attempt at documenting table crate

src/model/block.rs

@@ -18,7 +18,7 @@ use garage_rpc::membership::System;
 use garage_rpc::rpc_client::*;
 use garage_rpc::rpc_server::*;
 
-use garage_table::replication::{sharded::TableShardedReplication, TableReplication};
+use garage_table::replication::{TableReplication, TableShardedReplication};
 
 use crate::block_ref_table::*;
 

src/model/bucket_table.rs

@@ -100,6 +100,10 @@ impl TableSchema for BucketTable {
 	type E = Bucket;
 	type Filter = DeletedFilter;
 
+	fn updated(&self, _old: Option<Self::E>, _new: Option<Self::E>) {
+		// nothing to do when updated
+	}
+
 	fn matches_filter(entry: &Self::E, filter: &Self::Filter) -> bool {
 		filter.apply(entry.is_deleted())
 	}

src/model/garage.rs

@@ -7,8 +7,8 @@ use garage_rpc::membership::System;
 use garage_rpc::rpc_client::RpcHttpClient;
 use garage_rpc::rpc_server::RpcServer;
 
-use garage_table::replication::fullcopy::*;
-use garage_table::replication::sharded::*;
+use garage_table::replication::TableFullReplication;
+use garage_table::replication::TableShardedReplication;
 use garage_table::*;
 
 use crate::block::*;

src/model/key_table.rs

@@ -113,6 +113,10 @@ impl TableSchema for KeyTable {
 	type E = Key;
 	type Filter = KeyFilter;
 
+	fn updated(&self, _old: Option<Self::E>, _new: Option<Self::E>) {
+		// nothing to do when updated
+	}
+
 	fn matches_filter(entry: &Self::E, filter: &Self::Filter) -> bool {
 		match filter {
 			KeyFilter::Deleted(df) => df.apply(entry.deleted.get()),

src/model/object_table.rs

@@ -6,7 +6,7 @@ use garage_util::background::BackgroundRunner;
 use garage_util::data::*;
 
 use garage_table::crdt::*;
-use garage_table::replication::sharded::*;
+use garage_table::replication::TableShardedReplication;
 use garage_table::*;
 
 use crate::version_table::*;

src/model/version_table.rs

@@ -5,7 +5,7 @@ use garage_util::background::BackgroundRunner;
 use garage_util::data::*;
 
 use garage_table::crdt::*;
-use garage_table::replication::sharded::*;
+use garage_table::replication::TableShardedReplication;
 use garage_table::*;
 
 use crate::block_ref_table::*;

src/table/crdt/lww.rs

@@ -34,7 +34,7 @@ use crate::crdt::crdt::*;
 /// and may differ from what you observed with your atomic clock!
 ///
 /// This scheme is used by AWS S3 or Soundcloud and often without knowing
-/// in entreprise when reconciliating databases with ad-hoc scripts.
+/// in enterprise when reconciliating databases with ad-hoc scripts.
 #[derive(Clone, Debug, Serialize, Deserialize, PartialEq)]
 pub struct LWW<T> {
 	ts: u64,

src/table/crdt/map.rs

@@ -37,6 +37,7 @@ where
 		Self { vals: vec![(k, v)] }
 	}
 
+	/// Add a value to the map
 	pub fn put(&mut self, k: K, v: V) {
 		self.merge(&Self::put_mutator(k, v));
 	}

src/table/gc.rs

@@ -74,7 +74,7 @@ where
 		while !*must_exit.borrow() {
 			match self.gc_loop_iter().await {
 				Ok(true) => {
-					// Stuff was done, loop imediately
+					// Stuff was done, loop immediately
 					continue;
 				}
 				Ok(false) => {

src/table/lib.rs

@@ -1,3 +1,4 @@
+#![warn(missing_docs)]
 #![recursion_limit = "1024"]
 
 #[macro_use]
@@ -8,10 +9,10 @@ pub mod schema;
 pub mod util;
 
 pub mod data;
-pub mod gc;
-pub mod merkle;
+mod gc;
+mod merkle;
 pub mod replication;
-pub mod sync;
+mod sync;
 pub mod table;
 
 pub use schema::*;

src/table/replication/fullcopy.rs

@@ -6,19 +6,19 @@ use garage_util::data::*;
 
 use crate::replication::*;
 
+/// Full replication schema: all nodes store everything
+/// Writes are disseminated in an epidemic manner in the network
+/// Advantage: do all reads locally, extremely fast
+/// Inconvenient: only suitable to reasonably small tables
 #[derive(Clone)]
 pub struct TableFullReplication {
+	/// The membership manager of this node
 	pub system: Arc<System>,
+	/// Max number of faults allowed while replicating a record
 	pub max_faults: usize,
 }
 
 impl TableReplication for TableFullReplication {
-	// Full replication schema: all nodes store everything
-	// Writes are disseminated in an epidemic manner in the network
-
-	// Advantage: do all reads locally, extremely fast
-	// Inconvenient: only suitable to reasonably small tables
-
 	fn read_nodes(&self, _hash: &Hash) -> Vec<UUID> {
 		vec![self.system.id]
 	}

src/table/replication/mod.rs

@@ -1,6 +1,8 @@
 mod parameters;
 
-pub mod fullcopy;
-pub mod sharded;
+mod fullcopy;
+mod sharded;
 
+pub use fullcopy::TableFullReplication;
 pub use parameters::*;
+pub use sharded::TableShardedReplication;

src/table/replication/parameters.rs

@@ -2,20 +2,26 @@ use garage_rpc::ring::*;
 
 use garage_util::data::*;
 
+/// Trait to describe how a table shall be replicated
 pub trait TableReplication: Send + Sync {
 	// See examples in table_sharded.rs and table_fullcopy.rs
 	// To understand various replication methods
 
-	// Which nodes to send reads from
+	/// Which nodes to send read requests to
 	fn read_nodes(&self, hash: &Hash) -> Vec<UUID>;
+	/// Responses needed to consider a read succesfull
 	fn read_quorum(&self) -> usize;
 
-	// Which nodes to send writes to
+	/// Which nodes to send writes to
 	fn write_nodes(&self, hash: &Hash) -> Vec<UUID>;
+	/// Responses needed to consider a write succesfull
 	fn write_quorum(&self) -> usize;
+	// this feels like its write_nodes().len() - write_quorum()
 	fn max_write_errors(&self) -> usize;
 
 	// Accessing partitions, for Merkle tree & sync
+	/// Get partition for data with given hash
 	fn partition_of(&self, hash: &Hash) -> Partition;
+	/// List of existing partitions
 	fn partitions(&self) -> Vec<(Partition, Hash)>;
 }

src/table/replication/sharded.rs

@@ -6,22 +6,25 @@ use garage_util::data::*;
 
 use crate::replication::*;
 
+/// Sharded replication schema:
+/// - based on the ring of nodes, a certain set of neighbors
+///   store entries, given as a function of the position of the
+///   entry's hash in the ring
+/// - reads are done on all of the nodes that replicate the data
+/// - writes as well
 #[derive(Clone)]
 pub struct TableShardedReplication {
+	/// The membership manager of this node
 	pub system: Arc<System>,
+	/// How many time each data should be replicated
 	pub replication_factor: usize,
+	/// How many nodes to contact for a read, should be at most `replication_factor`
 	pub read_quorum: usize,
+	/// How many nodes to contact for a write, should be at most `replication_factor`
 	pub write_quorum: usize,
 }
 
 impl TableReplication for TableShardedReplication {
-	// Sharded replication schema:
-	// - based on the ring of nodes, a certain set of neighbors
-	//   store entries, given as a function of the position of the
-	//   entry's hash in the ring
-	// - reads are done on all of the nodes that replicate the data
-	// - writes as well
-
 	fn read_nodes(&self, hash: &Hash) -> Vec<UUID> {
 		let ring = self.system.ring.borrow().clone();
 		ring.walk_ring(&hash, self.replication_factor)

src/table/schema.rs

@@ -4,7 +4,9 @@ use garage_util::data::*;
 
 use crate::crdt::CRDT;
 
+/// Trait for partitionnable data
 pub trait PartitionKey {
+	/// Get the key used to partition
 	fn hash(&self) -> Hash;
 }
 
@@ -20,7 +22,9 @@ impl PartitionKey for Hash {
 	}
 }
 
+/// Trait for sortable data
 pub trait SortKey {
+	/// Get the key used to sort
 	fn sort_key(&self) -> &[u8];
 }
 
@@ -36,25 +40,34 @@ impl SortKey for Hash {
 	}
 }
 
+/// Trait for an entry in a table. It must be sortable and partitionnable.
 pub trait Entry<P: PartitionKey, S: SortKey>:
 	CRDT + PartialEq + Clone + Serialize + for<'de> Deserialize<'de> + Send + Sync
 {
+	/// Get the key used to partition
 	fn partition_key(&self) -> &P;
+	/// Get the key used to sort
 	fn sort_key(&self) -> &S;
 
+	/// Is the entry a tombstone? Default implementation always return false
 	fn is_tombstone(&self) -> bool {
 		false
 	}
 }
 
+/// Trait for the schema used in a table
 pub trait TableSchema: Send + Sync {
+	/// The partition key used in that table
 	type P: PartitionKey + Clone + PartialEq + Serialize + for<'de> Deserialize<'de> + Send + Sync;
+	/// The sort key used int that table
 	type S: SortKey + Clone + Serialize + for<'de> Deserialize<'de> + Send + Sync;
+	/// They type for an entry in that table
 	type E: Entry<Self::P, Self::S>;
 	type Filter: Clone + Serialize + for<'de> Deserialize<'de> + Send + Sync;
 
 	// Action to take if not able to decode current version:
 	// try loading from an older version
+	/// Try migrating an entry from an older version
 	fn try_migrate(_bytes: &[u8]) -> Option<Self::E> {
 		None
 	}
@@ -63,9 +76,7 @@ pub trait TableSchema: Send + Sync {
 	// as the update itself is an unchangeable fact that will never go back
 	// due to CRDT logic. Typically errors in propagation of info should be logged
 	// to stderr.
-	fn updated(&self, _old: Option<Self::E>, _new: Option<Self::E>) {}
+	fn updated(&self, old: Option<Self::E>, new: Option<Self::E>);
 
-	fn matches_filter(_entry: &Self::E, _filter: &Self::Filter) -> bool {
-		true
-	}
+	fn matches_filter(entry: &Self::E, filter: &Self::Filter) -> bool;
 }

src/web/lib.rs

@@ -1,6 +1,7 @@
 #[macro_use]
 extern crate log;
 
-pub mod error;
+mod error;
+pub use error::Error;
 
 pub mod web_server;