aeson-0.10.0.0 has been out since September 2015. Aeson is widely used, and adoption of the new version was fairly slow. The changelog claims, “the API changes below should be upwards compatible from older versions of aeson. If you run into upgrade problems, please file an issue with details.”
An issue was immediately filed: Breaking changes on 0.10 regarding Null field parsing on Maybe types. (This issue remains open.)
Not long after, another issue was filed: 0.10 causes Couldn’t match expected type `Data.Text.Internal.Text’ with actual type `Data.ByteString.Builder.Internal.Builder’ (This issue was addressed fairly quickly.)
I’m not going to rehash every single issue and bugfix, but the point is that a handful of bugs have been fixed since the 0.10.0.0 release, and a few regressions still haven’t been fixed. However, what bugfixes there are have not yet been published to Hackage.
When LTS Haskell 4.0 was released with aeson-0.10, one particularly nasty regression came back into the spotlight: GHC Killed with out of memory when using generics.
Stackage curators tend to agree that including aeson-0.10.0.0 in LTS Haskell was a mistake. So, where do we go from here? We have a few options.
* Revert to aeson-0.9
* Use a “compatibility layer” package
* Wait for a patch to aeson-0.10
How have you been dealing with aeson-0.10? Did its inclusion in LTS Haskell have any impact on you and your projects? How can we improve the way we deal with situations like this in the future?