BTW @lqd I've been thinking about this on and off and it'd maybe be useful to bounce some things off of someone. Interestingly, I think I wound up going back to the "pure subset"-based version, though for efficiency I suspect we'll want a sort of hybrid model. It's a bit out of cache now though. I'm trying to decide how best to proceed. I feel like I've been sitting on said posts too long though.
Well, maybe I'll make another effort to finish them off just to have some starting point for discussion.
is this something where we can prototype some ideas maybe ? (but I feel you've been doing that)
(I was myself using the posts, trying out the rules, etc to start familiarizing myself with these new (to me) problems)
I think I was "basically satisfied" with the rules, iirc, but some parts felt less elegant than I might liek
and I kind of couldn't decide which was better
but yeah I have to try and remember what I was concerned about
I remember — among other things — rediscovering for the umpteenth time why some of the polonius rules are the way they are, which got me to thinking that it'd be nice to do a blog post on some of the more subtle points
(notably the interaction of the subset/liveness/etc)
but also: we really should do some a renaming PR
I want to rename
contains, in particular
I saw the renaming in one of your later revisions
if this is still more or less what you want to do, I can do the renaming
oh and I've had a lot of thoughts here. Will try to write some of them up this afternoon, I think, but the TL;DR is that I think my original post -- the one you started implememting, @lqd -- will work out just fine, and the universe stuff we will handle as a separate step
I'll go back to finish this soon after the other couple issues we talked about
@nikomatsakis while I'm a bit stuck on the other issues, I've put up the PR for the missing subset errors (and bors is still ignoring us :joy:)
@lqd btw see my most recent blog post for some discussion of the general direction I am looking at as a "next step"
loving the title already :)
Could have used a "harrumph" in there.
it's going to be fun/challenging if we have to reimplement Stephen Dolan's thesis :) (or just even understand it in my case hehe)
I feel like I see a path forward, but I'm definitely open to finding better ways of doing things