0.8.3: Add an experimental feature to directly encode - #254
Conversation
...without using intermediate data structures. Also add --stringType as a preferred spelling of --string-type, and fix a bug in the encoding of packed "sint32" fields that harmed forward compatibility with "sint64" and increased message sizes.
...at the expense of some ambiguity in the type parameters of classes. Also expand test coverage.
Also reorganize Repetition and rename Encoding to MessageEncoding.
Previously we would be left with unreduced type family applications instead.
j6carey
changed the title
| pretty ==1.1.*, | ||
| pretty-show >= 1.6.12 && < 2.0, | ||
| proto3-wire >= 1.2.2 && < 1.5, | ||
| proto3-wire >= 1.4.4 && < 1.5, |
There was a problem hiding this comment.
Latest on hackage is 1.4.3, I think this broke cabal build
There was a problem hiding this comment.
Latest on hackage is 1.4.3, I think this broke
cabal build
Thanks for letting us know. I uploaded proto3-wire-1.4.4 and will get the rest of the versions missing packaged and published shortly.
There was a problem hiding this comment.
I see 1.4.4 and 1.4.5 have been released, which allowed me to fix the proto3-suite build in nixpkgs against 1.4.4 with NixOS/nixpkgs#458271 by pointing proto3-suite at #274. Thanks for that.
Ideally we could just use the hackage release, but latest for proto3-suite is 0.7.0, too far behind to even allow comfortably cherry-picking select commits (e.g. bounds) without conflicts.
Note nixpkgs relies on a couple proto3-wire overrides which are needed until a release of current master:
- 1.4.5: Add reverse encoders for packed repeated fields. proto3-wire#108
- Support LTS 24 proto3-wire#111
I noticed you use nixpkgs here as well. In case you're unaware, haskellPackages get most versions from stackage LTS, so it's often the same work getting proto3-* available on both rather than just one
4 participants
...without using intermediate data structures.
Also add --stringType as a preferred spelling of --string-type,
and fix a bug in the encoding of packed "sint32" fields that harmed
forward compatibility with "sint64" and increased message sizes.