GH-45521: [CI][Dev][R] Install required cyclocomp package to be used with R lintr

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.

[raulcd]

Rationale for this change

The linting jobs are failing due to the new version of lintr not installing cyclocomp anymore.
We use cyclocomp but this is not part of the default linters of lintr anymore. We should install it individually.

What changes are included in this PR?

Install cyclocomp as part of setting up the linting environment for R on our linting job.
Pin old version of lintr for R Windows job as it otherwise fails with a lot of new linter issues.

Are these changes tested?

Yes via CI.

Are there any user-facing changes?

No

• GitHub Issue: [CI][R] Lint failure #45521

[github-actions]

⚠️ GitHub issue #45521has been automatically assigned in GitHub to PR creator.

[raulcd]

This fixes the lintr job on Dev linting but seems to fail on the linting for Windows R release:

 Error: Error: Not lint free
R/arrow-info.R:87:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:95:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:103:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:111:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:119:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:127:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/arrow-info.R:135:5: style: [return_linter] Use implicit return behavior; explicit return() is not needed.
return(FALSE)
^~~~~~
R/dplyr-arrange

This is due to a new default linter on the new lintr version:

• New default linter return_linter()

I am unsure why this fails on this job and not the other one.

[raulcd]

return_linter seems to require many more changes as seen on other files failing. I am going to try and temporarily disable it.

[raulcd]

Once we disable return_linter there are other linters that fail:

 tests/testthat/test-dplyr-collapse.R:146:7: style: [commented_code_linter] Remove commented code.
# filter(dbl > 2) %>%
^~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:147:7: style: [commented_code_linter] Remove commented code.
# select(chr, int, lgl) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:148:7: style: [commented_code_linter] Remove commented code.
# mutate(twice = int * 2L) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:149:7: style: [commented_code_linter] Remove commented code.
# group_by(lgl) %>%
^~~~~~~~~~~~~~~~~
tests/testthat/test-dplyr-collapse.R:150:7: style: [commented_code_linter] Remove commented code.
# summarize(total = sum(int, na.rm = TRUE)) %>%
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In addition: Warning message:

I think the easier approach here is to temporarily pin lintr to the version we were using and open a new issue to unpin it and fix the new linters so we can fix CI which is broken at the moment.

[raulcd]

@jonkeane it seems we run linting differently from the Windows R release job and the Dev linting job. On the Dev linting job installing cyclocomp is enough but for Windows we start getting a bunch of linting failures with the new version as seen on the comments above. I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

[kou]

Can we remove this if we pin lintr to 3.1.2?

[raulcd]

I wasn't able to pin here to 3.1.2, CRAN says is not available:

2025-02-13T13:40:04.1633160Z #15 0.292 > install.packages('lintr@3.1.2')
2025-02-13T13:40:04.3161054Z #15 0.294 Installing package into '/usr/local/lib/R/site-library'
2025-02-13T13:40:04.3161668Z #15 0.294 (as 'lib' is unspecified)
2025-02-13T13:40:11.3252442Z #15 7.454 Warning message:
2025-02-13T13:40:11.3253097Z #15 7.454 package 'lintr@3.1.2' is not available for this version of R
2025-02-13T13:40:11.3253653Z #15 7.454 2025-02-13T13:40:11.3254046Z #15 7.454 A version of this package for your version of R might be available elsewhere,
2025-02-13T13:40:11.3254443Z #15 7.454 see the ideas at
2025-02-13T13:40:11.3254875Z #15 7.454 https://cran.r-project.org/doc/manuals/r-patched/R-admin.html#Installing-packages 2025-02-13T13:40:11.3255293Z #15 7.454 > 2025-02-13T13:40:11.3255459Z #15 7.454 > 

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[jonkeane]

I think we should do a follow up PR to fix those but we should probably merge this to fix CI as is currently broken for everyone. Let me know what do you think.

Strong agree that we should follow on later with fixes. Do we want a new issue or can we repurpose #45521 to be the follow on (and merge this PR as is)?

[raulcd]

@jonkeane I've created:

• [Dev][R] Update code to match new linters on lintr=3.2.0 #45536

[kou]

+1

[kou]

Ah, the .github/workflows/r.yml change isn't related to this change.

I thought that the pin in r.yml suppressed the lint error. But the lint error was caused by missing cyclocomp. (Found unused settings in config file (.lintr): unused_settings is just a warning. It's not an error.)

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 43d6f79.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 13 possible false positives for unstable benchmarks that are known to sometimes produce them.