not validate with the naming present.

This second change is related to a separate, minor issue with lintr:::is_valid_regex:

lintr:::is_valid_regex(lintr::default_settings$exclude_linter) # FALSE

[RasmusSkytte]

Also, here is the output for my current sessionInfo()

R version 4.3.1 (2023-06-16 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19045)

Matrix products: default


locale:
[1] LC_COLLATE=English_Denmark.1252  LC_CTYPE=English_Denmark.1252   
[3] LC_MONETARY=English_Denmark.1252 LC_NUMERIC=C                    
[5] LC_TIME=English_Denmark.1252    

time zone: Europe/Copenhagen
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
 [1] crayon_1.5.2       vctrs_0.6.4        cli_3.6.1          rlang_1.1.2        rex_1.2.1         
 [6] processx_3.8.2     xmlparsedata_1.0.5 data.table_1.14.8  glue_1.6.2         backports_1.4.1   
[11] rprojroot_2.0.4    ps_1.7.5           hms_1.1.3          fansi_1.0.5        tibble_3.2.1      
[16] tzdb_0.4.0         lifecycle_1.0.4    cyclocomp_1.1.1    compiler_4.3.1     lintr_3.1.1       
[21] pkgconfig_2.0.3    rstudioapi_0.15.0  R6_2.5.1           readr_2.1.4        utf8_1.2.4        
[26] pillar_1.9.0       callr_3.7.3        magrittr_2.0.3     tools_4.3.1        withr_2.5.2       
[31] lazyeval_0.2.2     xml2_1.3.5         remotes_2.4.2.1    desc_1.4.2        

[MichaelChirico]

This is WAI. You need the . at the end, otherwise it just gets matched to # nolint. Two ways to get the object_name_linter() hit to surface, hopefully this elucidates what's going on.

1. Include the .. lint(text = "N <- 2 # nolint: some_linter.", parse_settings=FALSE)
2. Change the # nolint marker, e.g. withr::with_options(c(lintr.exclude = "xxx"), lint(text = "N <- 2 # nolint: some_linter"))

Closing, but if you think there's some improvement we could add to the documentation, let us know.

[MichaelChirico]

a separate, minor issue with lintr:::is_valid_regex:

lintr:::is_valid_regex(lintr::default_settings$exclude_linter) # FALSE

Hmm, that's just because is_valid_regex() uses perl = FALSE, it works with perl = TRUE:

!inherits(tryCatch(grepl(lintr::default_settings$exclude_linter, "", perl = TRUE), condition = identity), "condition")
# [1] TRUE

#2375.