New linter: discourage the use of anonymous functions for trivial cases

[Bisaloo]

Example

Discouraged

lapply(l, function(x) mean(x)))

Recommended

lapply(l, mean)

Scope

• to limit the risk of false positive, I'm thinking to limit this to the apply() family function (+ replicate()), at least as a first step. Or do you think it wouldn't be much more difficult and riskier to do this in all contexts?

• ideally, this should detect also both tidyverse (~ mean(.x)) and base (\(x) mean(x)) lambda functions

• I'm thinking to limit it to cases with a single argument, i.e. exclude:

  lapply(l, function(x) mean(x, na.rm = TRUE))

  even though the anonymous function could be removed here as well by forwarding the ... because users might find it clearer to not forward them.

[Bisaloo]

Can you see any potential issues / pitfalls?

If you agree this would be a good addition, I can start working on a PR soon (now that I understand a little better how the package works).

[MichaelChirico]

Hmm it seems I never commented here.

We already have such a linter internally, so it'll be covered by #884

[MichaelChirico]

we already have this linter written internally for Google :)

…

On Tue, Sep 6, 2022, 9:29 AM Hugo Gruson ***@***.***> wrote: Can you see any potential issues / pitfalls? If you agree this would be a good addition, I can start working on a PR soon (now that I understand a little better how the package works). — Reply to this email directly, view it on GitHub <#1531 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB2BA5IYKDDDUBGQPGPWV5LV45WOHANCNFSM6AAAAAAQF7YDXU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>