> Maybe it's mostly locked up in tribal knowledge and you have to be in the right room with the right people to learn it.
This is sad.
Re:Crafting Interpreters: Classes
by munificent on Monday October 30, 2017 @08:35PM
(#15590263)
Tell me about it. As someone who works in programming languages professionally but has no academic background in it, it's maddening how much stuff is impossible to learn because "you can't publish a paper on it because everyone knows it already".
Years ago I dreamed that Lambda the Ultimate had added a wiki, so language implementation was quickening. Waking up... :(
Similarly in the sciences, the gaps can be surprisingly large between insightful discussion among principals at conferences, and "you have to know the people involved to evaluate them" narrow papers, and misleading texts, and at great remove, introductory material.
I'm assuming that you've seen this then? http://people.csail.mit.edu/jrb/Projects/pd.pdf
https://github.com/googoogaga/goo continued for a bit, then didn't.
OT: With Google Chrome's new Turbofan, and its altered inlining budgeting, there now seems a chance of inlining dispatch trees. Combined with PICs, and cpu branch prediction and speculation... there seems hope for wizzy multiple dispatch in javascript. Maybe. I don't know of anyone pushing on it.
Re:Crafting Interpreters: Classes
by munificent on Monday October 30, 2017 @06:44PM
(#15589646)
No, I haven't. Thanks for the reference!
You mention CLOS in your article; to dive deeper into implementation issues, see also https://en.wikipedia.org/wiki/The_Art_of_the_Metaobject_Prot....
Re:Crafting Interpreters: Classes
by munificent on Monday October 30, 2017 @07:48PM
(#15590031)
Yup, I read AMOP several years ago. But as I recall, it doesn't talk too much about low-level implementation shenanigans.
What I liked about it is the discussion about design decisions: where and how to let the protocol be open to extensions, etc. (btw, sorry about the pedantic tone of the previous comment).
No, the book will only go into implementing single dispatch. I'd love to talk about implementing multimethods, but I don't know how to implement them very well either. :) I've read a few papers on it, but there doesn't seem to be a lot of literature. Maybe it's mostly locked up in tribal knowledge and you have to be in the right room with the right people to learn it.