Exclusions read from .lintr are depedent of the current working directory

[joaopmatias]

Hi, I recently started using lintr and it's pretty great. Thank you to everyone involved!

I bumped into an issue related with the exclusions extracted from .lintr files. It seems that the lint report depends on the current working directory of the R console.

Running the following script while the working directory was in the root of the project (lintr/)

> lint_package(".")

does not show inst/example/bad.R in the report (since it's path is in .lintr). Then, if I run

> setwd("..")
> lint_package("lintr")

the file inst/example/bad.R shows in the report.

There is a similar issue with lint(). I added a file lintr/inst/example/.lintr with the content

exclusions: list("bad.R")

Then, I ran the commands

> setwd("lintr/inst/example")
> lint("bad.R")

and bad.R did not show up in the report (as expected). However, that was not the case when I ran

> setwd("..")
> lint("example/bad.R")

Furthermore, when run directly from the console, it seems that lint_dir() always ignores the exclusions in .lintr.

Can you clarify if any of these contrasting behaviors are expected, or what the expected behaviors should be?

Thank you
João

[russHyde]

Could you post an example .lintr file, please

[joaopmatias]

I'm using the code from master in the commands shown above, so the .lintr file is

linters: with_defaults(line_length_linter(120))
exclusions: list("inst/doc/creating_linters.R" = 1, "inst/example/bad.R", "tests/testthat/exclusions-test")

[russHyde]

Just seen this as well.
Seems to relate to read_settings

# in /path/to/lintr
# mimicking the code in `lint_package()`
path <- lintr:::find_package(".")
lintr:::read_settings(path)
lintr:::settings$exclusions
$`<HOME>/progs/pull_requests/lintr/inst/example/bad.R`
[1] Inf

$`<HOME>/progs/pull_requests/lintr/tests/testthat/exclusions-test`
[1] Inf

# In parent of lintr directory (<HOME>/progs/pull_requests/)
path <- lintr:::find_package("lintr")
lintr:::read_settings(path)
lintr:::settings$exclusions
named list()

[russHyde]

All calls to read_settings come from lint.R or cache.R; when called from lint.R either a single filename (from lint) or a directory path (from lint_dir or lint_package) is passed in. So we should be able to modify read_settings, appending the (assumed relative) filepaths from exclusions onto the dir-path whenever the input to read_settings is a directory

[joaopmatias]

I noticed that in both read_settings and the functions in lint.R, the exclusions always pass by the function normalize_exclusions.
This function gives the absolute paths for a list of files paths and removes non-existing files. I think that the issue is that it searches for the files paths relative to the working directory and not relatively to the path given in the input of a lint function.

I think that the issue won't be solved just by changing read_settings but it's a good point to start.