clean the interface of supports map (paritytech#9674)

* clean the interface of supports map, make it a bit cleaner and more efficients

* Fix stiff

* fix one test

* Fix warnings

frame/election-provider-multi-phase/src/benchmarking.rs

@@ -143,7 +143,7 @@ fn solution_with_size<T: Config>(
 
 	let solution =
 		<SolutionOf<T>>::from_assignment(&assignments, &voter_index, &target_index).unwrap();
-	let score = solution.clone().score(&winners, stake_of, voter_at, target_at).unwrap();
+	let score = solution.clone().score(stake_of, voter_at, target_at).unwrap();
 	let round = <MultiPhase<T>>::round();
 
 	assert!(score[0] > 0, "score is zero, this probably means that the stakes are not set.");

frame/election-provider-multi-phase/src/lib.rs

@@ -535,8 +535,6 @@ pub enum FeasibilityError {
 	InvalidVote,
 	/// A voter is invalid.
 	InvalidVoter,
-	/// A winner is invalid.
-	InvalidWinner,
 	/// The given score was invalid.
 	InvalidScore,
 	/// The provided round is incorrect.
@@ -1395,17 +1393,8 @@ impl<T: Config> Pallet<T> {
 		let target_at = helpers::target_at_fn::<T>(&snapshot_targets);
 		let voter_index = helpers::voter_index_fn_usize::<T>(&cache);
 
-		// First, make sure that all the winners are sane.
-		// OPTIMIZATION: we could first build the assignments, and then extract the winners directly
-		// from that, as that would eliminate a little bit of duplicate work. For now, we keep them
-		// separate: First extract winners separately from solution, and then assignments. This is
-		// also better, because we can reject solutions that don't meet `desired_targets` early on.
-		let winners = winners
-			.into_iter()
-			.map(|i| target_at(i).ok_or(FeasibilityError::InvalidWinner))
-			.collect::<Result<Vec<T::AccountId>, FeasibilityError>>()?;
-
-		// Then convert solution -> assignment. This will fail if any of the indices are gibberish.
+		// Then convert solution -> assignment. This will fail if any of the indices are gibberish,
+		// namely any of the voters or targets.
 		let assignments = solution
 			.into_assignment(voter_at, target_at)
 			.map_err::<FeasibilityError, _>(Into::into)?;
@@ -1441,14 +1430,10 @@ impl<T: Config> Pallet<T> {
 		// This might fail if the normalization fails. Very unlikely. See `integrity_test`.
 		let staked_assignments = assignment_ratio_to_staked_normalized(assignments, stake_of)
 			.map_err::<FeasibilityError, _>(Into::into)?;
-
-		// This might fail if one of the voter edges is pointing to a non-winner, which is not
-		// really possible anymore because all the winners come from the same `solution`.
-		let supports = sp_npos_elections::to_supports(&winners, &staked_assignments)
-			.map_err::<FeasibilityError, _>(Into::into)?;
+		let supports = sp_npos_elections::to_supports(&staked_assignments);
 
 		// Finally, check that the claimed score was indeed correct.
-		let known_score = (&supports).evaluate();
+		let known_score = supports.evaluate();
 		ensure!(known_score == score, FeasibilityError::InvalidScore);
 
 		Ok(ReadySolution { supports, compute, score })
@@ -1653,7 +1638,7 @@ mod feasibility_check {
 			});
 			assert_noop!(
 				MultiPhase::feasibility_check(raw, COMPUTE),
-				FeasibilityError::InvalidWinner
+				FeasibilityError::NposElection(sp_npos_elections::Error::SolutionInvalidIndex)
 			);
 		})
 	}

frame/election-provider-multi-phase/src/mock.rs

@@ -30,8 +30,8 @@ use sp_core::{
 	H256,
 };
 use sp_npos_elections::{
-	assignment_ratio_to_staked_normalized, seq_phragmen, to_supports, to_without_backing,
-	ElectionResult, EvaluateSupport, ExtendedBalance, NposSolution,
+	assignment_ratio_to_staked_normalized, seq_phragmen, to_supports, ElectionResult,
+	EvaluateSupport, ExtendedBalance, NposSolution,
 };
 use sp_runtime::{
 	testing::Header,
@@ -157,25 +157,24 @@ pub fn raw_solution() -> RawSolution<SolutionOf<Runtime>> {
 	let RoundSnapshot { voters, targets } = MultiPhase::snapshot().unwrap();
 	let desired_targets = MultiPhase::desired_targets().unwrap();
 
-	let ElectionResult { winners, assignments } = seq_phragmen::<_, SolutionAccuracyOf<Runtime>>(
-		desired_targets as usize,
-		targets.clone(),
-		voters.clone(),
-		None,
-	)
-	.unwrap();
+	let ElectionResult { winners: _, assignments } =
+		seq_phragmen::<_, SolutionAccuracyOf<Runtime>>(
+			desired_targets as usize,
+			targets.clone(),
+			voters.clone(),
+			None,
+		)
+		.unwrap();
 
 	// closures
 	let cache = helpers::generate_voter_cache::<Runtime>(&voters);
 	let voter_index = helpers::voter_index_fn_linear::<Runtime>(&voters);
 	let target_index = helpers::target_index_fn_linear::<Runtime>(&targets);
 	let stake_of = helpers::stake_of_fn::<Runtime>(&voters, &cache);
 
-	let winners = to_without_backing(winners);
-
 	let score = {
 		let staked = assignment_ratio_to_staked_normalized(assignments.clone(), &stake_of).unwrap();
-		to_supports(&winners, &staked).unwrap().evaluate()
+		to_supports(&staked).evaluate()
 	};
 	let solution =
 		<SolutionOf<Runtime>>::from_assignment(&assignments, &voter_index, &target_index).unwrap();

frame/election-provider-multi-phase/src/signed.rs

@@ -26,7 +26,6 @@ use codec::{Decode, Encode, HasCompact};
 use frame_support::{
 	storage::bounded_btree_map::BoundedBTreeMap,
 	traits::{Currency, Get, OnUnbalanced, ReservableCurrency},
-	DebugNoBound,
 };
 use sp_arithmetic::traits::SaturatedConversion;
 use sp_npos_elections::{is_score_better, ElectionScore, NposSolution};
@@ -113,7 +112,7 @@ pub enum InsertResult<T: Config> {
 /// Mask type which pretends to be a set of `SignedSubmissionOf<T>`, while in fact delegating to the
 /// actual implementations in `SignedSubmissionIndices<T>`, `SignedSubmissionsMap<T>`, and
 /// `SignedSubmissionNextIndex<T>`.
-#[cfg_attr(feature = "std", derive(DebugNoBound))]
+#[cfg_attr(feature = "std", derive(frame_support::DebugNoBound))]
 pub struct SignedSubmissions<T: Config> {
 	indices: SubmissionIndicesOf<T>,
 	next_idx: u32,

frame/election-provider-multi-phase/src/unsigned.rs

@@ -315,7 +315,7 @@ impl<T: Config> Pallet<T> {
 			SolutionOf::<T>::try_from(assignments).map(|s| s.encoded_size())
 		};
 
-		let ElectionResult { assignments, winners } = election_result;
+		let ElectionResult { assignments, winners: _ } = election_result;
 
 		// Reduce (requires round-trip to staked form)
 		let sorted_assignments = {
@@ -374,8 +374,7 @@ impl<T: Config> Pallet<T> {
 		let solution = SolutionOf::<T>::try_from(&index_assignments)?;
 
 		// re-calc score.
-		let winners = sp_npos_elections::to_without_backing(winners);
-		let score = solution.clone().score(&winners, stake_of, voter_at, target_at)?;
+		let score = solution.clone().score(stake_of, voter_at, target_at)?;
 
 		let round = Self::round();
 		Ok((RawSolution { solution, score, round }, size))

frame/election-provider-support/src/onchain.rs

@@ -81,14 +81,13 @@ impl<T: Config> ElectionProvider<T::AccountId, T::BlockNumber> for OnChainSequen
 		let stake_of =
 			|w: &T::AccountId| -> VoteWeight { stake_map.get(w).cloned().unwrap_or_default() };
 
-		let ElectionResult { winners, assignments } =
+		let ElectionResult { winners: _, assignments } =
 			seq_phragmen::<_, T::Accuracy>(desired_targets as usize, targets, voters, None)
 				.map_err(Error::from)?;
 
 		let staked = assignment_ratio_to_staked_normalized(assignments, &stake_of)?;
-		let winners = to_without_backing(winners);
 
-		to_supports(&winners, &staked).map_err(Error::from)
+		Ok(to_supports(&staked))
 	}
 }
 

frame/staking/src/pallet/mod.rs

@@ -40,8 +40,8 @@ pub use impls::*;
 use crate::{
 	migrations, slashing, weights::WeightInfo, ActiveEraInfo, BalanceOf, EraIndex, EraPayout,
 	EraRewardPoints, Exposure, Forcing, NegativeImbalanceOf, Nominations, PositiveImbalanceOf,
-	Releases, RewardDestination, SessionInterface, StakerStatus, StakingLedger, UnappliedSlash,
-	UnlockChunk, ValidatorPrefs,
+	Releases, RewardDestination, SessionInterface, StakingLedger, UnappliedSlash, UnlockChunk,
+	ValidatorPrefs,
 };
 
 pub const MAX_UNLOCKING_CHUNKS: usize = 32;
@@ -453,7 +453,8 @@ pub mod pallet {
 		pub force_era: Forcing,
 		pub slash_reward_fraction: Perbill,
 		pub canceled_payout: BalanceOf<T>,
-		pub stakers: Vec<(T::AccountId, T::AccountId, BalanceOf<T>, StakerStatus<T::AccountId>)>,
+		pub stakers:
+			Vec<(T::AccountId, T::AccountId, BalanceOf<T>, crate::StakerStatus<T::AccountId>)>,
 		pub min_nominator_bond: BalanceOf<T>,
 		pub min_validator_bond: BalanceOf<T>,
 	}
@@ -502,11 +503,11 @@ pub mod pallet {
 					RewardDestination::Staked,
 				));
 				frame_support::assert_ok!(match status {
-					StakerStatus::Validator => <Pallet<T>>::validate(
+					crate::StakerStatus::Validator => <Pallet<T>>::validate(
 						T::Origin::from(Some(controller.clone()).into()),
 						Default::default(),
 					),
-					StakerStatus::Nominator(votes) => <Pallet<T>>::nominate(
+					crate::StakerStatus::Nominator(votes) => <Pallet<T>>::nominate(
 						T::Origin::from(Some(controller.clone()).into()),
 						votes.iter().map(|l| T::Lookup::unlookup(l.clone())).collect(),
 					),

primitives/npos-elections/fuzzer/src/phragmen_balancing.rs

@@ -24,7 +24,7 @@ use honggfuzz::fuzz;
 use rand::{self, SeedableRng};
 use sp_npos_elections::{
 	assignment_ratio_to_staked_normalized, is_score_better, seq_phragmen, to_supports,
-	to_without_backing, EvaluateSupport, VoteWeight,
+	EvaluateSupport, VoteWeight,
 };
 use sp_runtime::Perbill;
 
@@ -58,8 +58,7 @@ fn main() {
 					&stake_of,
 				)
 				.unwrap();
-				let winners = to_without_backing(unbalanced.winners.clone());
-				let score = to_supports(winners.as_ref(), staked.as_ref()).unwrap().evaluate();
+				let score = to_supports(staked.as_ref()).evaluate();
 
 				if score[0] == 0 {
 					// such cases cannot be improved by balancing.
@@ -83,8 +82,7 @@ fn main() {
 						&stake_of,
 					)
 					.unwrap();
-					let winners = to_without_backing(balanced.winners);
-					to_supports(winners.as_ref(), staked.as_ref()).unwrap().evaluate()
+					to_supports(staked.as_ref()).evaluate()
 				};
 
 				let enhance = is_score_better(balanced_score, unbalanced_score, Perbill::zero());

primitives/npos-elections/fuzzer/src/phragmms_balancing.rs

@@ -23,8 +23,8 @@ use common::*;
 use honggfuzz::fuzz;
 use rand::{self, SeedableRng};
 use sp_npos_elections::{
-	assignment_ratio_to_staked_normalized, is_score_better, phragmms, to_supports,
-	to_without_backing, EvaluateSupport, VoteWeight,
+	assignment_ratio_to_staked_normalized, is_score_better, phragmms, to_supports, EvaluateSupport,
+	VoteWeight,
 };
 use sp_runtime::Perbill;
 
@@ -58,8 +58,7 @@ fn main() {
 					&stake_of,
 				)
 				.unwrap();
-				let winners = to_without_backing(unbalanced.winners.clone());
-				let score = to_supports(&winners, &staked).unwrap().evaluate();
+				let score = to_supports(&staked).evaluate();
 
 				if score[0] == 0 {
 					// such cases cannot be improved by balancing.
@@ -80,8 +79,7 @@ fn main() {
 				let staked =
 					assignment_ratio_to_staked_normalized(balanced.assignments.clone(), &stake_of)
 						.unwrap();
-				let winners = to_without_backing(balanced.winners);
-				to_supports(winners.as_ref(), staked.as_ref()).unwrap().evaluate()
+				to_supports(staked.as_ref()).evaluate()
 			};
 
 			let enhance = is_score_better(balanced_score, unbalanced_score, Perbill::zero());

primitives/npos-elections/fuzzer/src/reduce.rs

@@ -104,13 +104,11 @@ fn generate_random_phragmen_assignment(
 }
 
 fn assert_assignments_equal(
-	winners: &Vec<AccountId>,
 	ass1: &Vec<StakedAssignment<AccountId>>,
 	ass2: &Vec<StakedAssignment<AccountId>>,
 ) {
-	let support_1 = to_support_map::<AccountId>(winners, ass1).unwrap();
-	let support_2 = to_support_map::<AccountId>(winners, ass2).unwrap();
-
+	let support_1 = to_support_map::<AccountId>(ass1);
+	let support_2 = to_support_map::<AccountId>(ass2);
 	for (who, support) in support_1.iter() {
 		assert_eq!(support.total, support_2.get(who).unwrap().total);
 	}
@@ -134,7 +132,7 @@ fn reduce_and_compare(assignment: &Vec<StakedAssignment<AccountId>>, winners: &V
 		num_changed,
 	);
 
-	assert_assignments_equal(winners, &assignment, &altered_assignment);
+	assert_assignments_equal(&assignment, &altered_assignment);
 }
 
 fn assignment_len(assignments: &[StakedAssignment<AccountId>]) -> u32 {

primitives/npos-elections/src/helpers.rs

@@ -17,9 +17,7 @@
 
 //! Helper methods for npos-elections.
 
-use crate::{
-	Assignment, Error, IdentifierT, PerThing128, StakedAssignment, VoteWeight, WithApprovalOf,
-};
+use crate::{Assignment, Error, IdentifierT, PerThing128, StakedAssignment, VoteWeight};
 use sp_arithmetic::PerThing;
 use sp_std::prelude::*;
 
@@ -81,11 +79,6 @@ pub fn assignment_staked_to_ratio_normalized<A: IdentifierT, P: PerThing128>(
 	Ok(ratio)
 }
 
-/// consumes a vector of winners with backing stake to just winners.
-pub fn to_without_backing<A: IdentifierT>(winners: Vec<WithApprovalOf<A>>) -> Vec<A> {
-	winners.into_iter().map(|(who, _)| who).collect::<Vec<A>>()
-}
-
 #[cfg(test)]
 mod tests {
 	use super::*;