fractal witch, space dork 
~ she/her ~ 
works on compilers (rust-gpu) at @EmbarkStudios ~ "she's just like beyooom {good}" -@FreyaHolmer
You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
As an intern at microsoft working on Roslyn pinging around very silly naïve language ideas, a foundational language design lesson I learned from feedback was that "requiring whole-program analysis to determine behavior or even validity of code is very bad"
One simple, not-very-offensive example is using `var` as a return type - a caller of the method must now analyze the body of the method (and methods it calls!) to determine return type. This wreaks havoc on public API stability, predictability of code, etc.
and now I'm working on a SPIR-V compiler and SPIR-V constantly screams "I EAT WHOLE PROGRAM ANALYSIS FOR *BREAKFAST*" so, *so* many errors depend on whole-program analysis :'(
oh yeah I picked this up real quickly in pre 1.0 Rust when thinking about language design ideas super important
locality is a blessing. and not just for the computer, but for anyone having to read the code
i think my first experience with the downsides of this was c++ template errors (i used haskell before c++, but they strongly encouraged type signatures on all your functions. i didn't get why until later)
is it because it not only makes the compiler's job hard, but also makes the language harder to read for users?
Oh, crap. (Speaking as a maintainer of a whole-program analysis that determines the validity of code using GC-movable pointers.)
I'm pretty sure the GC rooting analysis isn't a whole program analysis because functions are analyzed independently. ie at a call site for Foo(), the contents of Foo() are not analyzed.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
