Haskell Platform Proposal: HLint

[Neil Mitchell]

Hi

>> I'm not entirely sure what the intended audience is for this simple
>> AST. Unless someone's going to use it, I don't see much point in it
>> existing.
>
> When you just use the AST, you can ignore the bits that
> you don't need. But when you need to transform it, there is
> no way to ignore parts of it. And that creates a huge amount
> of extra work and boilerplate code with the current
> feature-loaded AST.

If you aren't using a boilerplate removal library, and are working
with haskell-src-exts, then you are very probably doing it wrong! I
recommend Uniplate, but there are plenty of other options including
SYB and anything with the phrase generic and Haskell in it from the
last 15 years. I really cannot stress how much difference a
boilerplate removal library makes.

I'd also love a "Core Haskell" library, like HSE but with
desugaring/type-checking/name-resolution etc. That's not what HSE
does, but it is a niche that needs filling.

> In addition, a high-quality parser whose AST more faithfully
> reflects the Haskell 20nn standard is very useful in many ways
> in the area of verification, validation, and certification.

Making a parser faithfully reflect the standard is not a trivial
proposition, and one that I don't think anyone is actually going to
check. A parser that kinda-might reflect the standard doesn't help
with verification/validation/certification.

> If people are absolutely keen on getting haskell-src-exts
> into the platform immediately with its new name, another strategy
> could be to rename haskell-src to haskell98-src and haskell-src-exts to
> haskell-src.

I always talk about HSE, so if the name changes I'll need a new
acronym, and HS doesn't sound like a great choice :-) I'm happy with
any name, and I don't really have an opinion about including HSE or
not (it's awesome, but it being in the platform doesn't really benefit
me massively, as it's so easy to install with cabal). All I care about
is Niklas not using up his valuable time on a project that doesn't pay
off for users.

Thanks, Neil