Expose validators as host function

[ilblackdragon]

For voting contract #2474 it's required to query current stake by validator account id.

TBD the APIs.

[evgenykuzyakov]

It's not simple to provide all validators with their stakes into the wasm because of limited support for serialization. Our current design for WASM was serialization independent, so we didn't settle on Borsh.

I suggest the following API: validator_stake(account_id_len: u64, account_id_ptr: u64, stake_ptr: u64) -> [] where

• account_id_len and account_id_ptr is the account ID of the validator to check
• stake_ptr is the pointer to write the current active stake of a given validator Account ID

[evgenykuzyakov]

Another function we need is total_validator_stake(stake_ptr: u64) -> [] to return the total active stake balance. It's needed for denominator when resolving voting majority.

[evgenykuzyakov]

We can also expose the validators the following way: number_of_validators() -> u64 validator_id(validator_index: u64, register_id: u64) -> [] validator_stake(validator_index: u64, stake_ptr: u64) -> []

Which will make it flexible enough so long as we can pass validators in sorted order and the list of validators is not long enough.

[bowenwang1996]

We also need validator_id_to_account_id(validator_index: u64, account_id_ptr: u64) -> [].

[evgenykuzyakov]

Sorry, I meant validator_id(validator_index: u64, register_id: u64) -> [] is validator_account_id(validator_index: u64, register_id: u64) -> []

register_id contains the account ID

[bowenwang1996]

Do we actually need the list of all validators? It seems to me that having

validator_stake(account_id_len: u64, account_id_ptr: u64, stake_ptr: u64) -> []
total_validator_stake(stake_ptr: u64) -> []

is better because then if you want to know the stake of a specific validator, you don't need to iterate through all validators, which would be gas-consuming.

[evgenykuzyakov]

Let's create a spec change in NEPs. I think we can go with the following API:

number_of_validators() -> u64
validator_account_id(validator_index: u64, register_id: u64) -> []
validator_stake(validator_index: u64, stake_ptr: u64) -> []
validator_index(account_id_len: u64, account_id_ptr: u64) -> u64
total_validator_stake(stake_ptr: u64) -> []

[evgenykuzyakov]

Voting contract can't iterate through all validators, because they might not be elected in a given epoch. The total amount of validators who tried to stake (but didn't get elected into 100 seats) can be huge. We can't possibly know who was elected in a current epoch, unless we tracked every single stake proposal from the previous epoch, so this is not the case.

It leaves us with the only option for iterating through the list of currently active seats. Ideally they should be sorted in order from the largest stake to the lowest. Iterating through 100 seats is not a lot of gas, because it doesn't touch storage.

[bowenwang1996]

@evgenykuzyakov I don't understand why the voting contract has to iterate through all validators. From my perspective, it is sufficient to know for each account id, what their stake is (0 if they are not a validator) and the total stake. With this information it is sufficient to determine whether a proposal has the majority of votes.

[evgenykuzyakov]

For this you have to iterate through all possible votes within an epoch. It means if the validator doesn't call the voting contract within this epoch, his vote will not be counted, unless you want to iterate through casted votes. We need to design a contract in such a way that if the epoch is 5 minutes, then we still should be able to complete voting. We shouldn't require validators to re-vote every epoch.

[MaksymZavershynskyi]

This seems to be a duplicate of https://github.com/nearprotocol/nearcore/pull/2504 . Closing it.

[bowenwang1996]

@nearmax but there is no estimate on #2504, so I suggest keeping this one open until the PR is merged.