[racket] [racket-dev] Implementing contracts for async channels

From: Robby Findler (robby at eecs.northwestern.edu)
Date: Fri Jan 16 08:34:15 EST 2015

I agree that that makes more sense for the docs. Thanks for working on this!

Robby

On Fri, Jan 16, 2015 at 4:43 AM, Alexis King <lexi.lambda at gmail.com> wrote:
> I’ve written up the relevant documentation, so I think this is more or less
> complete. I’ve submitted a pull request here. I went ahead with my gut
> instincts and put the documentation with the racket/async-channel
> documentation rather than the contract and impersonator documentation, since
> I think it makes more sense there.
>
> The only remaining potential problem is the impersonate-evt issue outlined
> previously. Frankly, I’m not even sure how to test that, much less work
> around it, so if it’s a real issue, I’d probably need some advice in fixing
> it. (I’m not even really sure how to devise a contract that would produce an
> impersonator contract.)
>
> I’ve also implemented the accompanying changes in Typed Racket and submitted
> a pull request here. Obviously, since that depends on the Racket changes, it
> can’t be pulled until the changes are merged, but I figured I’d submit a
> pull request anyway in case there are any changes that I need to make to the
> Typed Racket implementation.
>
> On Jan 15, 2015, at 23:07, Alexis King <lexi.lambda at gmail.com> wrote:
>
> Sure thing, done. I’ve moved everything into racket/async-channel, added the
> missing functions, and added some tests. I squashed my commits into one, and
> the result is here:
> https://github.com/lexi-lambda/racket/commit/0074ba13b712a87c9d05948ae075bcd74c7651e7
>
> Two simple points remain:
>
> Where should the documentation go? Should it be under contracts or
> async-channels? I’d prefer the latter, but I’m not sure.
> Since there is no impersonate-evt function, I don’t think my
> impersonate-async-channel function will actually work. Is that an accurate
> concern? How could I fix it?
>
>
> Otherwise, I think this works fine and is probably ready to go.
>
> On Jan 15, 2015, at 15:09, Robby Findler <robby at eecs.northwestern.edu>
> wrote:
>
> I think they should probably all be exported from racket/async-channel.
>
> Unless there is some reason to modify the internals of racket/contract
> to support them?
>
> Robby
>
> On Thu, Jan 15, 2015 at 4:50 PM, Alexis King <lexi.lambda at gmail.com> wrote:
>
> Sorry, I wasn’t clear. The chaperone/impersonate-async-channel functions are
> exported from racket/async-channel. The async-channel contracts, however,
> are exported from racket/contract.
>
> On Jan 15, 2015, at 14:41, Robby Findler <robby at eecs.northwestern.edu>
> wrote:
>
> Just a small nit: why export that function from racket/contract and
> not an async-channel library?
>
> Robby
>
>
> On Thu, Jan 15, 2015 at 3:33 PM, Alexis King <lexi.lambda at gmail.com> wrote:
>
> As an update, I’ve made a bit more progress on this. I’ve implemented an
> impersonate-async-channel function, and I’ve actually included this in the
> exports from racket/contract. I also realized the blame information is
> correct, it works fine. Most of the other issues remain, as well as a few
> new questions:
>
> There is no impersonate-evt function, so I’m not sure that my implementation
> will work. What should I do about this?
> I’d assume this needs to be documented/tested as well. Where should those
> things be located?
>
>
> On Jan 14, 2015, at 23:44, Alexis King <lexi.lambda at gmail.com> wrote:
>
> Currently, async channels do not have contracts to check their contents.
> This is a problem for Typed Racket, and it prevents typed code from
> interacting with code that produces async channels.
>
> I started looking into how to add contracts to the language, and it seems to
> use the chaperones/impersonator system, as I suspected. However, async
> channels obviously cannot be impersonated, so I needed to implement that as
> well.
>
> I modified the async-channel struct to use generics to allow it to be
> impersonated or chaperoned, which I have exposed by implementing
> chaperone-async-channel. I then tried implementing async-channel/c. The
> internals of the contract system are a little beyond me, but I got a working
> solution by following the example of the box contracts.
>
> My work thus far can be found here:
> https://github.com/lexi-lambda/racket/commit/84b9f3604891f3f2061fb28ed4800af8afa4751b
>
> First of all, is this a correct approach? Have I done anything wrong, or
> should I have done anything differently? I didn’t find much documentation on
> the internals of either of these systems, so I mostly went about things as I
> found them.
>
> Second, obviously, not everything is implemented here. Among the additional
> necessary changes are:
>
> I need to implement support for impersonators and impersonator contracts
> (right now I’ve only bothered to do chaperones).
> I need to implement async-channel/c-first-order and
> async-channel/c-stronger. I can probably figure out the latter, but I’m not
> even sure what the former is supposed to do.
> I need to implement a wrap-async-channel/c macro for the export. I’m not
> sure how this works, either. From looking at wrap-box/c, it seems to add
> some syntax properties, but I’m not sure what they do or how they work.
> Somehow, the blame information has to be correct. Is that what the wrap
> function does? Or do the async-channel functions need to be updated to
> assign blame?
>
>
> I’d really like to get this working, and I think I’m close, but I’m a little
> inexperienced. I’d appreciate any help, even if it’s just pointing me in the
> right direction.
>
> Thanks!
>
>
>
> _________________________
> Racket Developers list:
> http://lists.racket-lang.org/dev
>
>
>
>
>
> ____________________
>   Racket Users list:
>   http://lists.racket-lang.org/users
>


Posted on the users mailing list.