diff --git a/Cargo.lock b/Cargo.lock index 76fabe80eb1be..7bb13ea3b958d 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -4417,6 +4417,21 @@ dependencies = [ "sp-std", ] +[[package]] +name = "pallet-scheduler" +version = "2.0.0-alpha.4" +dependencies = [ + "frame-benchmarking", + "frame-support", + "frame-system", + "parity-scale-codec", + "serde", + "sp-core", + "sp-io", + "sp-runtime", + "sp-std", +] + [[package]] name = "pallet-scored-pool" version = "2.0.0-alpha.5" diff --git a/Cargo.toml b/Cargo.toml index d86bab8abd533..2737214c08323 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -86,6 +86,7 @@ members = [ "frame/offences", "frame/randomness-collective-flip", "frame/recovery", + "frame/scheduler", "frame/scored-pool", "frame/session", "frame/session/benchmarking", diff --git a/frame/democracy/src/lib.rs b/frame/democracy/src/lib.rs index a4366c4ef23ee..1affba7706856 100644 --- a/frame/democracy/src/lib.rs +++ b/frame/democracy/src/lib.rs @@ -173,7 +173,7 @@ use frame_support::{ weights::{SimpleDispatchInfo, Weight, WeighData}, traits::{ Currency, ReservableCurrency, LockableCurrency, WithdrawReason, LockIdentifier, Get, - OnUnbalanced, BalanceStatus + OnUnbalanced, BalanceStatus, schedule } }; use frame_system::{self as system, ensure_signed, ensure_root}; @@ -270,6 +270,9 @@ pub trait Trait: frame_system::Trait + Sized { /// Handler for the unbalanced reduction when slashing a preimage deposit. type Slash: OnUnbalanced>; + + /// The Scheduler. + type Scheduler: schedule::Named; } decl_storage! { @@ -1628,6 +1631,7 @@ impl Module { let _ = Self::enact_proposal(status.proposal_hash, index); } else { let item = (now + status.delay, status.proposal_hash, index); + >::mutate(|queue| { let pos = queue.binary_search_by_key(&item.0, |x| x.0).unwrap_or_else(|e| e); queue.insert(pos, item); diff --git a/frame/scheduler/Cargo.toml b/frame/scheduler/Cargo.toml new file mode 100644 index 0000000000000..017518d882df4 --- /dev/null +++ b/frame/scheduler/Cargo.toml @@ -0,0 +1,35 @@ +[package] +name = "pallet-scheduler" +version = "2.0.0-alpha.4" +authors = ["Parity Technologies "] +edition = "2018" +license = "Unlicense" +homepage = "https://substrate.dev" +repository = "https://github.com/paritytech/substrate/" +description = "FRAME example pallet" + +[dependencies] +serde = { version = "1.0.101", optional = true } +codec = { package = "parity-scale-codec", version = "1.2.0", default-features = false } +frame-benchmarking = { version = "2.0.0-alpha.4", default-features = false, path = "../benchmarking" } +frame-support = { version = "2.0.0-alpha.4", default-features = false, path = "../support" } +frame-system = { version = "2.0.0-alpha.4", default-features = false, path = "../system" } +sp-runtime = { version = "2.0.0-alpha.4", default-features = false, path = "../../primitives/runtime" } +sp-std = { version = "2.0.0-alpha.4", default-features = false, path = "../../primitives/std" } +sp-io = { version = "2.0.0-alpha.4", default-features = false, path = "../../primitives/io" } + +[dev-dependencies] +sp-core = { version = "2.0.0-alpha.4", path = "../../primitives/core", default-features = false } + +[features] +default = ["std"] +std = [ + "serde", + "codec/std", + "sp-runtime/std", + "frame-benchmarking/std", + "frame-support/std", + "frame-system/std", + "sp-io/std", + "sp-std/std" +] diff --git a/frame/scheduler/src/lib.rs b/frame/scheduler/src/lib.rs new file mode 100644 index 0000000000000..7ccb43b4493bd --- /dev/null +++ b/frame/scheduler/src/lib.rs @@ -0,0 +1,517 @@ +// Copyright 2017-2020 Parity Technologies (UK) Ltd. +// This file is part of Substrate. + +// Substrate is free software: you can redistribute it and/or modify +// it under the terms of the GNU General Public License as published by +// the Free Software Foundation, either version 3 of the License, or +// (at your option) any later version. + +// Substrate is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +// GNU General Public License for more details. + +// You should have received a copy of the GNU General Public License +// along with Substrate. If not, see . + +//! # Scheduler +//! +//! \# Scheduler +//! +//! - \[`scheduler::Trait`](./trait.Trait.html) +//! - \[`Call`](./enum.Call.html) +//! - \[`Module`](./struct.Module.html) +//! +//! \## Overview +//! +//! // Short description of pallet's purpose. +//! // Links to Traits that should be implemented. +//! // What this pallet is for. +//! // What functionality the pallet provides. +//! // When to use the pallet (use case examples). +//! // How it is used. +//! // Inputs it uses and the source of each input. +//! // Outputs it produces. +//! +//! \## Terminology +//! +//! \## Goals +//! +//! \## Interface +//! +//! \### Dispatchable Functions + +// Ensure we're `no_std` when compiling for Wasm. +#![cfg_attr(not(feature = "std"), no_std)] + +use sp_std::prelude::*; +use codec::{Encode, Decode}; +use sp_runtime::{RuntimeDebug, traits::{Zero, One}}; +use frame_support::{ + dispatch::{Dispatchable, DispatchResult, Parameter}, decl_module, decl_storage, decl_event, + traits::{Get, schedule}, + weights::{GetDispatchInfo, Weight}, +}; +use frame_system::{self as system}; + +/// Our pallet's configuration trait. All our types and constants go in here. If the +/// pallet is dependent on specific other pallets, then their configuration traits +/// should be added to our implied traits list. +/// +/// `system::Trait` should always be included in our implied traits. +pub trait Trait: system::Trait { + /// The overarching event type. + type Event: From> + Into<::Event>; + + /// The aggregated origin which the dispatch will take. + type Origin: From>; + + /// The aggregated call type. + type Call: Parameter + Dispatchable::Origin> + GetDispatchInfo; + + /// The maximum weight that may be scheduled per block for any dispatchables of less priority + /// than 255. + type MaximumWeight: Get; +} + +/// Just a simple index for naming period tasks. +pub type PeriodicIndex = u32; +/// The location of a scheduled task that can be used to remove it. +pub type TaskAddress = (BlockNumber, u32); + +/// Information regarding an item to be executed in the future. +#[derive(Clone, RuntimeDebug, Encode, Decode)] +pub struct Scheduled { + /// The unique identity for this task, if there is one. + maybe_id: Option>, + /// This task's priority. + priority: schedule::Priority, + /// The call to be dispatched. + call: Call, + /// If the call is periodic, then this points to the information concerning that. + maybe_periodic: Option>, +} + +decl_storage! { + trait Store for Module as Scheduler { + /// Items to be executed, indexed by the block number that they should be executed on. + Agenda: map hasher(twox_64_concat) T::BlockNumber + => Vec::Call, T::BlockNumber>>>; + + /// Lookup from identity to the block number and index of the task. + Lookup: map hasher(twox_64_concat) Vec => Option>; + } +} + +decl_event!( + pub enum Event where ::BlockNumber { + Scheduled(BlockNumber), + Dispatched(TaskAddress, Option>, DispatchResult), + } +); + +decl_module! { + // Simple declaration of the `Module` type. Lets the macro know what its working on. + pub struct Module for enum Call where origin: ::Origin { + fn deposit_event() = default; + + fn on_initialize(now: T::BlockNumber) -> Weight { + let limit = T::MaximumWeight::get(); + let mut queued = Agenda::::take(now).into_iter() + .enumerate() + .filter_map(|(index, s)| s.map(|inner| (index as u32, inner))) + .collect::>(); + queued.sort_by_key(|(_, s)| s.priority); + let mut result = 0; + let unused_items = queued.into_iter() + .scan(0, |cumulative_weight, (index, s)| { + *cumulative_weight += s.call.get_dispatch_info().weight; + Some((index, *cumulative_weight, s)) + }) + .filter_map(|(index, cumulative_weight, mut s)| { + if s.priority <= schedule::HARD_DEADLINE || cumulative_weight <= limit { + let r = s.call.clone().dispatch(system::RawOrigin::Root.into()); + let maybe_id = s.maybe_id.clone(); + if let &Some((period, count)) = &s.maybe_periodic { + if count > 1 { + s.maybe_periodic = Some((period, count - 1)); + } else { + s.maybe_periodic = None; + } + let next = now + period; + if let Some(ref id) = s.maybe_id { + let next_index = Agenda::::decode_len(now + period).unwrap_or(0) as u32; + Lookup::::insert(id, (next, next_index)); + } + Agenda::::append_or_insert(next, &[Some(s)][..]); + } else { + if let Some(ref id) = s.maybe_id { + Lookup::::remove(id); + } + } + Self::deposit_event(RawEvent::Dispatched((now, index), maybe_id, r)); + result = cumulative_weight; + None + } else { + Some(Some(s)) + } + }) + .collect::>(); + if !unused_items.is_empty() { + let next = now + One::one(); + Agenda::::append_or_insert(next, &unused_items[..]); + } + result + } + } +} + +impl schedule::Anon::Call> for Module { + type Address = TaskAddress; + + fn schedule( + when: T::BlockNumber, + maybe_periodic: Option>, + priority: schedule::Priority, + call: ::Call + ) -> Self::Address { + // sanitize maybe_periodic + let maybe_periodic = maybe_periodic + .filter(|p| p.1 > 1 && !p.0.is_zero()) + // Remove one from the number of repetitions since we will schedule one now. + .map(|(p, c)| (p, c - 1)); + let s = Some(Scheduled { maybe_id: None, priority, call, maybe_periodic }); + Agenda::::append_or_insert(when, &[s][..]); + (when, Agenda::::decode_len(when).unwrap_or(1) as u32 - 1) + } + + fn cancel((when, index): Self::Address) -> Result<(), ()> { + if let Some(s) = Agenda::::mutate(when, |agenda| agenda.get_mut(index as usize).and_then(Option::take)) { + if let Some(id) = s.maybe_id { + Lookup::::remove(id) + } + Ok(()) + } else { + Err(()) + } + } +} + +impl schedule::Named::Call> for Module { + type Address = TaskAddress; + + fn schedule_named( + id: Id, + when: T::BlockNumber, + maybe_periodic: Option>, + priority: schedule::Priority, + call: ::Call, + ) -> Result { + // determine id and ensure it is unique + let id = id.encode(); + if Lookup::::contains_key(&id) { + return Err(()) + } + + // sanitize maybe_periodic + let maybe_periodic = maybe_periodic + .filter(|p| p.1 > 1 && !p.0.is_zero()) + // Remove one from the number of repetitions since we will schedule one now. + .map(|(p, c)| (p, c - 1)); + + let s = Scheduled { maybe_id: Some(id.clone()), priority, call, maybe_periodic }; + Agenda::::append_or_insert(when, &[Some(s)][..]); + let index = Agenda::::decode_len(when).unwrap_or(1) as u32 - 1; + let address = (when, index); + Lookup::::insert(&id, &address); + Ok(address) + } + + fn cancel_named(id: Id) -> Result<(), ()> { + if let Some((when, index)) = id.using_encoded(|d| Lookup::::take(d)) { + let i = index as usize; + Agenda::::mutate(when, |agenda| if let Some(s) = agenda.get_mut(i) { *s = None }); + Ok(()) + } else { + Err(()) + } + } +} + +#[cfg(test)] +mod tests { + use super::*; + + use frame_support::{ + impl_outer_event, impl_outer_origin, impl_outer_dispatch, parameter_types, assert_ok, + traits::{OnInitialize, OnFinalize, schedule::{Anon, Named}}, + weights::{DispatchClass, FunctionOf} + }; + use sp_core::H256; + // The testing primitives are very useful for avoiding having to work with signatures + // or public keys. `u64` is used as the `AccountId` and no `Signature`s are required. + use sp_runtime::{ + Perbill, + testing::Header, + traits::{BlakeTwo256, IdentityLookup}, + }; + use crate as scheduler; + + mod logger { + use super::*; + use std::cell::RefCell; + use frame_system::ensure_root; + + thread_local! { + static LOG: RefCell> = RefCell::new(Vec::new()); + } + pub fn log() -> Vec { + LOG.with(|log| log.borrow().clone()) + } + pub trait Trait: system::Trait { + type Event: From + Into<::Event>; + } + decl_storage! { + trait Store for Module as Logger { + } + } + decl_event! { + pub enum Event { + Logged(u32, Weight), + } + } + decl_module! { + // Simple declaration of the `Module` type. Lets the macro know what its working on. + pub struct Module for enum Call where origin: ::Origin { + fn deposit_event() = default; + + #[weight = FunctionOf( + |args: (&u32, &Weight)| *args.1, + |_: (&u32, &Weight)| DispatchClass::Normal, + true + )] + fn log(origin, i: u32, weight: Weight) { + ensure_root(origin)?; + Self::deposit_event(Event::Logged(i, weight)); + LOG.with(|log| { + log.borrow_mut().push(i); + }) + } + } + } + } + + impl_outer_origin! { + pub enum Origin for Test where system = frame_system {} + } + + impl_outer_dispatch! { + pub enum Call for Test where origin: Origin { + system::System, + logger::Logger, + } + } + + impl_outer_event! { + pub enum Event for Test { + system, + logger, + scheduler, + } + } + // For testing the pallet, we construct most of a mock runtime. This means + // first constructing a configuration type (`Test`) which `impl`s each of the + // configuration traits of pallets we want to use. + #[derive(Clone, Eq, PartialEq)] + pub struct Test; + parameter_types! { + pub const BlockHashCount: u64 = 250; + pub const MaximumBlockWeight: Weight = 1024; + pub const MaximumBlockLength: u32 = 2 * 1024; + pub const AvailableBlockRatio: Perbill = Perbill::one(); + } + impl system::Trait for Test { + type Origin = Origin; + type Call = (); + type Index = u64; + type BlockNumber = u64; + type Hash = H256; + type Hashing = BlakeTwo256; + type AccountId = u64; + type Lookup = IdentityLookup; + type Header = Header; + type Event = (); + type BlockHashCount = BlockHashCount; + type MaximumBlockWeight = MaximumBlockWeight; + type MaximumBlockLength = MaximumBlockLength; + type AvailableBlockRatio = AvailableBlockRatio; + type Version = (); + type ModuleToIndex = (); + type AccountData = (); + type OnNewAccount = (); + type OnKilledAccount = (); + } + impl logger::Trait for Test { + type Event = (); + } + parameter_types! { + pub const MaximumWeight: Weight = 10_000; + } + impl Trait for Test { + type Event = (); + type Origin = Origin; + type Call = Call; + type MaximumWeight = MaximumWeight; + } + type System = system::Module; + type Logger = logger::Module; + type Scheduler = Module; + + // This function basically just builds a genesis storage key/value store according to + // our desired mockup. + fn new_test_ext() -> sp_io::TestExternalities { + let t = system::GenesisConfig::default().build_storage::().unwrap(); + t.into() + } + + fn run_to_block(n: u64) { + while System::block_number() < n { + Scheduler::on_finalize(System::block_number()); + System::set_block_number(System::block_number() + 1); + Scheduler::on_initialize(System::block_number()); + } + } + + #[test] + fn basic_scheduling_works() { + new_test_ext().execute_with(|| { + Scheduler::schedule(4, None, 127, Call::Logger(logger::Call::log(42, 1000))); + run_to_block(3); + assert!(logger::log().is_empty()); + run_to_block(4); + assert_eq!(logger::log(), vec![42u32]); + run_to_block(100); + assert_eq!(logger::log(), vec![42u32]); + }); + } + + #[test] + fn periodic_scheduling_works() { + new_test_ext().execute_with(|| { + // at #4, every 3 blocks, 3 times. + Scheduler::schedule(4, Some((3, 3)), 127, Call::Logger(logger::Call::log(42, 1000))); + run_to_block(3); + assert!(logger::log().is_empty()); + run_to_block(4); + assert_eq!(logger::log(), vec![42u32]); + run_to_block(6); + assert_eq!(logger::log(), vec![42u32]); + run_to_block(7); + assert_eq!(logger::log(), vec![42u32, 42u32]); + run_to_block(9); + assert_eq!(logger::log(), vec![42u32, 42u32]); + run_to_block(10); + assert_eq!(logger::log(), vec![42u32, 42u32, 42u32]); + run_to_block(100); + assert_eq!(logger::log(), vec![42u32, 42u32, 42u32]); + }); + } + + #[test] + fn cancel_named_scheduling_works_with_normal_cancel() { + new_test_ext().execute_with(|| { + // at #4, every 3 blocks, 3 times. + Scheduler::schedule_named(1u32, 4, None, 127, Call::Logger(logger::Call::log(69, 1000))).unwrap(); + let i = Scheduler::schedule(4, None, 127, Call::Logger(logger::Call::log(42, 1000))); + run_to_block(3); + assert!(logger::log().is_empty()); + assert_ok!(Scheduler::cancel_named(1u32)); + assert_ok!(Scheduler::cancel(i)); + run_to_block(100); + assert!(logger::log().is_empty()); + }); + } + + #[test] + fn cancel_named_periodic_scheduling_works() { + new_test_ext().execute_with(|| { + // at #4, every 3 blocks, 3 times. + Scheduler::schedule_named(1u32, 4, Some((3, 3)), 127, Call::Logger(logger::Call::log(42, 1000))).unwrap(); + // same id results in error. + assert!(Scheduler::schedule_named(1u32, 4, None, 127, Call::Logger(logger::Call::log(69, 1000))).is_err()); + // different id is ok. + Scheduler::schedule_named(2u32, 8, None, 127, Call::Logger(logger::Call::log(69, 1000))).unwrap(); + run_to_block(3); + assert!(logger::log().is_empty()); + run_to_block(4); + assert_eq!(logger::log(), vec![42u32]); + run_to_block(6); + assert_ok!(Scheduler::cancel_named(1u32)); + run_to_block(100); + assert_eq!(logger::log(), vec![42u32, 69u32]); + }); + } + + #[test] + fn scheduler_respects_weight_limits() { + new_test_ext().execute_with(|| { + Scheduler::schedule(4, None, 127, Call::Logger(logger::Call::log(42, 6000))); + Scheduler::schedule(4, None, 127, Call::Logger(logger::Call::log(69, 6000))); + run_to_block(4); + assert_eq!(logger::log(), vec![42u32]); + run_to_block(5); + assert_eq!(logger::log(), vec![42u32, 69u32]); + }); + } + + #[test] + fn scheduler_respects_hard_deadlines_more() { + new_test_ext().execute_with(|| { + Scheduler::schedule(4, None, 0, Call::Logger(logger::Call::log(42, 6000))); + Scheduler::schedule(4, None, 0, Call::Logger(logger::Call::log(69, 6000))); + run_to_block(4); + assert_eq!(logger::log(), vec![42u32, 69u32]); + }); + } + + #[test] + fn scheduler_respects_priority_ordering() { + new_test_ext().execute_with(|| { + Scheduler::schedule(4, None, 1, Call::Logger(logger::Call::log(42, 6000))); + Scheduler::schedule(4, None, 0, Call::Logger(logger::Call::log(69, 6000))); + run_to_block(4); + assert_eq!(logger::log(), vec![69u32, 42u32]); + }); + } + + #[test] + fn scheduler_respects_priority_ordering_with_soft_deadlines() { + new_test_ext().execute_with(|| { + Scheduler::schedule(4, None, 255, Call::Logger(logger::Call::log(42, 5000))); + Scheduler::schedule(4, None, 127, Call::Logger(logger::Call::log(69, 5000))); + Scheduler::schedule(4, None, 126, Call::Logger(logger::Call::log(2600, 6000))); + run_to_block(4); + assert_eq!(logger::log(), vec![2600u32]); + run_to_block(5); + assert_eq!(logger::log(), vec![2600u32, 69u32, 42u32]); + }); + } + + #[test] + fn initialize_weight_is_correct() { + new_test_ext().execute_with(|| { + Scheduler::schedule(1, None, 255, Call::Logger(logger::Call::log(3, 1000))); + Scheduler::schedule(1, None, 128, Call::Logger(logger::Call::log(42, 5000))); + Scheduler::schedule(1, None, 127, Call::Logger(logger::Call::log(69, 5000))); + Scheduler::schedule(1, None, 126, Call::Logger(logger::Call::log(2600, 6000))); + let weight = Scheduler::on_initialize(1); + assert_eq!(weight, 6000); + let weight = Scheduler::on_initialize(2); + assert_eq!(weight, 10000); + let weight = Scheduler::on_initialize(3); + assert_eq!(weight, 1000); + let weight = Scheduler::on_initialize(4); + assert_eq!(weight, 0); + }); + } +} diff --git a/frame/support/src/traits.rs b/frame/support/src/traits.rs index bd1534bac506e..39b1c7571d184 100644 --- a/frame/support/src/traits.rs +++ b/frame/support/src/traits.rs @@ -19,7 +19,7 @@ //! NOTE: If you're looking for `parameter_types`, it has moved in to the top-level module. use sp_std::{prelude::*, result, marker::PhantomData, ops::Div, fmt::Debug}; -use codec::{FullCodec, Codec, Encode, Decode}; +use codec::{FullCodec, Codec, Encode, Decode, EncodeLike}; use sp_core::u32_trait::Value as U32; use sp_runtime::{ RuntimeDebug, @@ -1109,6 +1109,84 @@ pub trait OffchainWorker { fn offchain_worker(_n: BlockNumber) {} } +pub mod schedule { + use super::*; + + /// Information relating to the period of a scheduled task. First item is the length of the + /// period and the second is the number of times it should be executed in total before the task + /// is considered finished and removed. + pub type Period = (BlockNumber, u32); + + /// Priority with which a call is scheduled. It's just a linear amount with lowest values meaning + /// higher priority. + pub type Priority = u8; + + /// The highest priority. We invert the value so that normal sorting will place the highest + /// priority at the beginning of the list. + pub const HIGHEST_PRORITY: Priority = 0; + /// Anything of this value or lower will definitely be scheduled on the block that they ask for, even + /// if it breaches the `MaximumWeight` limitation. + pub const HARD_DEADLINE: Priority = 63; + /// The lowest priority. Most stuff should be around here. + pub const LOWEST_PRORITY: Priority = 255; + + /// A type that can be used as a scheduler. + pub trait Anon { + /// An address which can be used for removing a scheduled task. + type Address: Codec + Clone + Eq + EncodeLike + Debug; + + /// Schedule a one-off dispatch to happen at the beginning of some block in the future. + /// + /// This is not named. + /// + /// Infallible. + fn schedule( + when: BlockNumber, + maybe_periodic: Option>, + priority: Priority, + call: Call + ) -> Self::Address; + + /// Cancel a scheduled task. If periodic, then it will cancel all further instances of that, + /// also. + /// + /// Will return an error if the `address` is invalid. + /// + /// NOTE: This guaranteed to work only *before* the point that it is due to be executed. + /// If it ends up being delayed beyond the point of execution, then it cannot be cancelled. + /// + /// NOTE2: This will not work to cancel periodic tasks after their initial execution. For + /// that, you must name the task explicitly using the `Named` trait. + fn cancel(address: Self::Address) -> Result<(), ()>; + } + + /// A type that can be used as a scheduler. + pub trait Named { + /// An address which can be used for removing a scheduled task. + type Address: Codec + Clone + Eq + EncodeLike + sp_std::fmt::Debug; + + /// Schedule a one-off dispatch to happen at the beginning of some block in the future. + /// + /// - `id`: The identity of the task. This must be unique and will return an error if not. + fn schedule_named( + id: Id, + when: BlockNumber, + maybe_periodic: Option>, + priority: Priority, + call: Call + ) -> Result; + + /// Cancel a scheduled, named task. If periodic, then it will cancel all further instances + /// of that, also. + /// + /// Will return an error if the `id` is invalid. + /// + /// NOTE: This guaranteed to work only *before* the point that it is due to be executed. + /// If it ends up being delayed beyond the point of execution, then it cannot be cancelled. + fn cancel_named(id: Id) -> Result<(), ()>; + } +} + #[cfg(test)] mod tests { use super::*;