feat(primitive-types, serde): deserialize hex strings without 0x prefix

[shekhirin]

Support hex strings which are not prefixed with 0x.

Example from geth's debug_traceTransaction (storage key):

> debug.traceTransaction("0x2059dd53ecac9827faad14d364f9e04b1d5fe5b506e3acc886eff7a6f88a696a")
{
  gas: 85301,
  returnValue: "",
  structLogs: [{
      depth: 1,
      error: "",
      gas: 100000,
      gasCost: 0,
      memory: ["0000000000000000000000000000000000000000000000000000000000000006", "0000000000000000000000000000000000000000000000000000000000000000", "0000000000000000000000000000000000000000000000000000000000000060"],
      op: "STOP",
      pc: 120,
      stack: ["00000000000000000000000000000000000000000000000000000000d67cbec9"],
      storage: {
        0000000000000000000000000000000000000000000000000000000000000004: "8241fa522772837f0d05511f20caa6da1d5a3209000000000000000400000001",
        0000000000000000000000000000000000000000000000000000000000000006: "0000000000000000000000000000000000000000000000000000000000000001",
        f652222313e28459528d920b65115c16c04f3efc82aaedc97be59f3f377c0d3f: "00000000000000000000000002e816afc1b5c0f39852131959d946eb3b07b5ad"
      }
  }]

[shekhirin]

@ordian fixed the remaining tests

[ordian]

@tomusdrw is this going to break anything? We're less restrictive here, but also
assert_eq!(serialize(deserialize(s)?), s) is broken in some cases now.

Other than that, LGTM.

[tomusdrw]

We've been trying to be strict to avoid ambiguity between hex and decimal encoding. Also it used to be part of the Ethereum JSON-RPC spec to always prefix hex values with 0x. I guess it's not the case any more. I don't see a reason why it would break anything.

[shekhirin]

sounds good. Merge?

[niklasad1]

LGTM, the changelog needs to be updated not sure if any project is using these types in the ethereum JSON-RPC interface anymore but still...

[niklasad1]

dq: is (v.len() + 1) / 2 used to work in case the hex string is not divisible by 2?

when it's even we would allocate one extra byte but yeah it was like this before too..

[shekhirin]

yeah, there're some tests like

assert_eq!(from_hex("102"), Ok(vec![1, 2]));

which wouldn't pass if bytes.len() would be 1 instead of 2, so we need to round up. It was - 1 previously because we were always expecting 0x prefix to present.

[shekhirin]

@niklasad1 I updated the changelog

[ordian]

Ok, thanks. I'll merge and publish it today.

[ordian]

Published.

[shekhirin]

@ordian I forgot to update ethereum-types version since I've changed used impl-serde version inside it. Bumped here: #599