factor out codecs and core traits #325
Conversation
koehlma
changed the title
|
@koehlma any plans to continue on this PR? |
|
Yeah, I still think it makes sense to separate those out. I would need to have a look at the changes that happened in the meantime, as they seem to break part of this PR. There was only one failing test (prior to the conflicting changes), however, I did not understand why it was failing. It would be great if there was someone who would look into that, if that were to happen again, so that we can see this through together. I do not quite understand the A-semver-major label: This PR is designed to not be a breaking change. At least this was the intention and I don't see where it would break semver compatibility. |
I think it might be due to new crate added. |
1 similar comment
I think it might be due to new crate added. |
|
Can you resolve the conflicts please, so that it would trigger another CI run? |
This PR fixes #324.
It introduces two new crates:
compression-core: Contains theDecodeandEncodetraits.compression-codecs: Contains the formercodecmodule.I ran the changes through
cargo semver-checksand it did not complain.Along the way, I made some decisions that may require further discussion:
async-compressiondoes not expose any of the data structures of the other crates. In particular, I did not move theLevelenum and the moduleszstdandbrotliintocompression-codecs. This enables future breaking changes incompression-codecswithout breakingasync-compression.compression-codecsre-exports the underlying compression libraries. The root crate does not directly depend on any of the libraries anymore. I migrated everything, including the tests, to the re-exported libraries. It may be worthwhile to add additional abstractions tocompression-codecsto hide the underlying libraries. However, I think, that this should not be part of this PR, as it is already rather huge.