core/types: improve encoding/decoding

Author: fjl

core/types/tx_blob.go

@@ -19,6 +19,7 @@ package types
 import (
 	"bytes"
 	"crypto/sha256"
+	"errors"
 	"fmt"
 	"math/big"
 
@@ -119,22 +120,55 @@ func (sc *BlobTxSidecar) ValidateBlobCommitmentHashes(hashes []common.Hash) erro
 	return nil
 }
 
-// blobTxWithBlobs is used for encoding of transactions when blobs are present.
-type blobTxWithBlobs struct {
+// blobTxWithBlobs represents blob tx with its corresponding sidecar.
+// This is an interface because sidecars are versioned.
+type blobTxWithBlobs interface {
+	tx() *BlobTx
+	assign(*BlobTxSidecar) error
+}
+
+type blobTxWithBlobsV0 struct {
 	BlobTx      *BlobTx
 	Blobs       []kzg4844.Blob
 	Commitments []kzg4844.Commitment
 	Proofs      []kzg4844.Proof
 }
 
-type versionedBlobTxWithBlobs struct {
+type blobTxWithBlobsV1 struct {
 	BlobTx      *BlobTx
 	Version     byte
 	Blobs       []kzg4844.Blob
 	Commitments []kzg4844.Commitment
 	Proofs      []kzg4844.Proof
 }
 
+func (btx *blobTxWithBlobsV0) tx() *BlobTx {
+	return btx.BlobTx
+}
+
+func (btx *blobTxWithBlobsV0) assign(sc *BlobTxSidecar) error {
+	sc.Version = 0
+	sc.Blobs = btx.Blobs
+	sc.Commitments = btx.Commitments
+	sc.Proofs = btx.Proofs
+	return nil
+}
+
+func (btx *blobTxWithBlobsV1) tx() *BlobTx {
+	return btx.BlobTx
+}
+
+func (btx *blobTxWithBlobsV1) assign(sc *BlobTxSidecar) error {
+	if btx.Version != 1 {
+		return fmt.Errorf("unsupported blob tx version %d", btx.Version)
+	}
+	sc.Version = 1
+	sc.Blobs = btx.Blobs
+	sc.Commitments = btx.Commitments
+	sc.Proofs = btx.Proofs
+	return nil
+}
+
 // copy creates a deep copy of the transaction data and initializes all fields.
 func (tx *BlobTx) copy() TxData {
 	cpy := &BlobTx{
@@ -240,69 +274,98 @@ func (tx *BlobTx) withSidecar(sideCar *BlobTxSidecar) *BlobTx {
 }
 
 func (tx *BlobTx) encode(b *bytes.Buffer) error {
-	if tx.Sidecar == nil {
+	switch {
+	case tx.Sidecar == nil:
 		return rlp.Encode(b, tx)
-	}
-	// Encode a cell proof transaction
-	if tx.Sidecar.Version != 0 {
-		inner := &versionedBlobTxWithBlobs{
+
+	case tx.Sidecar.Version == 0:
+		return rlp.Encode(b, &blobTxWithBlobsV0{
+			BlobTx:      tx,
+			Blobs:       tx.Sidecar.Blobs,
+			Commitments: tx.Sidecar.Commitments,
+			Proofs:      tx.Sidecar.Proofs,
+		})
+
+	case tx.Sidecar.Version == 1:
+		return rlp.Encode(b, &blobTxWithBlobsV1{
 			BlobTx:      tx,
 			Version:     tx.Sidecar.Version,
 			Blobs:       tx.Sidecar.Blobs,
 			Commitments: tx.Sidecar.Commitments,
 			Proofs:      tx.Sidecar.Proofs,
-		}
-		return rlp.Encode(b, inner)
-	}
-	inner := &blobTxWithBlobs{
-		BlobTx:      tx,
-		Blobs:       tx.Sidecar.Blobs,
-		Commitments: tx.Sidecar.Commitments,
-		Proofs:      tx.Sidecar.Proofs,
+		})
+
+	default:
+		return errors.New("unsupported sidecar version")
 	}
-	return rlp.Encode(b, inner)
 }
 
 func (tx *BlobTx) decode(input []byte) error {
-	// Here we need to support two formats: the network protocol encoding of the tx (with
-	// blobs) or the canonical encoding without blobs.
+	// Here we need to support two outer formats: the network protocol encoding of the tx
+	// (with blobs) or the canonical encoding without blobs.
+	//
+	// The canonical encoding is just a list of fields:
+	//
+	//     [chainID, nonce, ...]
+	//
+	// The network encoding is a list where the first element is the tx in the canonical encoding,
+	// and the remaining elements are the 'sidecar':
 	//
-	// The two encodings can be distinguished by checking whether the first element of the
-	// input list is itself a list.
+	//     [[chainID, nonce, ...], ...]
+	//
+	// The two outer encodings can be distinguished by checking whether the first element
+	// of the input list is itself a list. If it's the canonical encoding, the first
+	// element is the chainID, which is a number.
 
-	outerList, _, err := rlp.SplitList(input)
+	firstElem, _, err := rlp.SplitList(input)
 	if err != nil {
 		return err
 	}
-	firstElemKind, _, _, err := rlp.Split(outerList)
+	firstElemKind, _, secondElem, err := rlp.Split(firstElem)
 	if err != nil {
 		return err
 	}
-
 	if firstElemKind != rlp.List {
+		// Blob tx without blobs.
 		return rlp.DecodeBytes(input, tx)
 	}
 
-	// It's a tx with blobs. Try to decode it as version 0.
-	var inner versionedBlobTxWithBlobs
-	if err := rlp.DecodeBytes(input, &inner); err != nil {
-		var innerV0 blobTxWithBlobs
-		if err := rlp.DecodeBytes(input, &innerV0); err != nil {
-			return err
-		}
-		inner.BlobTx = innerV0.BlobTx
-		inner.Version = 0
-		inner.Blobs = innerV0.Blobs
-		inner.Commitments = innerV0.Commitments
-		inner.Proofs = innerV0.Proofs
+	// Now we know it's the network encoding with the blob sidecar. Here we again need to
+	// support multiple encodings: legacy sidecars (v0) with a blob proof, and versioned
+	// sidecars.
+	//
+	// The legacy encoding is:
+	//
+	//     [tx, blobs, commitments, proofs]
+	//
+	// The versioned encoding is:
+	//
+	//     [tx, version, blobs, ...]
+	//
+	// We can tell the two apart by checking whether the second element is the version byte.
+	// For legacy sidecar the second element is a list of blobs.
+
+	secondElemKind, _, _, err := rlp.Split(secondElem)
+	if err != nil {
+		return err
+	}
+	var payload blobTxWithBlobs
+	if secondElemKind == rlp.List {
+		// No version byte: blob sidecar v0.
+		payload = new(blobTxWithBlobsV0)
+	} else {
+		// It has a version byte. Decode as v1, version is checked by assign()
+		payload = new(blobTxWithBlobsV1)
 	}
-	*tx = *inner.BlobTx
-	tx.Sidecar = &BlobTxSidecar{
-		Version:     inner.Version,
-		Blobs:       inner.Blobs,
-		Commitments: inner.Commitments,
-		Proofs:      inner.Proofs,
+	if err := rlp.DecodeBytes(input, payload); err != nil {
+		return err
+	}
+	sc := new(BlobTxSidecar)
+	if err := payload.assign(sc); err != nil {
+		return err
 	}
+	*tx = *payload.tx()
+	tx.Sidecar = sc
 	return nil
 }