Linear Types: ready for review

[Simon Peyton Jones]

Dear GHC developers and friends

I would like to urge you to review, and help to improve, this patch.

Adding linear types to Haskell is a big change that touches a lot of code.   That makes the patch a bit daunting.  But GHC is supposed to be a laboratory for advanced stuff, and this is a great example.   The authors (esp Arnaud, Krzysztof, and Jean-Philippe) have invested a lot of effort in the paper<https://www.microsoft.com/en-us/research/publication/linear-haskell-practical-linearity-higher-order-polymorphic-language/>, the proposal<https://github.com/tweag/ghc-proposals/blob/linear-types2/proposals/0000-linear-types.rst>, and now the patch<https://gitlab.haskell.org/ghc/ghc/merge_requests/852>.  And lots of people are interested in the result.  In short, we owe it to them to give their work serious attention.

But please don’t think “oh, I’m not qualified and anyway Simon will look at it”.  I will, of course, but like you I’m too busy.  And a far-reaching change like this needs many eyes.   Moreover, assuming it’s committed, this code will be us for a long time;       most future developers will be no better qualified than you, but they’ll still have to deal with it!  So if you get bamboozled, make constructive suggestions about what sorts of Notes or explanation would un-bamboozle you.  Don’t just assume you are too stupid; that way likes an un-maintainable code base.

So please jump in there.  This is  your compiler.

Thanks

Simon

From: ghc-devs <ghc-devs-bounces at haskell.org> On Behalf Of Spiwack, Arnaud
Sent: 03 May 2019 10:52
To: GHC developers <ghc-devs at haskell.org>
Cc: Krzysztof Gogolewski <krzysztof.gogolewski at tweag.io>
Subject: Linear Types: ready for review

Dear GHC devs,

The linear types patch is ready for review!
https://gitlab.haskell.org/ghc/ghc/merge_requests/852

It is a fairly large patch (~3500loc of change to the compiler itself,
with a ~2000loc diff on the testsuite). And it's gone through a lot of
redesigns. The issue with a patch like this is that it is rather easy
to get reviewer fatigue and just let things through.

This is why I want to encourage as many people as possible to give a
thorough review of part of the patch. Even 15 minutes of your time
would be tremendously helpful!

Here is what I suggest: pick a part of the compiler you know well, and
look at how the patch affects it. If there is something that you don't
understand let us know! Either in the MR discussion, or send an email
to Krzysztof and me. We will answer and update the documentation.

---

If you are still unsure where to start, here are what we consider to
be the main entry points to the patch:

- The file Multiplicity, which defines type Mult = Type, the Scaled
  type and functions unrestricted, linear, pattern synonyms One and
  Omega, quick submultiplicity test submult
- The change to FunTy and new mkFunTy in TyCoRep, where the
  multiplicity argument is added
- The change to funTyCon in TysPrim
- The new unrestrictedFunTyCon and multiplicityTy in TysWiredIn
- Var has now a new field varMult; functions such as
  updateVarTypeAndMult can be used to update multiplicity (e.g. when
  zonking). In Core, the multiplicity of a lambda is indicated by its
  varMult; multiplicity of a case expression is indicated by varMult
  of the case binder.
- The functions mkDataConRepXin basicTypes/MkId and and
  dataConUserType function in basicTypes/DataCon, which are adding
  multiplicity variables to constructors. The change to wrapper_reqd
  means that all data constructors have a wrapper now.
- The file UsageEnv defines a map from a variable to its multiplicity,
  used when typechecking
- Calls to submult and tcSubMult, which guide multiplicity checking in
  the type checker
- ensureSubMult and its calls, which guide multiplicity checking in
  Lint

---

A more in-depth documentation and overview of the patch is on the
wiki: https://gitlab.haskell.org/ghc/ghc/wikis/linear-types/implementation

It has a table of contents which you can use as yet another way to
find a starting point.

---

If any of these resources are lacking, let us know too! We'll make
sure to update them.

---

Finally, I'll be holding a project session on Linear Types in ZüriHac
next month. Where we can have further, more in-depth discussion about
the patch, finalise its documentation, and address outstanding
requests from reviews.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-devs/attachments/20190507/71e82470/attachment.html>