Leave serde as the only serialization method

[fjarri]

Fixes #109
Fixes #95 (or at least invalidates it)

See nucypher/nucypher-core#41 for the preview of how it will affect nucypher-core

Changes:

• Remove HasTypeName trait, and use any::type_name() where some name is still required.
• Add serde::Serialize/Deserialize implementations for Caspule, (Verified)CapsuleFrag, (Verified)KeyFrag, PublicKey, and Signature.
• Remove VerifiedCapsuleFrag::from_verified_bytes() and VerifiedKeyFrag::from_verified_bytes(). For this behavior, deserialize into CapsuleFrag or KeyFrag and call skip_verification().
• Remove DeserializableFromArray, RepresentableAsArray, SerializableToArray, SerializableToSecretArray.
• Add DefaultSerialize/DefaultDeserialize for Umbral objects (Caspule, (Verified)CapsuleFrag, (Verified)KeyFrag) which uses MessagePack via serde (gated by default-serialization feature). This is what the serialization methods in the bindings use.
• Add to_compressed_bytes()/try_from_compressed_bytes() for PublicKey to give access to non-serde representation (just 33 bytes). These are exposed in the bindings in lieu of the "default" methods (e.g. __bytes__() or toBytes()).
• Add to_der_bytes()/try_from_der_bytes() for Signature to give access to non-serde representation. These are exposed in the bindings in lieu of the "default" methods (e.g. __bytes__() or toBytes()).
• Add SecretKeyFactory::make_secret() to provide a more straightforward replacement of SecretKeyFactory::make_secret_key()->SecretKey::to_secret_bytes().
• Capsule no longer implements Copy (it's ~128 bytes, I don't think it's wise to have it)

Internal changes:

• Add Serialize/Deserialize implementations for CurveScalar and CurvePoint.
• Hash in Python bindings is calculated just with SHA2(byte_representation), instead of bothering GIL and Python builtins.

Remains to be decided:

• serde_bytes module may be better extracted into its own crate or added to some existing bytes-handling crate for serde

[codecov-commenter]

Codecov Report

Merging #110 (b3fd37f) into master (fbf2c03) will decrease coverage by 2.67%.
The diff coverage is 44.38%.

@@            Coverage Diff             @@
##           master     #110      +/-   ##
==========================================
- Coverage   51.12%   48.44%   -2.68%     
==========================================
  Files          17       17              
  Lines        1160      997     -163     
==========================================
- Hits          593      483     -110     
+ Misses        567      514      -53     

Impacted Files	Coverage Δ
umbral-pre/src/bindings_python.rs	0.00% <0.00%> (ø)
umbral-pre/src/bindings_wasm.rs	0.00% <0.00%> (ø)
umbral-pre/src/capsule_frag.rs	79.68% <0.00%> (-5.27%)	⬇️
umbral-pre/src/dem.rs	78.37% <ø> (ø)
umbral-pre/src/params.rs	100.00% <ø> (ø)
umbral-pre/src/secret_box.rs	80.00% <0.00%> (-20.00%)	⬇️
umbral-pre/src/serde_bytes.rs	41.86% <ø> (-16.48%)	⬇️
umbral-pre/src/traits.rs	0.00% <0.00%> (-58.00%)	⬇️
umbral-pre/src/keys.rs	70.10% <61.29%> (+1.17%)	⬆️
umbral-pre/src/key_frag.rs	84.94% <77.77%> (-1.52%)	⬇️
... and 7 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[theref]

what does der in to_der_bytes()/try_from_der_bytes() mean?

[fjarri]

It's the standard format for serializing signatures or certificates, see https://en.wikipedia.org/wiki/X.690#DER_encoding

[piotr-roslaniec]

I personally have nothing against fully-qualified names in this case.

[piotr-roslaniec]

LGTM

serde_bytes module may be better extracted into its own crate or added to some existing bytes-handling crate for serde

Rings a bell, serde-bytes-repr

[fjarri]

Yes, there is a bunch - see serde-rs/bytes#37 (comment). serde-bytes-repr in particular requires one to write boilerplate, which I would like to avoid.