[bdk_chain_redesign] Consistent ChainOracle

The problem with the previous `ChainOracle` interface is that it had no
guarantee for consistency. For example, a block deemed to be part of the
"best chain" can be reorged out. So when `ChainOracle` is called
multiple times for an operation (such as getting the UTXO set), the
returned result may be inconsistent.

This PR changes `ChainOracle::is_block_in_chain` to take in another
input `static_block`, ensuring `block` is an ancestor of `static_block`.
Thus, if `static_block` is consistent across the operation, the result
will be consistent also.

`is_block_in_chain` now returns `Option<bool>`. The `None` case means
that the oracle implementation cannot determine whether block is an
ancestor of static block. `IndexedTxGraph::list_chain_txouts` handles
this case by checking child spends that are in chain, and if so, the
parent tx must be in chain too.

Author: evanlinjin

crates/chain/src/chain_data.rs

@@ -248,7 +248,7 @@ impl<A: BlockAnchor> FullTxOut<ObservedAs<A>> {
     /// [`ObservedAs<A>`] where `A` implements [`BlockAnchor`].
     ///
     /// [`is_mature`]: Self::is_mature
-    pub fn is_observed_as_mature(&self, tip: u32) -> bool {
+    pub fn is_observed_as_confirmed_and_mature(&self, tip: u32) -> bool {
         if !self.is_on_coinbase {
             return false;
         }
@@ -275,8 +275,8 @@ impl<A: BlockAnchor> FullTxOut<ObservedAs<A>> {
     /// being a [`ObservedAs<A>`] where `A` implements [`BlockAnchor`].
     ///
     /// [`is_spendable_at`]: Self::is_spendable_at
-    pub fn is_observed_as_spendable(&self, tip: u32) -> bool {
-        if !self.is_observed_as_mature(tip) {
+    pub fn is_observed_as_confirmed_and_spendable(&self, tip: u32) -> bool {
+        if !self.is_observed_as_confirmed_and_mature(tip) {
             return false;
         }
 

crates/chain/src/chain_oracle.rs

@@ -1,162 +1,77 @@
-use core::{convert::Infallible, marker::PhantomData};
+use crate::collections::HashSet;
+use core::marker::PhantomData;
 
-use alloc::collections::BTreeMap;
+use alloc::{collections::VecDeque, vec::Vec};
 use bitcoin::BlockHash;
 
 use crate::BlockId;
 
-/// Represents a service that tracks the best chain history.
-/// TODO: How do we ensure the chain oracle is consistent across a single call?
-/// * We need to somehow lock the data! What if the ChainOracle is remote?
-/// * Get tip method! And check the tip still exists at the end! And every internal call
-///   does not go beyond the initial tip.
+/// Represents a service that tracks the blockchain.
+///
+/// The main method is [`is_block_in_chain`] which determines whether a given block of [`BlockId`]
+/// is an ancestor of another "static block".
+///
+/// [`is_block_in_chain`]: Self::is_block_in_chain
 pub trait ChainOracle {
     /// Error type.
     type Error: core::fmt::Debug;
 
-    /// Get the height and hash of the tip in the best chain.
-    fn get_tip_in_best_chain(&self) -> Result<Option<BlockId>, Self::Error>;
-
-    /// Returns the block hash (if any) of the given `height`.
-    fn get_block_in_best_chain(&self, height: u32) -> Result<Option<BlockHash>, Self::Error>;
-
-    /// Determines whether the block of [`BlockId`] exists in the best chain.
-    fn is_block_in_best_chain(&self, block_id: BlockId) -> Result<bool, Self::Error> {
-        Ok(matches!(self.get_block_in_best_chain(block_id.height)?, Some(h) if h == block_id.hash))
-    }
-}
-
-// [TODO] We need stuff for smart pointers. Maybe? How does rust lib do this?
-// Box<dyn ChainOracle>, Arc<dyn ChainOracle> ????? I will figure it out
-impl<C: ChainOracle> ChainOracle for &C {
-    type Error = C::Error;
-
-    fn get_tip_in_best_chain(&self) -> Result<Option<BlockId>, Self::Error> {
-        <C as ChainOracle>::get_tip_in_best_chain(self)
-    }
-
-    fn get_block_in_best_chain(&self, height: u32) -> Result<Option<BlockHash>, Self::Error> {
-        <C as ChainOracle>::get_block_in_best_chain(self, height)
-    }
-
-    fn is_block_in_best_chain(&self, block_id: BlockId) -> Result<bool, Self::Error> {
-        <C as ChainOracle>::is_block_in_best_chain(self, block_id)
-    }
+    /// Determines whether `block` of [`BlockId`] exists as an ancestor of `static_block`.
+    ///
+    /// If `None` is returned, it means the implementation cannot determine whether `block` exists.
+    fn is_block_in_chain(
+        &self,
+        block: BlockId,
+        static_block: BlockId,
+    ) -> Result<Option<bool>, Self::Error>;
 }
 
-/// This structure increases the performance of getting chain data.
-#[derive(Debug)]
-pub struct Cache<C> {
-    assume_final_depth: u32,
-    tip_height: u32,
-    cache: BTreeMap<u32, BlockHash>,
+/// A cache structure increases the performance of getting chain data.
+///
+/// A simple FIFO cache replacement policy is used. Something more efficient and advanced can be
+/// implemented later.
+#[derive(Debug, Default)]
+pub struct CacheBackend<C> {
+    cache: HashSet<(BlockHash, BlockHash)>,
+    fifo: VecDeque<(BlockHash, BlockHash)>,
     marker: PhantomData<C>,
 }
 
-impl<C> Cache<C> {
-    /// Creates a new [`Cache`].
-    ///
-    /// `assume_final_depth` represents the minimum number of blocks above the block in question
-    /// when we can assume the block is final (reorgs cannot happen). I.e. a value of 0 means the
-    /// tip is assumed to be final. The cache only caches blocks that are assumed to be final.
-    pub fn new(assume_final_depth: u32) -> Self {
-        Self {
-            assume_final_depth,
-            tip_height: 0,
-            cache: Default::default(),
-            marker: Default::default(),
-        }
-    }
-}
-
-impl<C: ChainOracle> Cache<C> {
-    /// This is the topmost (highest) block height that we assume as final (no reorgs possible).
-    ///
-    /// Blocks higher than this height are not cached.
-    pub fn assume_final_height(&self) -> u32 {
-        self.tip_height.saturating_sub(self.assume_final_depth)
+impl<C> CacheBackend<C> {
+    /// Get the number of elements in the cache.
+    pub fn cache_size(&self) -> usize {
+        self.cache.len()
     }
 
-    /// Update the `tip_height` with the [`ChainOracle`]'s tip.
+    /// Prunes the cache to reach the `max_size` target.
     ///
-    /// `tip_height` is used with `assume_final_depth` to determine whether we should cache a
-    /// certain block height (`tip_height` - `assume_final_depth`).
-    pub fn try_update_tip_height(&mut self, chain: C) -> Result<(), C::Error> {
-        let tip = chain.get_tip_in_best_chain()?;
-        if let Some(BlockId { height, .. }) = tip {
-            self.tip_height = height;
-        }
-        Ok(())
+    /// Returns pruned elements.
+    pub fn prune(&mut self, max_size: usize) -> Vec<(BlockHash, BlockHash)> {
+        let prune_count = self.cache.len().saturating_sub(max_size);
+        (0..prune_count)
+            .filter_map(|_| self.fifo.pop_front())
+            .filter(|k| self.cache.remove(k))
+            .collect()
     }
 
-    /// Get a block from the cache with the [`ChainOracle`] as fallback.
-    ///
-    /// If the block does not exist in cache, the logic fallbacks to fetching from the internal
-    /// [`ChainOracle`]. If the block is at or below the "assume final height", we will also store
-    /// the missing block in the cache.
-    pub fn try_get_block(&mut self, chain: C, height: u32) -> Result<Option<BlockHash>, C::Error> {
-        if let Some(&hash) = self.cache.get(&height) {
-            return Ok(Some(hash));
+    pub fn contains(&self, static_block: BlockId, block: BlockId) -> bool {
+        if static_block.height < block.height
+            || static_block.height == block.height && static_block.hash != block.hash
+        {
+            return false;
         }
 
-        let hash = chain.get_block_in_best_chain(height)?;
-
-        if hash.is_some() && height > self.tip_height {
-            self.tip_height = height;
-        }
-
-        // only cache block if at least as deep as `assume_final_depth`
-        let assume_final_height = self.tip_height.saturating_sub(self.assume_final_depth);
-        if height <= assume_final_height {
-            if let Some(hash) = hash {
-                self.cache.insert(height, hash);
-            }
-        }
-
-        Ok(hash)
-    }
-
-    /// Determines whether the block of `block_id` is in the chain using the cache.
-    ///
-    /// This uses [`try_get_block`] internally.
-    ///
-    /// [`try_get_block`]: Self::try_get_block
-    pub fn try_is_block_in_chain(&mut self, chain: C, block_id: BlockId) -> Result<bool, C::Error> {
-        match self.try_get_block(chain, block_id.height)? {
-            Some(hash) if hash == block_id.hash => Ok(true),
-            _ => Ok(false),
-        }
-    }
-}
-
-impl<C: ChainOracle<Error = Infallible>> Cache<C> {
-    /// Updates the `tip_height` with the [`ChainOracle`]'s tip.
-    ///
-    /// This is the no-error version of [`try_update_tip_height`].
-    ///
-    /// [`try_update_tip_height`]: Self::try_update_tip_height
-    pub fn update_tip_height(&mut self, chain: C) {
-        self.try_update_tip_height(chain)
-            .expect("chain oracle error is infallible")
+        self.cache.contains(&(static_block.hash, block.hash))
     }
 
-    /// Get a block from the cache with the [`ChainOracle`] as fallback.
-    ///
-    /// This is the no-error version of [`try_get_block`].
-    ///
-    /// [`try_get_block`]: Self::try_get_block
-    pub fn get_block(&mut self, chain: C, height: u32) -> Option<BlockHash> {
-        self.try_get_block(chain, height)
-            .expect("chain oracle error is infallible")
-    }
+    pub fn insert(&mut self, static_block: BlockId, block: BlockId) -> bool {
+        let cache_key = (static_block.hash, block.hash);
 
-    /// Determines whether the block at `block_id` is in the chain using the cache.
-    ///
-    /// This is the no-error version of [`try_is_block_in_chain`].
-    ///
-    /// [`try_is_block_in_chain`]: Self::try_is_block_in_chain
-    pub fn is_block_in_best_chain(&mut self, chain: C, block_id: BlockId) -> bool {
-        self.try_is_block_in_chain(chain, block_id)
-            .expect("chain oracle error is infallible")
+        if self.cache.insert(cache_key) {
+            self.fifo.push_back(cache_key);
+            true
+        } else {
+            false
+        }
     }
 }