These are chat archives for non/algebra

27th
Apr 2016
Denis Rosset
@denisrosset
Apr 27 2016 02:32
@johnynek, @non: can we keep (at minima) Java 1.7 support ? My code interfaces with Matlab, which does not support recent versions (I know...). And Java 1.6 is still important for colleagues using older versions of Matlab (depending on the funding, a few are not spending money for the upgrade). Bonus: it would keep compatibility with older versions of Android.
@non: regarding typelevel/cats#1001 thanks A LOT. I'll be very happy to help migrate Spire when the cats-kernel package stabilizes.
Denis Rosset
@denisrosset
Apr 27 2016 05:53
@non, @johnynek : regarding the (semi)group instances for integers (Int, ...)... shouldn't we have an optional adapter that instantiates a typeclass in the Semigroup tower from either the additive or the multiplicative version, if one does not exists already (using the NoImplicit trick) ? That would enable the merge of Map[Key, Field] using Monoid[Map[Key, Field]], but the specific operation would depend on the algebraic structure: addition with 0 as the map default value, or multiplication with 1 as the map default value. I can see usages for both.
So we would have two kind of adapters: those from AdditiveXXX and those from MultiplicativeXXX. We could use the same design to instantiate a Semilattice either from the meet or the join variant.
This is more a design suggestion than anything else.
Erik Osheim
@non
Apr 27 2016 12:35
@denisrosset i think we can definitely keep 1.7 compatibility. the only reason to want to drop 1.6 is that 1.7 introduces some useful comparemethods (which I think map to intrinsics) which aren't available on 1.6 (e.g. https://docs.oracle.com/javase/7/docs/api/java/lang/Integer.html#compare(int,%20int) )
Denis Rosset
@denisrosset
Apr 27 2016 13:06
I see, I'm fine with that.
Is the performance gain worth dropping Android support ?
Erik Osheim
@non
Apr 27 2016 13:08
i'm not sure.
i was mostly trying to figure out if we definitely could not drop 1.6, or if it was an option.
Denis Rosset
@denisrosset
Apr 27 2016 13:08
(I'm not so much concerned, but as algebra/cats-kernel aims to be "standard", I worry about fragmentation)
P. Oscar Boykin
@johnynek
Apr 27 2016 19:47
I’m +1 conservative on java support. lots of hadoop folks run old jvms
Erik Osheim
@non
Apr 27 2016 19:49
yeah same here, i'm reluctant to leave people out in the cold