[ghc-steering-committee] Split -Wunused-imports (recommend accept)

[Simon Peyton Jones]

I'm good with accepting this.

Simon

On Thu, 15 Jun 2023 at 16:51, Chris Dornan <chris at chrisdornan.com> wrote:

> Dear Committee,
>
> Georgi Lyubenov has a simple proposal to improve the quality of life for
> many package maintainers that I recommend that we accept.
>
> ## Split -Wunused-imports
>
> Rendered proposal: <
> https://github.com/googleson78/ghc-proposals/blob/split-unused-imports/proposals/0000-split-wunused-imports.md
> >
> Discussion of proposal: <
> https://github.com/ghc-proposals/ghc-proposals/pull/586>
>
> ### Summary
>
> (Straight from the proposal.)
>
> We propose to split the -Wunused-imports warning into two parts:
>
> - warn on actual unused imports
> - warn on duplicate imports
>
> while also not including the second in -Wall. This will greatly reduce
> "churn" for library authors.
>
>
> ## In more Detail
>
> As we know, many Haskell developers, including significant commercial
> projects, incorporate -Wall
> -Werror into their workflow at some stage. To this end it is important
> that we minimise 'false
> negatives', whereby the compiler emits 'frivolous' warnings.
>
> This proposal identifies an important class of -Wall warnings that many
> package maintainers
> appear to regard quite bothersome, namely when the compiler warns that
> import statement is
> superflous even when it is importing an entity that is being used by the
> module. The issue is
> the way `-Wunused-imports` works, which will issue a warning for the
> following program.
>
> ```haskell
> import X -- exports some function @f@
> import y -- exports the same function @f@
>
> g = f
> ```
>
> The compiler will warn that the second Y import is redundant.
>
> Many developers do not want to be warned about this. As far as they are
> concerned each import is
> yielding enties that are in use and the fact that one of them can be
> removed is just not
> interesting.
>
> In contrast, developers that use -Wall do want to be warned about the
> following.
>
> ```haskell
> import X -- exports some function @f@
> import Z -- does not export f
>
> g = f
> ```
>
> Here our module has a completely false dependency on Z which should be
> cleaned up at the point that
> warnings are cleaned up.
>
> This proposal proposes modifying -Wunused-imports so that it will warn
> about the second example but
> stay silent for the first, and adds a -Wduplicate-imports that will issue
> a warning for the the
> first example. In effect the enabling of both warnings in the new proposal
> will restore the current
> -Wunused-imports behaviour. -Wall should *not* enable -Wduplicate-imports
> but is avilable to those
> developers that rely on the current behaviour of -Wunused-imports (which
> some clearly do).
>
>
> ## Recommendation
>
> I recommend that we accept this proposal.
>
> _______________________________________________
> ghc-steering-committee mailing list
> ghc-steering-committee at haskell.org
> https://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-steering-committee
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-steering-committee/attachments/20230615/caa0601d/attachment.html>