chore(trie): refactor existing header encoding

internal/trie/node/README.md

@@ -0,0 +1,31 @@
+# Trie node
+
+Package node defines the `Node` structure with methods to be used in the modified Merkle-Patricia Radix-16 trie.
+
+## Codec
+
+The following sub-sections precise the encoding of a node.
+This encoding is formally described in [the Polkadot specification](https://spec.polkadot.network/#sect-state-storage).
+
+### Header
+
+Each node encoding has a header of one or more bytes.
+The first byte contains the node variant and some or all of the partial key length of the node.
+If the partial key length cannot fit in the first byte, additional bytes are added to the header to represent the total partial key length.
+
+### Partial key
+
+The header is then concatenated with the partial key of the node, encoded as Little Endian bytes.
+
+### Remaining bytes
+
+The remaining bytes appended depend on the node variant.
+
+- For leaves, the SCALE-encoded leaf value is appended.
+- For branches, the following elements are concatenated in this order and appended to the previous header+partial key:
+  - Children bitmap (2 bytes)
+  - SCALE-encoded node value
+  - Hash(Encoding(Child[0]))
+  - Hash(Encoding(Child[1]))
+  - ...
+  - Hash(Encoding(Child[15]))

internal/trie/node/decode.go

@@ -9,63 +9,68 @@ import (
 	"fmt"
 	"io"
 
-	"github.com/ChainSafe/gossamer/internal/trie/pools"
 	"github.com/ChainSafe/gossamer/pkg/scale"
 )
 
 var (
-	ErrReadHeaderByte     = errors.New("cannot read header byte")
-	ErrUnknownNodeType    = errors.New("unknown node type")
+	// ErrDecodeValue is defined since no sentinel error is defined
+	// in the scale package.
+	// TODO remove once the following issue is done:
+	// https://github.com/ChainSafe/gossamer/issues/2631 .
 	ErrDecodeValue        = errors.New("cannot decode value")
 	ErrReadChildrenBitmap = errors.New("cannot read children bitmap")
-	ErrDecodeChildHash    = errors.New("cannot decode child hash")
+	// ErrDecodeChildHash is defined since no sentinel error is defined
+	// in the scale package.
+	// TODO remove once the following issue is done:
+	// https://github.com/ChainSafe/gossamer/issues/2631 .
+	ErrDecodeChildHash = errors.New("cannot decode child hash")
 )
 
 // Decode decodes a node from a reader.
+// The encoding format is documented in the README.md
+// of this package, and specified in the Polkadot spec at
+// https://spec.polkadot.network/#sect-state-storage
 // For branch decoding, see the comments on decodeBranch.
 // For leaf decoding, see the comments on decodeLeaf.
 func Decode(reader io.Reader) (n *Node, err error) {
-	buffer := pools.SingleByteBuffers.Get().(*bytes.Buffer)
-	defer pools.SingleByteBuffers.Put(buffer)
-	oneByteBuf := buffer.Bytes()
-	_, err = reader.Read(oneByteBuf)
+	variant, partialKeyLength, err := decodeHeader(reader)
 	if err != nil {
-		return nil, fmt.Errorf("%w: %s", ErrReadHeaderByte, err)
+		return nil, fmt.Errorf("decoding header: %w", err)
 	}
-	header := oneByteBuf[0]
 
-	nodeTypeHeaderByte := header >> 6
-	switch nodeTypeHeaderByte {
-	case leafHeader:
-		n, err = decodeLeaf(reader, header)
+	switch variant {
+	case leafVariant.bits:
+		n, err = decodeLeaf(reader, partialKeyLength)
 		if err != nil {
 			return nil, fmt.Errorf("cannot decode leaf: %w", err)
 		}
 		return n, nil
-	case branchHeader, branchWithValueHeader:
-		n, err = decodeBranch(reader, header)
+	case branchVariant.bits, branchWithValueVariant.bits:
+		n, err = decodeBranch(reader, variant, partialKeyLength)
 		if err != nil {
 			return nil, fmt.Errorf("cannot decode branch: %w", err)
 		}
 		return n, nil
 	default:
-		return nil, fmt.Errorf("%w: %d", ErrUnknownNodeType, nodeTypeHeaderByte)
+		// this is a programming error, an unknown node variant
+		// should be caught by decodeHeader.
+		panic(fmt.Sprintf("not implemented for node variant %08b", variant))
 	}
 }
 
-// decodeBranch reads and decodes from a reader with the encoding specified in internal/trie/node/encode_doc.go.
+// decodeBranch reads from a reader and decodes to a node branch.
 // Note that since the encoded branch stores the hash of the children nodes, we are not
 // reconstructing the child nodes from the encoding. This function instead stubs where the
 // children are known to be with an empty leaf. The children nodes hashes are then used to
 // find other values using the persistent database.
-func decodeBranch(reader io.Reader, header byte) (node *Node, err error) {
+func decodeBranch(reader io.Reader, variant byte, partialKeyLength uint16) (
+	node *Node, err error) {
 	node = &Node{
 		Dirty:    true,
 		Children: make([]*Node, ChildrenCapacity),
 	}
 
-	keyLen := header & keyLenOffset
-	node.Key, err = decodeKey(reader, keyLen)
+	node.Key, err = decodeKey(reader, partialKeyLength)
 	if err != nil {
 		return nil, fmt.Errorf("cannot decode key: %w", err)
 	}
@@ -78,18 +83,14 @@ func decodeBranch(reader io.Reader, header byte) (node *Node, err error) {
 
 	sd := scale.NewDecoder(reader)
 
-	nodeType := header >> 6
-	if nodeType == branchWithValueHeader {
-		var value []byte
-		// branch w/ value
-		err := sd.Decode(&value)
+	if variant == branchWithValueVariant.bits {
+		err := sd.Decode(&node.Value)
 		if err != nil {
 			return nil, fmt.Errorf("%w: %s", ErrDecodeValue, err)
 		}
-		node.Value = value
 	}
 
-	for i := 0; i < 16; i++ {
+	for i := 0; i < ChildrenCapacity; i++ {
 		if (childrenBitmap[i/8]>>(i%8))&1 != 1 {
 			continue
 		}
@@ -101,37 +102,38 @@ func decodeBranch(reader io.Reader, header byte) (node *Node, err error) {
 				ErrDecodeChildHash, i, err)
 		}
 
-		// Handle inlined leaf nodes.
 		const hashLength = 32
-		nodeTypeHeaderByte := hash[0] >> 6
-		if nodeTypeHeaderByte == leafHeader && len(hash) < hashLength {
-			leaf, err := decodeLeaf(bytes.NewReader(hash[1:]), hash[0])
-			if err != nil {
-				return nil, fmt.Errorf("%w: at index %d: %s",
-					ErrDecodeValue, i, err)
+		childNode := &Node{
+			HashDigest: hash,
+			Dirty:      true,
+		}
+		if len(hash) < hashLength {
+			// Handle inlined nodes
+			reader = bytes.NewReader(hash)
+			variant, partialKeyLength, err := decodeHeader(reader)
+			if err == nil && variant == leafVariant.bits {
+				childNode, err = decodeLeaf(reader, partialKeyLength)
+				if err != nil {
+					return nil, fmt.Errorf("%w: at index %d: %s",
+						ErrDecodeValue, i, err)
+				}
 			}
-			node.Descendants++
-			node.Children[i] = leaf
-			continue
 		}
 
 		node.Descendants++
-		node.Children[i] = &Node{
-			HashDigest: hash,
-		}
+		node.Children[i] = childNode
 	}
 
 	return node, nil
 }
 
-// decodeLeaf reads and decodes from a reader with the encoding specified in lib/trie/node/encode_doc.go.
-func decodeLeaf(reader io.Reader, header byte) (node *Node, err error) {
+// decodeLeaf reads from a reader and decodes to a leaf node.
+func decodeLeaf(reader io.Reader, partialKeyLength uint16) (node *Node, err error) {
 	node = &Node{
 		Dirty: true,
 	}
 
-	keyLen := header & keyLenOffset
-	node.Key, err = decodeKey(reader, keyLen)
+	node.Key, err = decodeKey(reader, partialKeyLength)
 	if err != nil {
 		return nil, fmt.Errorf("cannot decode key: %w", err)
 	}