Closed matthewleon closed 6 years ago
:+1: but since this is a breaking change, possibly best to wait until we're doing the next big round of breaking changes? Do we know how much stuff depends on this library?
My suspicion is that there is very little out there that is using the new Homogeneous stuff. In any case, if you increment the major version number to mark the breaking change, shouldn't the conventional bower format ensure that the dependent libs don't break?
True, I guess we could release this as a breaking change and then update any dependent libraries to loosen the upper bounds. According to https://github.com/purescript/package-sets/blob/c375ba150ee53c4646b869804e2494ca26679dde/packages.json that's quickcheck
, quickserve
, simple-json
, typelevel
, and variant
.
I see... Seems like either option here is unpleasant in some way or another. :(
Only one of those is a core library though, and 5 downstream libraries is not very many, so that's actually not a bad option imo. What do other maintainers think?
I'd say make a breaking release and bump deps in the other libraries. package-sets are a great tool to figure out what to bump. I think this might also be a good chance to see if we can build a bit of tooling around this?
Around loosening upper bounds for downstream libraries, you mean?
Also figuring out a linear order in which to bump versions and maybe even automatically make PRs which bump dependencies. If we have CI on these repos we can then just merge or intervene if compilation fails.
Ah right, yeah, that would be nice. Also what kind of version bump do you think we should use for downstream libraries if they don't use the Homogeneous
stuff, i.e. these changes don't actually mean anything for downstream libraries?
Normally I'd just say 'patch' but the fact that you can implicitly depend on stuff transitively is perhaps a bit of a problem..?
It needs to be major, since bumping the minor version on the downstream library would pull in the breaking change.
Edited.
I feel pretty bad about this 😕
I think I copied and pasted the suggestion to rename this from FieldOf and didn't realize that the spelling was different.
@kRITZCREEK yeah, I suppose you're right. I'd really like to change this state of affairs, I think it makes this type of change much more difficult than it really needs to be.
Updated the instance spellings.
@paulyoung I don't think there's anything to feel too bad about. This kind of thing happens: https://en.wikipedia.org/wiki/HTTP_referer
Since there is a 0.12 branch where breaking things are being merged, does this perhaps belong in there?
This is very picky, but I feel like it should probably be done now, while this is less than a week old, than sit until it becomes entrenched.