Torturing Rustc by Emulating HKTs

3 comments

• mmastrac 13 hours ago
  I've been playing around with type mapping like this in a highly experimental crate. The ability to map, filter and reduce types is insanely powerful but it's tough to get ergonomic syntax.

  https://docs.rs/tuplemagic/latest/tuplemagic/index.html
• j-krieger 15 hours ago
  I also hit a funny compiler end-of-the road last month when I wanted to try out generic_const_exprs for my RFC8366 implementation:

    pub struct VoucherDetails<C>
      where
        C: CapacityConfig,
        [(); C::SERIAL_NUMBER_MAX]:,
        [(); C::MANUFACTURER_PRIVATE_MAX]:,
        [(); C::IDEVID_ISSUER_MAX]:,
        // <imagine a dozen more properties here>
        [(); C::NONCE_B64_MAX]:,
    { ... }
  

  The heavy const eval + trait resolution was too much even for my M2 Pro.
  [-]
  • amluto 13 hours ago
    Nice.

    You inspired me to look up RFC8366. That's a remarkable document, apparently written in English, where at least the abstract and introduction look like they might be the result of a particularly nerdy game of Mad-Libs. :)

    I, personally, have never observed a pledge doing anything other than being heard, although I observed Lemon Pledge smelling unpleasant. But, in RFC8366, pledges can join domains!

    (I assume that a manufacturer makes a device and the device somehow contains a "pledge" from the manufacturer that the device is what it says it is...)
• cptroot 16 hours ago
  This was incredible, even as it rapidly outpaced my PL and mathematics knowledge.
  [-]
  • PeterWhittaker 15 hours ago
    Like I said to a friend, I know just enough category theory to know that I do not understand. Perhaps upon Nth reading.
  • spartanatreyu 13 hours ago
    I think I'll just point to this post the next time some asks me what getting nerdsniped means.