Do I know anyone in Ireland? I'm going to be visiting Cork and then Galway before FOSDEM, and then Dublin afterwards, and if you want to meet up, let me know!
Do I know anyone in Ireland? I'm going to be visiting Cork and then Galway before FOSDEM, and then Dublin afterwards, and if you want to meet up, let me know!
@judofyr yeah, I think that is a super interesting question! carrying a lot of data in the weights while parsing a very ambiguous grammar would be terrible for performance (I think you could easily push it out of O(n^3) territory, probably?) but it should still give a _correct_ result as long as you've actually satisfied the semiring laws.
also I want to back up and say again that I still think what you've written is awesome, even if you go in a different direction than I wanted to explore. you've captured an algorithm that made no sense to me and made it much more clear.
@judofyr I tend to think that if the underlying theory has fewer special cases, that there will be more surprising things you'll discover you can do with it? for example, I feel like unparsing should be almost as easy as parsing in my library, although I haven't worked out how.
I'm particularly interested in the semiring stuff because that can encode a lot of context-sensitivity in ways I think feel reasonable--like carrying a symbol table around to disambiguate C's type-vs-variable thing.
@lifning That's a fascinating question and I don't know the answer, but: I feel fairly confident in saying at least that you can't have function declarations that refer to external definitions without making those definitions available to the linker. So I think bindgen would have to generate entirely different code, involving explicit calls to dlsym, in order to do what you're asking for. But maybe it does offer that option, or maybe I'm wrong!
@audrey There are a few posts I've chosen to drop because in hindsight they were too personal or I regret having written them. But I was generally deliberately writing for a public audience on each of these platforms, and for the most part I feel I did pretty well at that. When I didn't, I personally still want to acknowledge that I'm a different person today than I was then.
But I'm certainly not going to say anyone else should do the same since everyone's circumstances differ!
currently manually importing some of my old Google+ posts into my blog. I previously imported my LiveJournal, Blogger, and ancient files from my personal website, but I forgot I'd ever used Google+ for anything interesting, and there's one post there I _really_ want to save.
That would make a very reassuring sign at the entrance of a bridge:
ยซ This bridge is open in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANDABILITY or FITNESS FOR A CROSSING PURPOSE. ยป
we build infra, nerds. bridges and roads, bolts and screws made of light and will. the consequences of our craft require higher standards than open source alone. we must accept the whole burden of stewardship, of governance, if we hope to ever make good software. that means binding ourselves and our communities of practice to both technical and ethical expectations.
you wouldn't trust a bridge that its builders consider their pet project, but that's so much modern software.
I remember being excited that Java enums could have both fields and methods, although you only get one copy of the fields per enum value so really they need to be constants set when the enum is declared. But then getting to go to Rust's proper sum-typed enums is sooo much better still!