Add TaprootBuilder class

This class functions as a utility for building taproot outputs, from
internal key and script leaves.

src/script/interpreter.cpp

@@ -1484,8 +1484,8 @@ template PrecomputedTransactionData::PrecomputedTransactionData(const CTransacti
 template PrecomputedTransactionData::PrecomputedTransactionData(const CMutableTransaction& txTo);
 
 static const CHashWriter HASHER_TAPSIGHASH = TaggedHash("TapSighash");
-static const CHashWriter HASHER_TAPLEAF = TaggedHash("TapLeaf");
-static const CHashWriter HASHER_TAPBRANCH = TaggedHash("TapBranch");
+const CHashWriter HASHER_TAPLEAF = TaggedHash("TapLeaf");
+const CHashWriter HASHER_TAPBRANCH = TaggedHash("TapBranch");
 
 static bool HandleMissingData(MissingDataBehavior mdb)
 {

src/script/interpreter.h

@@ -6,6 +6,7 @@
 #ifndef BITCOIN_SCRIPT_INTERPRETER_H
 #define BITCOIN_SCRIPT_INTERPRETER_H
 
+#include <hash.h>
 #include <script/script_error.h>
 #include <span.h>
 #include <primitives/transaction.h>
@@ -218,6 +219,9 @@ static constexpr size_t TAPROOT_CONTROL_NODE_SIZE = 32;
 static constexpr size_t TAPROOT_CONTROL_MAX_NODE_COUNT = 128;
 static constexpr size_t TAPROOT_CONTROL_MAX_SIZE = TAPROOT_CONTROL_BASE_SIZE + TAPROOT_CONTROL_NODE_SIZE * TAPROOT_CONTROL_MAX_NODE_COUNT;
 
+extern const CHashWriter HASHER_TAPLEAF;    //!< Hasher with tag "TapLeaf" pre-fed to it.
+extern const CHashWriter HASHER_TAPBRANCH;  //!< Hasher with tag "TapBranch" pre-fed to it.
+
 template <class T>
 uint256 SignatureHash(const CScript& scriptCode, const T& txTo, unsigned int nIn, int nHashType, const CAmount& amount, SigVersion sigversion, const PrecomputedTransactionData* cache = nullptr);
 

src/script/standard.cpp

@@ -6,8 +6,11 @@
 #include <script/standard.h>
 
 #include <crypto/sha256.h>
+#include <hash.h>
 #include <pubkey.h>
+#include <script/interpreter.h>
 #include <script/script.h>
+#include <util/strencodings.h>
 
 #include <string>
 
@@ -370,3 +373,99 @@ CScript GetScriptForMultisig(int nRequired, const std::vector<CPubKey>& keys)
 bool IsValidDestination(const CTxDestination& dest) {
     return dest.index() != 0;
 }
+
+/*static*/ TaprootBuilder::NodeInfo TaprootBuilder::Combine(NodeInfo&& a, NodeInfo&& b)
+{
+    NodeInfo ret;
+    /* Lexicographically sort a and b's hash, and compute parent hash. */
+    if (a.hash < b.hash) {
+        ret.hash = (CHashWriter(HASHER_TAPBRANCH) << a.hash << b.hash).GetSHA256();
+    } else {
+        ret.hash = (CHashWriter(HASHER_TAPBRANCH) << b.hash << a.hash).GetSHA256();
+    }
+    return ret;
+}
+
+void TaprootBuilder::Insert(TaprootBuilder::NodeInfo&& node, int depth)
+{
+    assert(depth >= 0 && (size_t)depth <= TAPROOT_CONTROL_MAX_NODE_COUNT);
+    /* We cannot insert a leaf at a lower depth while a deeper branch is unfinished. Doing
+     * so would mean the Add() invocations do not correspond to a DFS traversal of a
+     * binary tree. */
+    if ((size_t)depth + 1 < m_branch.size()) {
+        m_valid = false;
+        return;
+    }
+    /* As long as an entry in the branch exists at the specified depth, combine it and propagate up.
+     * The 'node' variable is overwritten here with the newly combined node. */
+    while (m_valid && m_branch.size() > (size_t)depth && m_branch[depth].has_value()) {
+        node = Combine(std::move(node), std::move(*m_branch[depth]));
+        m_branch.pop_back();
+        if (depth == 0) m_valid = false; /* Can't propagate further up than the root */
+        --depth;
+    }
+    if (m_valid) {
+        /* Make sure the branch is big enough to place the new node. */
+        if (m_branch.size() <= (size_t)depth) m_branch.resize((size_t)depth + 1);
+        assert(!m_branch[depth].has_value());
+        m_branch[depth] = std::move(node);
+    }
+}
+
+/*static*/ bool TaprootBuilder::ValidDepths(const std::vector<int>& depths)
+{
+    std::vector<bool> branch;
+    for (int depth : depths) {
+        // This inner loop corresponds to effectively the same logic on branch
+        // as what Insert() performs on the m_branch variable. Instead of
+        // storing a NodeInfo object, just remember whether or not there is one
+        // at that depth.
+        if (depth < 0 || (size_t)depth > TAPROOT_CONTROL_MAX_NODE_COUNT) return false;
+        if ((size_t)depth + 1 < branch.size()) return false;
+        while (branch.size() > (size_t)depth && branch[depth]) {
+            branch.pop_back();
+            if (depth == 0) return false;
+            --depth;
+        }
+        if (branch.size() <= (size_t)depth) branch.resize((size_t)depth + 1);
+        assert(!branch[depth]);
+        branch[depth] = true;
+    }
+    // And this check corresponds to the IsComplete() check on m_branch.
+    return branch.size() == 0 || (branch.size() == 1 && branch[0]);
+}
+
+TaprootBuilder& TaprootBuilder::Add(int depth, const CScript& script, int leaf_version)
+{
+    assert((leaf_version & ~TAPROOT_LEAF_MASK) == 0);
+    if (!IsValid()) return *this;
+    /* Construct NodeInfo object with leaf hash. */
+    NodeInfo node;
+    node.hash = (CHashWriter{HASHER_TAPLEAF} << uint8_t(leaf_version) << script).GetSHA256();
+    /* Insert into the branch. */
+    Insert(std::move(node), depth);
+    return *this;
+}
+
+TaprootBuilder& TaprootBuilder::AddOmitted(int depth, const uint256& hash)
+{
+    if (!IsValid()) return *this;
+    /* Construct NodeInfo object with the hash directly, and insert it into the branch. */
+    NodeInfo node;
+    node.hash = hash;
+    Insert(std::move(node), depth);
+    return *this;
+}
+
+TaprootBuilder& TaprootBuilder::Finalize(const XOnlyPubKey& internal_key)
+{
+    /* Can only call this function when IsComplete() is true. */
+    assert(IsComplete());
+    m_internal_key = internal_key;
+    auto ret = m_internal_key.CreateTapTweak(m_branch.size() == 0 ? nullptr : &m_branch[0]->hash);
+    assert(ret.has_value());
+    std::tie(m_output_key, std::ignore) = *ret;
+    return *this;
+}
+
+WitnessV1Taproot TaprootBuilder::GetOutput() { return WitnessV1Taproot{m_output_key}; }

src/script/standard.h

@@ -209,4 +209,82 @@ CScript GetScriptForRawPubKey(const CPubKey& pubkey);
 /** Generate a multisig script. */
 CScript GetScriptForMultisig(int nRequired, const std::vector<CPubKey>& keys);
 
+/** Utility class to construct Taproot outputs from internal key and script tree. */
+class TaprootBuilder
+{
+private:
+    /** Information associated with a node in the Merkle tree. */
+    struct NodeInfo
+    {
+        /** Merkle hash of this node. */
+        uint256 hash;
+    };
+    /** Whether the builder is in a valid state so far. */
+    bool m_valid = true;
+
+    /** The current state of the builder.
+     *
+     * For each level in the tree, one NodeInfo object may be present. m_branch[0]
+     * is information about the root; further values are for deeper subtrees being
+     * explored.
+     *
+     * For every right branch taken to reach the position we're currently
+     * working in, there will be a (non-nullopt) entry in m_branch corresponding
+     * to the left branch at that level.
+     *
+     * For example, imagine this tree:     - N0 -
+     *                                    /      \
+     *                                   N1      N2
+     *                                  /  \    /  \
+     *                                 A    B  C   N3
+     *                                            /  \
+     *                                           D    E
+     *
+     * Initially, m_branch is empty. After processing leaf A, it would become
+     * {nullopt, nullopt, A}. When processing leaf B, an entry at level 2 already
+     * exists, and it would thus be combined with it to produce a level 1 one,
+     * resulting in {nullopt, N1}. Adding C and D takes us to {nullopt, N1, C}
+     * and {nullopt, N1, C, D} respectively. When E is processed, it is combined
+     * with D, and then C, and then N1, to produce the root, resulting in {N0}.
+     *
+     * This structure allows processing with just O(log n) overhead if the leaves
+     * are computed on the fly.
+     *
+     * As an invariant, there can never be nullopt entries at the end. There can
+     * also not be more than 128 entries (as that would mean more than 128 levels
+     * in the tree). The depth of newly added entries will always be at least
+     * equal to the current size of m_branch (otherwise it does not correspond
+     * to a depth-first traversal of a tree). m_branch is only empty if no entries
+     * have ever be processed. m_branch having length 1 corresponds to being done.
+     */
+    std::vector<std::optional<NodeInfo>> m_branch;
+
+    XOnlyPubKey m_internal_key;  //!< The internal key, set when finalizing.
+    XOnlyPubKey m_output_key; //!< The output key, computed when finalizing. */
+
+    /** Combine information about a parent Merkle tree node from its child nodes. */
+    static NodeInfo Combine(NodeInfo&& a, NodeInfo&& b);
+    /** Insert information about a node at a certain depth, and propagate information up. */
+    void Insert(NodeInfo&& node, int depth);
+
+public:
+    /** Add a new script at a certain depth in the tree. Add() operations must be called
+     *  in depth-first traversal order of binary tree. */
+    TaprootBuilder& Add(int depth, const CScript& script, int leaf_version);
+    /** Like Add(), but for a Merkle node with a given hash to the tree. */
+    TaprootBuilder& AddOmitted(int depth, const uint256& hash);
+    /** Finalize the construction. Can only be called when IsComplete() is true.
+        internal_key.IsFullyValid() must be true. */
+    TaprootBuilder& Finalize(const XOnlyPubKey& internal_key);
+
+    /** Return true if so far all input was valid. */
+    bool IsValid() const { return m_valid; }
+    /** Return whether there were either no leaves, or the leaves form a Huffman tree. */
+    bool IsComplete() const { return m_valid && (m_branch.size() == 0 || (m_branch.size() == 1 && m_branch[0].has_value())); }
+    /** Compute scriptPubKey (after Finalize()). */
+    WitnessV1Taproot GetOutput();
+    /** Check if a list of depths is legal (will lead to IsComplete()). */
+    static bool ValidDepths(const std::vector<int>& depths);
+};
+
 #endif // BITCOIN_SCRIPT_STANDARD_H