> On Jan 23, 2017, at 4:54 PM, Andrew Kent <[email protected]> wrote:
>
> I personally find the performance overhead added by the dependent contracts
> in racket/dict to be too burdensome for general use in large applications. I
> would be much happier if the default interface was instead comprised of the
> simple contracts the documentation for gen:dict seems to suggest.
>
> Am I in a small minority with this opinion? Would people be amicable to a PR
> that simplified these contracts? Or would losing the extra contract checking
> be a noticeable drawback for some projects?
>
> As an anecdotal side note: this overhead initially gave me a bad taste in my
> mouth for Racket's generics (I initially thought they were to blame for the
> performance slowdown). It wasn't until later that I realized the overhead
> from using generics is _quite reasonable_ (I've used them in Typed Racket to
> speed things up in fact) and that really this is an issue of the pros and
> cons of using expressive, dependent contracts. Anyway, I hope others haven't
> been similarly turned off to generics (which are awesome!) if they've noticed
> undesirable overhead while trying out racket/dict.
>
> Best,
> Andrew
>
> P.S. I think ideally both interfaces should be provided (one with simple
> contracts and one without) but doing this right might require some tools that
> haven't been invented yet (in particular, this seems like another use case
> for what I like to call "negotiable contracts," which would allow modules to
> provide multiple contract options for identifiers that a requiring module
> could then specify at compile time which they want with a require form -- I
> should really get on this…)
1. I think this is a place where we could experiment with ‘option contracts’.
2. I think this also makes a great place where we could experiment with
supplying two variants:
— gen:dict-with
- gen:dict-out
or whatever. They can use the same source module in a private directory and
offer two interfaces.
We could even go to three. The interesting question would then become whether
people
switch to . . . -with when things blow up inside of gen:dict so that they get
better
error messages. That would partly confirm our ‘contract pressure’ conjecture.
In general I don’t see why we need more work per se. We can offer several
options,
including option contracts.
— Matthias
--
You received this message because you are subscribed to the Google Groups
"Racket Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/racket-dev/D27A2061-70D9-40D6-BF82-0F64F10A5660%40ccs.neu.edu.
For more options, visit https://groups.google.com/d/optout.
