Introduce XCM matcher for writing barriers (#6756)

* Introduce XCM matcher for writing barriers

* Fix compilation

* cargo fmt

* Add more doc comments

* Add mod doc comment

* More doc comments

* Add tests and fix logic

* Remove redundant syntax

* Add more doc comments

* Add more doc comments

* Add more doc comments

polkadot/xcm/src/lib.rs

@@ -23,6 +23,7 @@
 #![no_std]
 extern crate alloc;
 
+use core::ops::ControlFlow;
 use derivative::Derivative;
 use parity_scale_codec::{Decode, Encode, Error as CodecError, Input, MaxEncodedLen};
 use scale_info::TypeInfo;
@@ -47,6 +48,89 @@ pub const MAX_XCM_DECODE_DEPTH: u32 = 8;
 /// A version of XCM.
 pub type Version = u32;
 
+/// Creates an instruction matcher from an XCM. Since XCM versions differ, we need to make a trait
+/// here to unify the interfaces among them.
+pub trait CreateMatcher {
+	/// The concrete matcher type.
+	type Matcher;
+
+	/// Method that creates and returns the matcher type from `Self`.
+	fn matcher(self) -> Self::Matcher;
+}
+
+/// API that allows to pattern-match against anything that is contained within an XCM.
+///
+/// The intended usage of the matcher API is to enable the ability to chain successive methods of
+/// this trait together, along with the ? operator for the purpose of facilitating the writing,
+/// maintenance and auditability of XCM barriers.
+///
+/// Example:
+/// ```rust
+/// use xcm::{
+/// 	v3::{Instruction, Matcher},
+/// 	CreateMatcher, MatchXcm,
+/// };
+///
+/// let mut msg = [Instruction::<()>::ClearOrigin];
+/// let res = msg
+/// 	.matcher()
+/// 	.assert_remaining_insts(1)?
+/// 	.match_next_inst(|inst| match inst {
+/// 		Instruction::<()>::ClearOrigin => Ok(()),
+/// 		_ => Err(()),
+/// 	});
+/// assert!(res.is_ok());
+///
+/// Ok::<(), ()>(())
+/// ```
+pub trait MatchXcm {
+	/// The concrete instruction type. Necessary to specify as it changes between XCM versions.
+	type Inst;
+	/// The `MultiLocation` type. Necessary to specify as it changes between XCM versions.
+	type Loc;
+	/// The error type to throw when errors happen during matching.
+	type Error;
+
+	/// Returns success if the number of instructions that still have not been iterated over
+	/// equals `n`, otherwise returns an error.
+	fn assert_remaining_insts(self, n: usize) -> Result<Self, Self::Error>
+	where
+		Self: Sized;
+
+	/// Accepts a closure `f` that contains an argument signifying the next instruction to be
+	/// iterated over. The closure can then be used to check whether the instruction matches a
+	/// given condition, and can also be used to mutate the fields of an instruction.
+	///
+	/// The closure `f` returns success when the instruction passes the condition, otherwise it
+	/// returns an error, which will ultimately be returned by this function.
+	fn match_next_inst<F>(self, f: F) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+		F: FnMut(&mut Self::Inst) -> Result<(), Self::Error>;
+
+	/// Attempts to continuously iterate through the instructions while applying `f` to each of
+	/// them, until either the last instruction or `cond` returns false.
+	///
+	/// If `f` returns an error, then iteration halts and the function returns that error.
+	/// Otherwise, `f` returns a `ControlFlow` which signifies whether the iteration breaks or
+	/// continues.
+	fn match_next_inst_while<C, F>(self, cond: C, f: F) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+		C: Fn(&Self::Inst) -> bool,
+		F: FnMut(&mut Self::Inst) -> Result<ControlFlow<()>, Self::Error>;
+
+	/// Iterate instructions forward until `cond` returns false. When there are no more instructions
+	/// to be read, an error is returned.
+	fn skip_inst_while<C>(self, cond: C) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+		C: Fn(&Self::Inst) -> bool,
+	{
+		Self::match_next_inst_while(self, cond, |_| Ok(ControlFlow::Continue(())))
+	}
+}
+
 #[derive(Clone, Eq, PartialEq, Debug)]
 pub enum Unsupported {}
 impl Encode for Unsupported {}

polkadot/xcm/src/v3/matcher.rs

@@ -0,0 +1,109 @@
+// Copyright 2023 Parity Technologies (UK) Ltd.
+// This file is part of Polkadot.
+
+// 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 Polkadot.  If not, see <http://www.gnu.org/licenses/>.
+
+//! XCM matcher API, used primarily for writing barrier conditions.
+
+use super::{Instruction, MultiLocation};
+use crate::{CreateMatcher, MatchXcm};
+use core::ops::ControlFlow;
+
+impl<'a, Call> CreateMatcher for &'a mut [Instruction<Call>] {
+	type Matcher = Matcher<'a, Call>;
+
+	fn matcher(self) -> Self::Matcher {
+		let total_inst = self.len();
+
+		Matcher { xcm: self, current_idx: 0, total_inst }
+	}
+}
+
+/// Struct created from calling `fn matcher()` on a mutable slice of `Instruction`s.
+///
+/// Implements `MatchXcm` to allow an iterator-like API to match against each `Instruction`
+/// contained within the slice, which facilitates the building of XCM barriers.
+pub struct Matcher<'a, Call> {
+	pub(crate) xcm: &'a mut [Instruction<Call>],
+	pub(crate) current_idx: usize,
+	pub(crate) total_inst: usize,
+}
+
+impl<'a, Call> MatchXcm for Matcher<'a, Call> {
+	type Error = ();
+	type Inst = Instruction<Call>;
+	type Loc = MultiLocation;
+
+	fn assert_remaining_insts(self, n: usize) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+	{
+		if self.total_inst - self.current_idx != n {
+			return Err(())
+		}
+
+		Ok(self)
+	}
+
+	fn match_next_inst<F>(mut self, mut f: F) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+		F: FnMut(&mut Self::Inst) -> Result<(), Self::Error>,
+	{
+		if self.current_idx < self.total_inst {
+			f(&mut self.xcm[self.current_idx])?;
+			self.current_idx += 1;
+			Ok(self)
+		} else {
+			Err(())
+		}
+	}
+
+	fn match_next_inst_while<C, F>(mut self, cond: C, mut f: F) -> Result<Self, Self::Error>
+	where
+		Self: Sized,
+		C: Fn(&Self::Inst) -> bool,
+		F: FnMut(&mut Self::Inst) -> Result<ControlFlow<()>, Self::Error>,
+	{
+		if self.current_idx >= self.total_inst {
+			return Err(())
+		}
+
+		while self.current_idx < self.total_inst && cond(&self.xcm[self.current_idx]) {
+			if let ControlFlow::Break(()) = f(&mut self.xcm[self.current_idx])? {
+				break
+			}
+			self.current_idx += 1;
+		}
+
+		Ok(self)
+	}
+}
+
+#[cfg(test)]
+mod tests {
+	use super::*;
+	use crate::v3::prelude::*;
+	use alloc::{vec, vec::Vec};
+
+	#[test]
+	fn match_next_inst_while_works() {
+		let mut xcm: Vec<Instruction<()>> = vec![ClearOrigin];
+
+		let _ = xcm
+			.matcher()
+			.match_next_inst_while(|_| true, |_| Ok(ControlFlow::Continue(())))
+			.unwrap();
+	}
+}

polkadot/xcm/src/v3/mod.rs

@@ -1,5 +1,5 @@
 // Copyright 2020 Parity Technologies (UK) Ltd.
-// This file is part of Cumulus.
+// This file is part of Polkadot.
 
 // Substrate is free software: you can redistribute it and/or modify
 // it under the terms of the GNU General Public License as published by
@@ -12,7 +12,7 @@
 // GNU General Public License for more details.
 
 // You should have received a copy of the GNU General Public License
-// along with Cumulus.  If not, see <http://www.gnu.org/licenses/>.
+// along with Polkadot.  If not, see <http://www.gnu.org/licenses/>.
 
 //! Version 3 of the Cross-Consensus Message format data structures.
 
@@ -34,12 +34,14 @@ use scale_info::TypeInfo;
 
 mod junction;
 pub(crate) mod junctions;
+mod matcher;
 mod multiasset;
 mod multilocation;
 mod traits;
 
 pub use junction::{BodyId, BodyPart, Junction, NetworkId};
 pub use junctions::Junctions;
+pub use matcher::Matcher;
 pub use multiasset::{
 	AssetId, AssetInstance, Fungibility, MultiAsset, MultiAssetFilter, MultiAssets,
 	WildFungibility, WildMultiAsset,

polkadot/xcm/src/v3/traits.rs

@@ -1,5 +1,5 @@
 // Copyright 2020 Parity Technologies (UK) Ltd.
-// This file is part of Cumulus.
+// This file is part of Polkadot.
 
 // Substrate is free software: you can redistribute it and/or modify
 // it under the terms of the GNU General Public License as published by
@@ -12,7 +12,7 @@
 // GNU General Public License for more details.
 
 // You should have received a copy of the GNU General Public License
-// along with Cumulus.  If not, see <http://www.gnu.org/licenses/>.
+// along with Polkadot.  If not, see <http://www.gnu.org/licenses/>.
 
 //! Cross-Consensus Message format data structures.
 

polkadot/xcm/xcm-builder/src/barriers.rs

@@ -21,13 +21,16 @@ use frame_support::{
 	traits::{Contains, Get},
 };
 use polkadot_parachain::primitives::IsSystem;
-use sp_std::{marker::PhantomData, result::Result};
-use xcm::latest::{
-	Instruction::{self, *},
-	InteriorMultiLocation, Junction, Junctions,
-	Junctions::X1,
-	MultiLocation, Weight,
-	WeightLimit::*,
+use sp_std::{cell::Cell, marker::PhantomData, ops::ControlFlow, result::Result};
+use xcm::{
+	latest::{
+		Instruction::{self, *},
+		InteriorMultiLocation, Junction, Junctions,
+		Junctions::X1,
+		MultiLocation, Weight,
+		WeightLimit::*,
+	},
+	CreateMatcher, MatchXcm,
 };
 use xcm_executor::traits::{OnResponse, ShouldExecute};
 
@@ -77,32 +80,31 @@ impl<T: Contains<MultiLocation>> ShouldExecute for AllowTopLevelPaidExecutionFro
 		// We will read up to 5 instructions. This allows up to 3 `ClearOrigin` instructions. We
 		// allow for more than one since anything beyond the first is a no-op and it's conceivable
 		// that composition of operations might result in more than one being appended.
-		let mut iter = instructions.iter_mut().take(5);
-		let i = iter.next().ok_or(())?;
-		match i {
-			ReceiveTeleportedAsset(..) |
-			WithdrawAsset(..) |
-			ReserveAssetDeposited(..) |
-			ClaimAsset { .. } => (),
-			_ => return Err(()),
-		}
-		let mut i = iter.next().ok_or(())?;
-		while let ClearOrigin = i {
-			i = iter.next().ok_or(())?;
-		}
-		match i {
-			BuyExecution { weight_limit: Limited(ref mut weight), .. }
-				if weight.all_gte(max_weight) =>
-			{
-				*weight = max_weight;
-				Ok(())
-			},
-			BuyExecution { ref mut weight_limit, .. } if weight_limit == &Unlimited => {
-				*weight_limit = Limited(max_weight);
-				Ok(())
-			},
-			_ => Err(()),
-		}
+		let end = instructions.len().min(5);
+		instructions[..end]
+			.matcher()
+			.match_next_inst(|inst| match inst {
+				ReceiveTeleportedAsset(..) |
+				WithdrawAsset(..) |
+				ReserveAssetDeposited(..) |
+				ClaimAsset { .. } => Ok(()),
+				_ => Err(()),
+			})?
+			.skip_inst_while(|inst| matches!(inst, ClearOrigin))?
+			.match_next_inst(|inst| match inst {
+				BuyExecution { weight_limit: Limited(ref mut weight), .. }
+					if weight.all_gte(max_weight) =>
+				{
+					*weight = max_weight;
+					Ok(())
+				},
+				BuyExecution { ref mut weight_limit, .. } if weight_limit == &Unlimited => {
+					*weight_limit = Limited(max_weight);
+					Ok(())
+				},
+				_ => Err(()),
+			})?;
+		Ok(())
 	}
 }
 
@@ -172,29 +174,34 @@ impl<
 			origin, instructions, max_weight, weight_credit,
 		);
 		let mut actual_origin = *origin;
-		let mut skipped = 0;
+		let skipped = Cell::new(0usize);
 		// NOTE: We do not check the validity of `UniversalOrigin` here, meaning that a malicious
 		// origin could place a `UniversalOrigin` in order to spoof some location which gets free
 		// execution. This technical could get it past the barrier condition, but the execution
 		// would instantly fail since the first instruction would cause an error with the
 		// invalid UniversalOrigin.
-		while skipped < MaxPrefixes::get() as usize {
-			match instructions.get(skipped) {
-				Some(UniversalOrigin(new_global)) => {
-					// Note the origin is *relative to local consensus*! So we need to escape local
-					// consensus with the `parents` before diving in into the `universal_location`.
-					actual_origin = X1(*new_global).relative_to(&LocalUniversal::get());
-				},
-				Some(DescendOrigin(j)) => {
-					actual_origin.append_with(*j).map_err(|_| ())?;
-				},
-				_ => break,
-			}
-			skipped += 1;
-		}
+		instructions.matcher().match_next_inst_while(
+			|_| skipped.get() < MaxPrefixes::get() as usize,
+			|inst| {
+				match inst {
+					UniversalOrigin(new_global) => {
+						// Note the origin is *relative to local consensus*! So we need to escape
+						// local consensus with the `parents` before diving in into the
+						// `universal_location`.
+						actual_origin = X1(*new_global).relative_to(&LocalUniversal::get());
+					},
+					DescendOrigin(j) => {
+						let Ok(_) = actual_origin.append_with(*j) else { return Err(()) };
+					},
+					_ => return Ok(ControlFlow::Break(())),
+				};
+				skipped.set(skipped.get() + 1);
+				Ok(ControlFlow::Continue(()))
+			},
+		)?;
 		InnerBarrier::should_execute(
 			&actual_origin,
-			&mut instructions[skipped..],
+			&mut instructions[skipped.get()..],
 			max_weight,
 			weight_credit,
 		)
@@ -241,12 +248,12 @@ impl<T: Contains<MultiLocation>> ShouldExecute for AllowExplicitUnpaidExecutionF
 			origin, instructions, max_weight, _weight_credit,
 		);
 		ensure!(T::contains(origin), ());
-		match instructions.first() {
-			Some(UnpaidExecution { weight_limit: Limited(m), .. }) if m.all_gte(max_weight) =>
-				Ok(()),
-			Some(UnpaidExecution { weight_limit: Unlimited, .. }) => Ok(()),
+		instructions.matcher().match_next_inst(|inst| match inst {
+			UnpaidExecution { weight_limit: Limited(m), .. } if m.all_gte(max_weight) => Ok(()),
+			UnpaidExecution { weight_limit: Unlimited, .. } => Ok(()),
 			_ => Err(()),
-		}
+		})?;
+		Ok(())
 	}
 }
 
@@ -276,13 +283,16 @@ impl<ResponseHandler: OnResponse> ShouldExecute for AllowKnownQueryResponses<Res
 			"AllowKnownQueryResponses origin: {:?}, instructions: {:?}, max_weight: {:?}, weight_credit: {:?}",
 			origin, instructions, _max_weight, _weight_credit,
 		);
-		ensure!(instructions.len() == 1, ());
-		match instructions.first() {
-			Some(QueryResponse { query_id, querier, .. })
-				if ResponseHandler::expecting_response(origin, *query_id, querier.as_ref()) =>
-				Ok(()),
-			_ => Err(()),
-		}
+		instructions
+			.matcher()
+			.assert_remaining_insts(1)?
+			.match_next_inst(|inst| match inst {
+				QueryResponse { query_id, querier, .. }
+					if ResponseHandler::expecting_response(origin, *query_id, querier.as_ref()) =>
+					Ok(()),
+				_ => Err(()),
+			})?;
+		Ok(())
 	}
 }
 
@@ -302,9 +312,13 @@ impl<T: Contains<MultiLocation>> ShouldExecute for AllowSubscriptionsFrom<T> {
 			origin, instructions, _max_weight, _weight_credit,
 		);
 		ensure!(T::contains(origin), ());
-		match instructions {
-			&mut [SubscribeVersion { .. } | UnsubscribeVersion] => Ok(()),
-			_ => Err(()),
-		}
+		instructions
+			.matcher()
+			.assert_remaining_insts(1)?
+			.match_next_inst(|inst| match inst {
+				SubscribeVersion { .. } | UnsubscribeVersion => Ok(()),
+				_ => Err(()),
+			})?;
+		Ok(())
 	}
 }