keyring: remove lazy_static public keys hash maps

[michalkucharczyk]

The lazy_static package does not work well in no-std: it requires spin_no_std feature, which also will propagate into std if enabled. This is not what we want.

This PR removes public/private key hash-maps and replaces them with simple static byte arrays.

&T versions of AsRef/Deref/From traits implementation were removed.

Little const helper for converting hex strings into array during compile time was also added. (somewhat similar to hex_literal).

[davxy]

The code obviously works. There is not much to say about it.

The only thing that I still don't like (but maybe is subjective) is that there are two obvious ways to do the same thing.

That is, getting a public key using:

• Keyring::pair().public()
• From<Keyring> or [u8; 32]

In practice this new approach is the same as the one used in the previous iteration only that the code that previously was in a member function has not being moved in From<Keyring> for [u8; 32]`.

Again, IMO would be way better to have one way or another.

• One which is used for std or full_crypto or whatever you want to call the feature in this crate (which also includes the Pair` and capability to sign).
• Another where fn public internally uses this new approach of pre-constructed arrays.

If at some point you want to include this stuff in no_std you may want to remove fn pair() anyway as to sign with a Schnorr key you need randomness.

[michalkucharczyk]

IUC: the consequence of your proposal will be calling the pair().public() everytime (w/o a PUBLICK_KESY cache which was just removed) in full_crypto and std.

Having the public key in form of raw bytes prevents this call (and is closer to cached approach).

[michalkucharczyk]

ok, I think I got it, you simply want this function to use predefined raw bytes, right?

polkadot-sdk/substrate/primitives/keyring/src/sr25519.rs

Lines 94 to 96 in 88ae185

	pub fn public(self) -> Public {
	self.pair().public()
	}

[davxy]

IUC: the consequence of your proposal will be calling the pair().public() everytime (w/o a PUBLICK_KESY cache which was just removed) in full_crypto and std.

Having the public key in form of raw bytes prevents this call (and is closer to cached approach).

Caching is pointless here. Alice, Bob and co keyring is used for tests, templates and a like. Never in production

[davxy]

ok, I think I got it, you simply want this function to use predefined raw bytes, right?

The ideal is:

• if I can compute from pair() => compute from pair
• if I can't => use precomputed

As I said above, caching is not really important for these keys

[michalkucharczyk]

IUC: the consequence of your proposal will be calling the pair().public() everytime (w/o a PUBLICK_KESY cache which was just removed) in full_crypto and std.
Having the public key in form of raw bytes prevents this call (and is closer to cached approach).

Caching is pointless here. Alice, Bob and co keyring is used for tests, templates and a like. Never in production

This will be done in chainspec generation (for all testnets, zombienets, etc...), which will be done in runtime. Why not keep this optimization?

[michalkucharczyk]

Is the cost of calling pair() actually negligible?

[davxy]

IUC: the consequence of your proposal will be calling the pair().public() everytime (w/o a PUBLICK_KESY cache which was just removed) in full_crypto and std.
Having the public key in form of raw bytes prevents this call (and is closer to cached approach).

Caching is pointless here. Alice, Bob and co keyring is used for tests, templates and a like. Never in production

This will be done in chainspec generation (for all testnets, zombienets, etc...), which will be done in runtime. Why not keep this optimization?

1. I'm not saying that is negligible. But for sure we do a bazilion of hashes and computations around which probably makes it negligible in the contexts where is done (I haven't done any benchmarks BTW)
2. Haven't you already removed the cache with this PR? (i.e. PUBLIC_KEYS and PRIVATE_KEYS)

My suggestion was such that in the runtime you end up using the fn public() version which doesn't use pair(). I.e. the one which uses the pre-computed stuff. There is not real value exposing fn pair() in the runtime as you can't use fn sign() for Schnorr and ECDSA (which in general requires randomness)

[michalkucharczyk]

Support for no-std (e.g removal of sign method) is added in:
#2044

[michalkucharczyk]

Yeah, sorry my oversight, Obviously fn public() method should have been using the raw bytes. Fixed in 25a0a89.

[davxy]

Support for no-std (e.g removal of sign method) is added in: #2044

Ok I remember. But I also was thinking that with the introduction of statically computed Public having the Pair trait for no_std was not required.

What is the usage of Pair in no_std at this point? Wasn't all that work done because you were requiring to compute Public? Maybe offchain workers who doesn't want to use the keystore via HF?

[michalkucharczyk]

We can gate pair() method also. I will add it in #2044 (as it introduces support of Keyring in runtimes).

Probably there is no use case for this in runtime. On the other hand we will have this functionality for free, and if not used linker should strip it off.

[michalkucharczyk]

Sorry for bikeshedding on this

It was different at the beggining, I spotted my mistake here: #2387 (comment)

[michalkucharczyk]

bot update-ui

[command-bot]

@michalkucharczyk https://gitlab.parity.io/parity/mirrors/polkadot-sdk/-/jobs/4662342 was started for your command "$PIPELINE_SCRIPTS_DIR/commands/update-ui/update-ui.sh". Check out https://gitlab.parity.io/parity/mirrors/polkadot-sdk/-/pipelines?page=1&scope=all&username=group_605_bot to know what else is being executed currently.

Comment bot cancel 1-3f5db3cb-98b2-4490-b5c9-da31f43ceeaf to cancel this command or bot cancel to cancel all commands in this pull request.

[command-bot]

@michalkucharczyk Command "$PIPELINE_SCRIPTS_DIR/commands/update-ui/update-ui.sh" has finished. Result: https://gitlab.parity.io/parity/mirrors/polkadot-sdk/-/jobs/4662342 has finished. If any artifacts were generated, you can download them from https://gitlab.parity.io/parity/mirrors/polkadot-sdk/-/jobs/4662342/artifacts/download.