diff options
author | Pieter Wuille <pieter.wuille@gmail.com> | 2019-08-28 16:23:03 -0700 |
---|---|---|
committer | Antoine Poinsot <darosior@protonmail.com> | 2022-03-17 14:09:07 +0100 |
commit | 1ddaa66eae67b102f5e37d212d366a5dcad4aa26 (patch) | |
tree | 3650c4074dc021e0d763b412feff7a4ac4f374f1 /src/script | |
parent | 4fe29368c0ded0e62f437cab3a7c904f7fd3ad67 (diff) |
Miniscript: type system, script creation, text notation, tests
More information about Miniscript can be found at https://bitcoin.sipa.be/miniscript/ (the
website source is hosted at https://github.com/sipa/miniscript/).
This commit defines all fragments, their composition, parsing from
string representation and conversion to Script.
Co-Authored-By: Antoine Poinsot <darosior@protonmail.com>
Co-Authored-By: Sanket Kanjalkar <sanket1729@gmail.com>
Co-Authored-By: Samuel Dobson <dobsonsa68@gmail.com>
Diffstat (limited to 'src/script')
-rw-r--r-- | src/script/miniscript.cpp | 295 | ||||
-rw-r--r-- | src/script/miniscript.h | 1020 |
2 files changed, 1315 insertions, 0 deletions
diff --git a/src/script/miniscript.cpp b/src/script/miniscript.cpp new file mode 100644 index 0000000000..8074be6cde --- /dev/null +++ b/src/script/miniscript.cpp @@ -0,0 +1,295 @@ +// Copyright (c) 2019 The Bitcoin Core developers +// Distributed under the MIT software license, see the accompanying +// file COPYING or http://www.opensource.org/licenses/mit-license.php. + +#include <string> +#include <vector> +#include <script/script.h> +#include <script/miniscript.h> + +#include <assert.h> + +namespace miniscript { +namespace internal { + +Type SanitizeType(Type e) { + int num_types = (e << "K"_mst) + (e << "V"_mst) + (e << "B"_mst) + (e << "W"_mst); + if (num_types == 0) return ""_mst; // No valid type, don't care about the rest + assert(num_types == 1); // K, V, B, W all conflict with each other + bool ok = // Work around a GCC 4.8 bug that breaks user-defined literals in macro calls. + (!(e << "z"_mst) || !(e << "o"_mst)) && // z conflicts with o + (!(e << "n"_mst) || !(e << "z"_mst)) && // n conflicts with z + (!(e << "n"_mst) || !(e << "W"_mst)) && // n conflicts with W + (!(e << "V"_mst) || !(e << "d"_mst)) && // V conflicts with d + (!(e << "K"_mst) || (e << "u"_mst)) && // K implies u + (!(e << "V"_mst) || !(e << "u"_mst)) && // V conflicts with u + (!(e << "e"_mst) || !(e << "f"_mst)) && // e conflicts with f + (!(e << "e"_mst) || (e << "d"_mst)) && // e implies d + (!(e << "V"_mst) || !(e << "e"_mst)) && // V conflicts with e + (!(e << "d"_mst) || !(e << "f"_mst)) && // d conflicts with f + (!(e << "V"_mst) || (e << "f"_mst)) && // V implies f + (!(e << "K"_mst) || (e << "s"_mst)) && // K implies s + (!(e << "z"_mst) || (e << "m"_mst)); // z implies m + assert(ok); + return e; +} + +Type ComputeType(Fragment nodetype, Type x, Type y, Type z, const std::vector<Type>& sub_types, uint32_t k, size_t data_size, size_t n_subs, size_t n_keys) { + // Sanity check on data + if (nodetype == Fragment::SHA256 || nodetype == Fragment::HASH256) { + assert(data_size == 32); + } else if (nodetype == Fragment::RIPEMD160 || nodetype == Fragment::HASH160) { + assert(data_size == 20); + } else { + assert(data_size == 0); + } + // Sanity check on k + if (nodetype == Fragment::OLDER || nodetype == Fragment::AFTER) { + assert(k >= 1 && k < 0x80000000UL); + } else if (nodetype == Fragment::MULTI) { + assert(k >= 1 && k <= n_keys); + } else if (nodetype == Fragment::THRESH) { + assert(k >= 1 && k <= n_subs); + } else { + assert(k == 0); + } + // Sanity check on subs + if (nodetype == Fragment::AND_V || nodetype == Fragment::AND_B || nodetype == Fragment::OR_B || + nodetype == Fragment::OR_C || nodetype == Fragment::OR_I || nodetype == Fragment::OR_D) { + assert(n_subs == 2); + } else if (nodetype == Fragment::ANDOR) { + assert(n_subs == 3); + } else if (nodetype == Fragment::WRAP_A || nodetype == Fragment::WRAP_S || nodetype == Fragment::WRAP_C || + nodetype == Fragment::WRAP_D || nodetype == Fragment::WRAP_V || nodetype == Fragment::WRAP_J || + nodetype == Fragment::WRAP_N) { + assert(n_subs == 1); + } else if (nodetype != Fragment::THRESH) { + assert(n_subs == 0); + } + // Sanity check on keys + if (nodetype == Fragment::PK_K || nodetype == Fragment::PK_H) { + assert(n_keys == 1); + } else if (nodetype == Fragment::MULTI) { + assert(n_keys >= 1 && n_keys <= 20); + } else { + assert(n_keys == 0); + } + + // Below is the per-nodetype logic for computing the expression types. + // It heavily relies on Type's << operator (where "X << a_mst" means + // "X has all properties listed in a"). + switch (nodetype) { + case Fragment::PK_K: return "Konudemsxk"_mst; + case Fragment::PK_H: return "Knudemsxk"_mst; + case Fragment::OLDER: return + "g"_mst.If(k & CTxIn::SEQUENCE_LOCKTIME_TYPE_FLAG) | + "h"_mst.If(!(k & CTxIn::SEQUENCE_LOCKTIME_TYPE_FLAG)) | + "Bzfmxk"_mst; + case Fragment::AFTER: return + "i"_mst.If(k >= LOCKTIME_THRESHOLD) | + "j"_mst.If(k < LOCKTIME_THRESHOLD) | + "Bzfmxk"_mst; + case Fragment::SHA256: return "Bonudmk"_mst; + case Fragment::RIPEMD160: return "Bonudmk"_mst; + case Fragment::HASH256: return "Bonudmk"_mst; + case Fragment::HASH160: return "Bonudmk"_mst; + case Fragment::JUST_1: return "Bzufmxk"_mst; + case Fragment::JUST_0: return "Bzudemsxk"_mst; + case Fragment::WRAP_A: return + "W"_mst.If(x << "B"_mst) | // W=B_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "udfems"_mst) | // u=u_x, d=d_x, f=f_x, e=e_x, m=m_x, s=s_x + "x"_mst; // x + case Fragment::WRAP_S: return + "W"_mst.If(x << "Bo"_mst) | // W=B_x*o_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "udfemsx"_mst); // u=u_x, d=d_x, f=f_x, e=e_x, m=m_x, s=s_x, x=x_x + case Fragment::WRAP_C: return + "B"_mst.If(x << "K"_mst) | // B=K_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "ondfem"_mst) | // o=o_x, n=n_x, d=d_x, f=f_x, e=e_x, m=m_x + "us"_mst; // u, s + case Fragment::WRAP_D: return + "B"_mst.If(x << "Vz"_mst) | // B=V_x*z_x + "o"_mst.If(x << "z"_mst) | // o=z_x + "e"_mst.If(x << "f"_mst) | // e=f_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "ms"_mst) | // m=m_x, s=s_x + "nudx"_mst; // n, u, d, x + case Fragment::WRAP_V: return + "V"_mst.If(x << "B"_mst) | // V=B_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "zonms"_mst) | // z=z_x, o=o_x, n=n_x, m=m_x, s=s_x + "fx"_mst; // f, x + case Fragment::WRAP_J: return + "B"_mst.If(x << "Bn"_mst) | // B=B_x*n_x + "e"_mst.If(x << "f"_mst) | // e=f_x + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "oums"_mst) | // o=o_x, u=u_x, m=m_x, s=s_x + "ndx"_mst; // n, d, x + case Fragment::WRAP_N: return + (x & "ghijk"_mst) | // g=g_x, h=h_x, i=i_x, j=j_x, k=k_x + (x & "Bzondfems"_mst) | // B=B_x, z=z_x, o=o_x, n=n_x, d=d_x, f=f_x, e=e_x, m=m_x, s=s_x + "ux"_mst; // u, x + case Fragment::AND_V: return + (y & "KVB"_mst).If(x << "V"_mst) | // B=V_x*B_y, V=V_x*V_y, K=V_x*K_y + (x & "n"_mst) | (y & "n"_mst).If(x << "z"_mst) | // n=n_x+z_x*n_y + ((x | y) & "o"_mst).If((x | y) << "z"_mst) | // o=o_x*z_y+z_x*o_y + (x & y & "dmz"_mst) | // d=d_x*d_y, m=m_x*m_y, z=z_x*z_y + ((x | y) & "s"_mst) | // s=s_x+s_y + "f"_mst.If((y << "f"_mst) || (x << "s"_mst)) | // f=f_y+s_x + (y & "ux"_mst) | // u=u_y, x=x_y + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + "k"_mst.If(((x & y) << "k"_mst) && + !(((x << "g"_mst) && (y << "h"_mst)) || + ((x << "h"_mst) && (y << "g"_mst)) || + ((x << "i"_mst) && (y << "j"_mst)) || + ((x << "j"_mst) && (y << "i"_mst)))); // k=k_x*k_y*!(g_x*h_y + h_x*g_y + i_x*j_y + j_x*i_y) + case Fragment::AND_B: return + (x & "B"_mst).If(y << "W"_mst) | // B=B_x*W_y + ((x | y) & "o"_mst).If((x | y) << "z"_mst) | // o=o_x*z_y+z_x*o_y + (x & "n"_mst) | (y & "n"_mst).If(x << "z"_mst) | // n=n_x+z_x*n_y + (x & y & "e"_mst).If((x & y) << "s"_mst) | // e=e_x*e_y*s_x*s_y + (x & y & "dzm"_mst) | // d=d_x*d_y, z=z_x*z_y, m=m_x*m_y + "f"_mst.If(((x & y) << "f"_mst) || (x << "sf"_mst) || (y << "sf"_mst)) | // f=f_x*f_y + f_x*s_x + f_y*s_y + ((x | y) & "s"_mst) | // s=s_x+s_y + "ux"_mst | // u, x + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + "k"_mst.If(((x & y) << "k"_mst) && + !(((x << "g"_mst) && (y << "h"_mst)) || + ((x << "h"_mst) && (y << "g"_mst)) || + ((x << "i"_mst) && (y << "j"_mst)) || + ((x << "j"_mst) && (y << "i"_mst)))); // k=k_x*k_y*!(g_x*h_y + h_x*g_y + i_x*j_y + j_x*i_y) + case Fragment::OR_B: return + "B"_mst.If(x << "Bd"_mst && y << "Wd"_mst) | // B=B_x*d_x*W_x*d_y + ((x | y) & "o"_mst).If((x | y) << "z"_mst) | // o=o_x*z_y+z_x*o_y + (x & y & "m"_mst).If((x | y) << "s"_mst && (x & y) << "e"_mst) | // m=m_x*m_y*e_x*e_y*(s_x+s_y) + (x & y & "zse"_mst) | // z=z_x*z_y, s=s_x*s_y, e=e_x*e_y + "dux"_mst | // d, u, x + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + (x & y & "k"_mst); // k=k_x*k_y + case Fragment::OR_D: return + (y & "B"_mst).If(x << "Bdu"_mst) | // B=B_y*B_x*d_x*u_x + (x & "o"_mst).If(y << "z"_mst) | // o=o_x*z_y + (x & y & "m"_mst).If(x << "e"_mst && (x | y) << "s"_mst) | // m=m_x*m_y*e_x*(s_x+s_y) + (x & y & "zes"_mst) | // z=z_x*z_y, e=e_x*e_y, s=s_x*s_y + (y & "ufd"_mst) | // u=u_y, f=f_y, d=d_y + "x"_mst | // x + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + (x & y & "k"_mst); // k=k_x*k_y + case Fragment::OR_C: return + (y & "V"_mst).If(x << "Bdu"_mst) | // V=V_y*B_x*u_x*d_x + (x & "o"_mst).If(y << "z"_mst) | // o=o_x*z_y + (x & y & "m"_mst).If(x << "e"_mst && (x | y) << "s"_mst) | // m=m_x*m_y*e_x*(s_x+s_y) + (x & y & "zs"_mst) | // z=z_x*z_y, s=s_x*s_y + "fx"_mst | // f, x + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + (x & y & "k"_mst); // k=k_x*k_y + case Fragment::OR_I: return + (x & y & "VBKufs"_mst) | // V=V_x*V_y, B=B_x*B_y, K=K_x*K_y, u=u_x*u_y, f=f_x*f_y, s=s_x*s_y + "o"_mst.If((x & y) << "z"_mst) | // o=z_x*z_y + ((x | y) & "e"_mst).If((x | y) << "f"_mst) | // e=e_x*f_y+f_x*e_y + (x & y & "m"_mst).If((x | y) << "s"_mst) | // m=m_x*m_y*(s_x+s_y) + ((x | y) & "d"_mst) | // d=d_x+d_y + "x"_mst | // x + ((x | y) & "ghij"_mst) | // g=g_x+g_y, h=h_x+h_y, i=i_x+i_y, j=j_x+j_y + (x & y & "k"_mst); // k=k_x*k_y + case Fragment::ANDOR: return + (y & z & "BKV"_mst).If(x << "Bdu"_mst) | // B=B_x*d_x*u_x*B_y*B_z, K=B_x*d_x*u_x*K_y*K_z, V=B_x*d_x*u_x*V_y*V_z + (x & y & z & "z"_mst) | // z=z_x*z_y*z_z + ((x | (y & z)) & "o"_mst).If((x | (y & z)) << "z"_mst) | // o=o_x*z_y*z_z+z_x*o_y*o_z + (y & z & "u"_mst) | // u=u_y*u_z + (z & "f"_mst).If((x << "s"_mst) || (y << "f"_mst)) | // f=(s_x+f_y)*f_z + (z & "d"_mst) | // d=d_z + (x & z & "e"_mst).If(x << "s"_mst || y << "f"_mst) | // e=e_x*e_z*(s_x+f_y) + (x & y & z & "m"_mst).If(x << "e"_mst && (x | y | z) << "s"_mst) | // m=m_x*m_y*m_z*e_x*(s_x+s_y+s_z) + (z & (x | y) & "s"_mst) | // s=s_z*(s_x+s_y) + "x"_mst | // x + ((x | y | z) & "ghij"_mst) | // g=g_x+g_y+g_z, h=h_x+h_y+h_z, i=i_x+i_y+i_z, j=j_x+j_y_j_z + "k"_mst.If(((x & y & z) << "k"_mst) && + !(((x << "g"_mst) && (y << "h"_mst)) || + ((x << "h"_mst) && (y << "g"_mst)) || + ((x << "i"_mst) && (y << "j"_mst)) || + ((x << "j"_mst) && (y << "i"_mst)))); // k=k_x*k_y*k_z* !(g_x*h_y + h_x*g_y + i_x*j_y + j_x*i_y) + case Fragment::MULTI: return "Bnudemsk"_mst; + case Fragment::THRESH: { + bool all_e = true; + bool all_m = true; + uint32_t args = 0; + uint32_t num_s = 0; + Type acc_tl = "k"_mst; + for (size_t i = 0; i < sub_types.size(); ++i) { + Type t = sub_types[i]; + if (!(t << (i ? "Wdu"_mst : "Bdu"_mst))) return ""_mst; // Require Bdu, Wdu, Wdu, ... + if (!(t << "e"_mst)) all_e = false; + if (!(t << "m"_mst)) all_m = false; + if (t << "s"_mst) num_s += 1; + args += (t << "z"_mst) ? 0 : (t << "o"_mst) ? 1 : 2; + acc_tl = ((acc_tl | t) & "ghij"_mst) | + // Thresh contains a combination of timelocks if it has threshold > 1 and + // it contains two different children that have different types of timelocks + // Note how if any of the children don't have "k", the parent also does not have "k" + "k"_mst.If(((acc_tl & t) << "k"_mst) && ((k <= 1) || + ((k > 1) && !(((acc_tl << "g"_mst) && (t << "h"_mst)) || + ((acc_tl << "h"_mst) && (t << "g"_mst)) || + ((acc_tl << "i"_mst) && (t << "j"_mst)) || + ((acc_tl << "j"_mst) && (t << "i"_mst)))))); + } + return "Bdu"_mst | + "z"_mst.If(args == 0) | // z=all z + "o"_mst.If(args == 1) | // o=all z except one o + "e"_mst.If(all_e && num_s == n_subs) | // e=all e and all s + "m"_mst.If(all_e && all_m && num_s >= n_subs - k) | // m=all e, >=(n-k) s + "s"_mst.If(num_s >= n_subs - k + 1) | // s= >=(n-k+1) s + acc_tl; // timelock info + } + } + assert(false); + return ""_mst; +} + +size_t ComputeScriptLen(Fragment nodetype, Type sub0typ, size_t subsize, uint32_t k, size_t n_subs, size_t n_keys) { + switch (nodetype) { + case Fragment::JUST_1: + case Fragment::JUST_0: return 1; + case Fragment::PK_K: return 34; + case Fragment::PK_H: return 3 + 21; + case Fragment::OLDER: + case Fragment::AFTER: return 1 + BuildScript(k).size(); + case Fragment::HASH256: + case Fragment::SHA256: return 4 + 2 + 33; + case Fragment::HASH160: + case Fragment::RIPEMD160: return 4 + 2 + 21; + case Fragment::MULTI: return 3 + (n_keys > 16) + (k > 16) + 34 * n_keys; + case Fragment::AND_V: return subsize; + case Fragment::WRAP_V: return subsize + (sub0typ << "x"_mst); + case Fragment::WRAP_S: + case Fragment::WRAP_C: + case Fragment::WRAP_N: + case Fragment::AND_B: + case Fragment::OR_B: return subsize + 1; + case Fragment::WRAP_A: + case Fragment::OR_C: return subsize + 2; + case Fragment::WRAP_D: + case Fragment::OR_D: + case Fragment::OR_I: + case Fragment::ANDOR: return subsize + 3; + case Fragment::WRAP_J: return subsize + 4; + case Fragment::THRESH: return subsize + n_subs + BuildScript(k).size(); + } + assert(false); + return 0; +} + +int FindNextChar(Span<const char> sp, const char m) +{ + for (int i = 0; i < (int)sp.size(); ++i) { + if (sp[i] == m) return i; + // We only search within the current parentheses + if (sp[i] == ')') break; + } + return -1; +} + +} // namespace internal +} // namespace miniscript diff --git a/src/script/miniscript.h b/src/script/miniscript.h new file mode 100644 index 0000000000..a394aed146 --- /dev/null +++ b/src/script/miniscript.h @@ -0,0 +1,1020 @@ +// Copyright (c) 2019 The Bitcoin Core developers +// Distributed under the MIT software license, see the accompanying +// file COPYING or http://www.opensource.org/licenses/mit-license.php. + +#ifndef BITCOIN_SCRIPT_MINISCRIPT_H +#define BITCOIN_SCRIPT_MINISCRIPT_H + +#include <algorithm> +#include <numeric> +#include <memory> +#include <optional> +#include <string> +#include <vector> + +#include <stdlib.h> +#include <assert.h> + +#include <policy/policy.h> +#include <primitives/transaction.h> +#include <script/script.h> +#include <span.h> +#include <util/spanparsing.h> +#include <util/strencodings.h> +#include <util/string.h> +#include <util/vector.h> + +namespace miniscript { + +/** This type encapsulates the miniscript type system properties. + * + * Every miniscript expression is one of 4 basic types, and additionally has + * a number of boolean type properties. + * + * The basic types are: + * - "B" Base: + * - Takes its inputs from the top of the stack. + * - When satisfied, pushes a nonzero value of up to 4 bytes onto the stack. + * - When dissatisfied, pushes a 0 onto the stack. + * - This is used for most expressions, and required for the top level one. + * - For example: older(n) = <n> OP_CHECKSEQUENCEVERIFY. + * - "V" Verify: + * - Takes its inputs from the top of the stack. + * - When satisfactied, pushes nothing. + * - Cannot be dissatisfied. + * - This can be obtained by adding an OP_VERIFY to a B, modifying the last opcode + * of a B to its -VERIFY version (only for OP_CHECKSIG, OP_CHECKSIGVERIFY + * and OP_EQUAL), or by combining a V fragment under some conditions. + * - For example vc:pk_k(key) = <key> OP_CHECKSIGVERIFY + * - "K" Key: + * - Takes its inputs from the top of the stack. + * - Becomes a B when followed by OP_CHECKSIG. + * - Always pushes a public key onto the stack, for which a signature is to be + * provided to satisfy the expression. + * - For example pk_h(key) = OP_DUP OP_HASH160 <Hash160(key)> OP_EQUALVERIFY + * - "W" Wrapped: + * - Takes its input from one below the top of the stack. + * - When satisfied, pushes a nonzero value (like B) on top of the stack, or one below. + * - When dissatisfied, pushes 0 op top of the stack or one below. + * - Is always "OP_SWAP [B]" or "OP_TOALTSTACK [B] OP_FROMALTSTACK". + * - For example sc:pk_k(key) = OP_SWAP <key> OP_CHECKSIG + * + * There a type properties that help reasoning about correctness: + * - "z" Zero-arg: + * - Is known to always consume exactly 0 stack elements. + * - For example after(n) = <n> OP_CHECKLOCKTIMEVERIFY + * - "o" One-arg: + * - Is known to always consume exactly 1 stack element. + * - Conflicts with property 'z' + * - For example sha256(hash) = OP_SIZE 32 OP_EQUALVERIFY OP_SHA256 <hash> OP_EQUAL + * - "n" Nonzero: + * - For every way this expression can be satisfied, a satisfaction exists that never needs + * a zero top stack element. + * - Conflicts with property 'z' and with type 'W'. + * - "d" Dissatisfiable: + * - There is an easy way to construct a dissatisfaction for this expression. + * - Conflicts with type 'V'. + * - "u" Unit: + * - In case of satisfaction, an exact 1 is put on the stack (rather than just nonzero). + * - Conflicts with type 'V'. + * + * Additional type properties help reasoning about nonmalleability: + * - "e" Expression: + * - This implies property 'd', but the dissatisfaction is nonmalleable. + * - This generally requires 'e' for all subexpressions which are invoked for that + * dissatifsaction, and property 'f' for the unexecuted subexpressions in that case. + * - Conflicts with type 'V'. + * - "f" Forced: + * - Dissatisfactions (if any) for this expression always involve at least one signature. + * - Is always true for type 'V'. + * - "s" Safe: + * - Satisfactions for this expression always involve at least one signature. + * - "m" Nonmalleable: + * - For every way this expression can be satisfied (which may be none), + * a nonmalleable satisfaction exists. + * - This generally requires 'm' for all subexpressions, and 'e' for all subexpressions + * which are dissatisfied when satisfying the parent. + * + * One type property is an implementation detail: + * - "x" Expensive verify: + * - Expressions with this property have a script whose last opcode is not EQUAL, CHECKSIG, or CHECKMULTISIG. + * - Not having this property means that it can be converted to a V at no cost (by switching to the + * -VERIFY version of the last opcode). + * + * Five more type properties for representing timelock information. Spend paths + * in miniscripts containing conflicting timelocks and heightlocks cannot be spent together. + * This helps users detect if miniscript does not match the semantic behaviour the + * user expects. + * - "g" Whether the branch contains a relative time timelock + * - "h" Whether the branch contains a relative height timelock + * - "i" Whether the branch contains an absolute time timelock + * - "j" Whether the branch contains an absolute height timelock + * - "k" + * - Whether all satisfactions of this expression don't contain a mix of heightlock and timelock + * of the same type. + * - If the miniscript does not have the "k" property, the miniscript template will not match + * the user expectation of the corresponding spending policy. + * For each of these properties the subset rule holds: an expression with properties X, Y, and Z, is also + * valid in places where an X, a Y, a Z, an XY, ... is expected. +*/ +class Type { + //! Internal bitmap of properties (see ""_mst operator for details). + uint32_t m_flags; + + //! Internal constructor used by the ""_mst operator. + explicit constexpr Type(uint32_t flags) : m_flags(flags) {} + +public: + //! The only way to publicly construct a Type is using this literal operator. + friend constexpr Type operator"" _mst(const char* c, size_t l); + + //! Compute the type with the union of properties. + constexpr Type operator|(Type x) const { return Type(m_flags | x.m_flags); } + + //! Compute the type with the intersection of properties. + constexpr Type operator&(Type x) const { return Type(m_flags & x.m_flags); } + + //! Check whether the left hand's properties are superset of the right's (= left is a subtype of right). + constexpr bool operator<<(Type x) const { return (x.m_flags & ~m_flags) == 0; } + + //! Comparison operator to enable use in sets/maps (total ordering incompatible with <<). + constexpr bool operator<(Type x) const { return m_flags < x.m_flags; } + + //! Equality operator. + constexpr bool operator==(Type x) const { return m_flags == x.m_flags; } + + //! The empty type if x is false, itself otherwise. + constexpr Type If(bool x) const { return Type(x ? m_flags : 0); } +}; + +//! Literal operator to construct Type objects. +inline constexpr Type operator"" _mst(const char* c, size_t l) { + Type typ{0}; + + for (const char *p = c; p < c + l; p++) { + typ = typ | Type( + *p == 'B' ? 1 << 0 : // Base type + *p == 'V' ? 1 << 1 : // Verify type + *p == 'K' ? 1 << 2 : // Key type + *p == 'W' ? 1 << 3 : // Wrapped type + *p == 'z' ? 1 << 4 : // Zero-arg property + *p == 'o' ? 1 << 5 : // One-arg property + *p == 'n' ? 1 << 6 : // Nonzero arg property + *p == 'd' ? 1 << 7 : // Dissatisfiable property + *p == 'u' ? 1 << 8 : // Unit property + *p == 'e' ? 1 << 9 : // Expression property + *p == 'f' ? 1 << 10 : // Forced property + *p == 's' ? 1 << 11 : // Safe property + *p == 'm' ? 1 << 12 : // Nonmalleable property + *p == 'x' ? 1 << 13 : // Expensive verify + *p == 'g' ? 1 << 14 : // older: contains relative time timelock (csv_time) + *p == 'h' ? 1 << 15 : // older: contains relative height timelock (csv_height) + *p == 'i' ? 1 << 16 : // after: contains time timelock (cltv_time) + *p == 'j' ? 1 << 17 : // after: contains height timelock (cltv_height) + *p == 'k' ? 1 << 18 : // does not contain a combination of height and time locks + (throw std::logic_error("Unknown character in _mst literal"), 0) + ); + } + + return typ; +} + +template<typename Key> struct Node; +template<typename Key> using NodeRef = std::shared_ptr<const Node<Key>>; + +//! Construct a miniscript node as a shared_ptr. +template<typename Key, typename... Args> +NodeRef<Key> MakeNodeRef(Args&&... args) { return std::make_shared<const Node<Key>>(std::forward<Args>(args)...); } + +//! The different node types in miniscript. +enum class Fragment { + JUST_0, //!< OP_0 + JUST_1, //!< OP_1 + PK_K, //!< [key] + PK_H, //!< OP_DUP OP_HASH160 [keyhash] OP_EQUALVERIFY + OLDER, //!< [n] OP_CHECKSEQUENCEVERIFY + AFTER, //!< [n] OP_CHECKLOCKTIMEVERIFY + SHA256, //!< OP_SIZE 32 OP_EQUALVERIFY OP_SHA256 [hash] OP_EQUAL + HASH256, //!< OP_SIZE 32 OP_EQUALVERIFY OP_HASH256 [hash] OP_EQUAL + RIPEMD160, //!< OP_SIZE 32 OP_EQUALVERIFY OP_RIPEMD160 [hash] OP_EQUAL + HASH160, //!< OP_SIZE 32 OP_EQUALVERIFY OP_HASH160 [hash] OP_EQUAL + WRAP_A, //!< OP_TOALTSTACK [X] OP_FROMALTSTACK + WRAP_S, //!< OP_SWAP [X] + WRAP_C, //!< [X] OP_CHECKSIG + WRAP_D, //!< OP_DUP OP_IF [X] OP_ENDIF + WRAP_V, //!< [X] OP_VERIFY (or -VERIFY version of last opcode in X) + WRAP_J, //!< OP_SIZE OP_0NOTEQUAL OP_IF [X] OP_ENDIF + WRAP_N, //!< [X] OP_0NOTEQUAL + AND_V, //!< [X] [Y] + AND_B, //!< [X] [Y] OP_BOOLAND + OR_B, //!< [X] [Y] OP_BOOLOR + OR_C, //!< [X] OP_NOTIF [Y] OP_ENDIF + OR_D, //!< [X] OP_IFDUP OP_NOTIF [Y] OP_ENDIF + OR_I, //!< OP_IF [X] OP_ELSE [Y] OP_ENDIF + ANDOR, //!< [X] OP_NOTIF [Z] OP_ELSE [Y] OP_ENDIF + THRESH, //!< [X1] ([Xn] OP_ADD)* [k] OP_EQUAL + MULTI, //!< [k] [key_n]* [n] OP_CHECKMULTISIG + // AND_N(X,Y) is represented as ANDOR(X,Y,0) + // WRAP_T(X) is represented as AND_V(X,1) + // WRAP_L(X) is represented as OR_I(0,X) + // WRAP_U(X) is represented as OR_I(X,0) +}; + +namespace internal { + +//! Helper function for Node::CalcType. +Type ComputeType(Fragment nodetype, Type x, Type y, Type z, const std::vector<Type>& sub_types, uint32_t k, size_t data_size, size_t n_subs, size_t n_keys); + +//! Helper function for Node::CalcScriptLen. +size_t ComputeScriptLen(Fragment nodetype, Type sub0typ, size_t subsize, uint32_t k, size_t n_subs, size_t n_keys); + +//! A helper sanitizer/checker for the output of CalcType. +Type SanitizeType(Type x); + +} // namespace internal + +//! A node in a miniscript expression. +template<typename Key> +struct Node { + //! What node type this node is. + const Fragment nodetype; + //! The k parameter (time for OLDER/AFTER, threshold for THRESH(_M)) + const uint32_t k = 0; + //! The keys used by this expression (only for PK_K/PK_H/MULTI) + const std::vector<Key> keys; + //! The data bytes in this expression (only for HASH160/HASH256/SHA256/RIPEMD10). + const std::vector<unsigned char> data; + //! Subexpressions (for WRAP_*/AND_*/OR_*/ANDOR/THRESH) + const std::vector<NodeRef<Key>> subs; + +private: + //! Cached expression type (computed by CalcType and fed through SanitizeType). + const Type typ; + //! Cached script length (computed by CalcScriptLen). + const size_t scriptlen; + + //! Compute the length of the script for this miniscript (including children). + size_t CalcScriptLen() const { + size_t subsize = 0; + for (const auto& sub : subs) { + subsize += sub->ScriptSize(); + } + Type sub0type = subs.size() > 0 ? subs[0]->GetType() : ""_mst; + return internal::ComputeScriptLen(nodetype, sub0type, subsize, k, subs.size(), keys.size()); + } + + /* Apply a recursive algorithm to a Miniscript tree, without actual recursive calls. + * + * The algorithm is defined by two functions: downfn and upfn. Conceptually, the + * result can be thought of as first using downfn to compute a "state" for each node, + * from the root down to the leaves. Then upfn is used to compute a "result" for each + * node, from the leaves back up to the root, which is then returned. In the actual + * implementation, both functions are invoked in an interleaved fashion, performing a + * depth-first traversal of the tree. + * + * In more detail, it is invoked as node.TreeEvalMaybe<Result>(root, downfn, upfn): + * - root is the state of the root node, of type State. + * - downfn is a callable (State&, const Node&, size_t) -> State, which given a + * node, its state, and an index of one of its children, computes the state of that + * child. It can modify the state. Children of a given node will have downfn() + * called in order. + * - upfn is a callable (State&&, const Node&, Span<Result>) -> std::optional<Result>, + * which given a node, its state, and a Span of the results of its children, + * computes the result of the node. If std::nullopt is returned by upfn, + * TreeEvalMaybe() immediately returns std::nullopt. + * The return value of TreeEvalMaybe is the result of the root node. + */ + template<typename Result, typename State, typename DownFn, typename UpFn> + std::optional<Result> TreeEvalMaybe(State root_state, DownFn downfn, UpFn upfn) const + { + /** Entries of the explicit stack tracked in this algorithm. */ + struct StackElem + { + const Node& node; //!< The node being evaluated. + size_t expanded; //!< How many children of this node have been expanded. + State state; //!< The state for that node. + + StackElem(const Node& node_, size_t exp_, State&& state_) : + node(node_), expanded(exp_), state(std::move(state_)) {} + }; + /* Stack of tree nodes being explored. */ + std::vector<StackElem> stack; + /* Results of subtrees so far. Their order and mapping to tree nodes + * is implicitly defined by stack. */ + std::vector<Result> results; + stack.emplace_back(*this, 0, std::move(root_state)); + + /* Here is a demonstration of the algorithm, for an example tree A(B,C(D,E),F). + * State variables are omitted for simplicity. + * + * First: stack=[(A,0)] results=[] + * stack=[(A,1),(B,0)] results=[] + * stack=[(A,1)] results=[B] + * stack=[(A,2),(C,0)] results=[B] + * stack=[(A,2),(C,1),(D,0)] results=[B] + * stack=[(A,2),(C,1)] results=[B,D] + * stack=[(A,2),(C,2),(E,0)] results=[B,D] + * stack=[(A,2),(C,2)] results=[B,D,E] + * stack=[(A,2)] results=[B,C] + * stack=[(A,3),(F,0)] results=[B,C] + * stack=[(A,3)] results=[B,C,F] + * Final: stack=[] results=[A] + */ + while (stack.size()) { + const Node& node = stack.back().node; + if (stack.back().expanded < node.subs.size()) { + /* We encounter a tree node with at least one unexpanded child. + * Expand it. By the time we hit this node again, the result of + * that child (and all earlier children) will be at the end of `results`. */ + size_t child_index = stack.back().expanded++; + State child_state = downfn(stack.back().state, node, child_index); + stack.emplace_back(*node.subs[child_index], 0, std::move(child_state)); + continue; + } + // Invoke upfn with the last node.subs.size() elements of results as input. + assert(results.size() >= node.subs.size()); + std::optional<Result> result{upfn(std::move(stack.back().state), node, + Span<Result>{results}.last(node.subs.size()))}; + // If evaluation returns std::nullopt, abort immediately. + if (!result) return {}; + // Replace the last node.subs.size() elements of results with the new result. + results.erase(results.end() - node.subs.size(), results.end()); + results.push_back(std::move(*result)); + stack.pop_back(); + } + // The final remaining results element is the root result, return it. + assert(results.size() == 1); + return std::move(results[0]); + } + + /** Like TreeEvalMaybe, but always produces a result. upfn must return Result. */ + template<typename Result, typename State, typename DownFn, typename UpFn> + Result TreeEval(State root_state, DownFn&& downfn, UpFn upfn) const + { + // Invoke TreeEvalMaybe with upfn wrapped to return std::optional<Result>, and then + // unconditionally dereference the result (it cannot be std::nullopt). + return std::move(*TreeEvalMaybe<Result>(std::move(root_state), + std::forward<DownFn>(downfn), + [&upfn](State&& state, const Node& node, Span<Result> subs) { + Result res{upfn(std::move(state), node, subs)}; + return std::optional<Result>(std::move(res)); + } + )); + } + + //! Compute the type for this miniscript. + Type CalcType() const { + using namespace internal; + + // THRESH has a variable number of subexpressions + std::vector<Type> sub_types; + if (nodetype == Fragment::THRESH) { + for (const auto& sub : subs) sub_types.push_back(sub->GetType()); + } + // All other nodes than THRESH can be computed just from the types of the 0-3 subexpressions. + Type x = subs.size() > 0 ? subs[0]->GetType() : ""_mst; + Type y = subs.size() > 1 ? subs[1]->GetType() : ""_mst; + Type z = subs.size() > 2 ? subs[2]->GetType() : ""_mst; + + return SanitizeType(ComputeType(nodetype, x, y, z, sub_types, k, data.size(), subs.size(), keys.size())); + } + +public: + template<typename Ctx> + CScript ToScript(const Ctx& ctx) const + { + // To construct the CScript for a Miniscript object, we use the TreeEval algorithm. + // The State is a boolean: whether or not the node's script expansion is followed + // by an OP_VERIFY (which may need to be combined with the last script opcode). + auto downfn = [](bool verify, const Node& node, size_t index) { + // For WRAP_V, the subexpression is certainly followed by OP_VERIFY. + if (node.nodetype == Fragment::WRAP_V) return true; + // The subexpression of WRAP_S, and the last subexpression of AND_V + // inherit the followed-by-OP_VERIFY property from the parent. + if (node.nodetype == Fragment::WRAP_S || + (node.nodetype == Fragment::AND_V && index == 1)) return verify; + return false; + }; + // The upward function computes for a node, given its followed-by-OP_VERIFY status + // and the CScripts of its child nodes, the CScript of the node. + auto upfn = [&ctx](bool verify, const Node& node, Span<CScript> subs) -> CScript { + switch (node.nodetype) { + case Fragment::PK_K: return BuildScript(ctx.ToPKBytes(node.keys[0])); + case Fragment::PK_H: return BuildScript(OP_DUP, OP_HASH160, ctx.ToPKHBytes(node.keys[0]), OP_EQUALVERIFY); + case Fragment::OLDER: return BuildScript(node.k, OP_CHECKSEQUENCEVERIFY); + case Fragment::AFTER: return BuildScript(node.k, OP_CHECKLOCKTIMEVERIFY); + case Fragment::SHA256: return BuildScript(OP_SIZE, 32, OP_EQUALVERIFY, OP_SHA256, node.data, verify ? OP_EQUALVERIFY : OP_EQUAL); + case Fragment::RIPEMD160: return BuildScript(OP_SIZE, 32, OP_EQUALVERIFY, OP_RIPEMD160, node.data, verify ? OP_EQUALVERIFY : OP_EQUAL); + case Fragment::HASH256: return BuildScript(OP_SIZE, 32, OP_EQUALVERIFY, OP_HASH256, node.data, verify ? OP_EQUALVERIFY : OP_EQUAL); + case Fragment::HASH160: return BuildScript(OP_SIZE, 32, OP_EQUALVERIFY, OP_HASH160, node.data, verify ? OP_EQUALVERIFY : OP_EQUAL); + case Fragment::WRAP_A: return BuildScript(OP_TOALTSTACK, subs[0], OP_FROMALTSTACK); + case Fragment::WRAP_S: return BuildScript(OP_SWAP, subs[0]); + case Fragment::WRAP_C: return BuildScript(std::move(subs[0]), verify ? OP_CHECKSIGVERIFY : OP_CHECKSIG); + case Fragment::WRAP_D: return BuildScript(OP_DUP, OP_IF, subs[0], OP_ENDIF); + case Fragment::WRAP_V: { + if (node.subs[0]->GetType() << "x"_mst) { + return BuildScript(std::move(subs[0]), OP_VERIFY); + } else { + return std::move(subs[0]); + } + } + case Fragment::WRAP_J: return BuildScript(OP_SIZE, OP_0NOTEQUAL, OP_IF, subs[0], OP_ENDIF); + case Fragment::WRAP_N: return BuildScript(std::move(subs[0]), OP_0NOTEQUAL); + case Fragment::JUST_1: return BuildScript(OP_1); + case Fragment::JUST_0: return BuildScript(OP_0); + case Fragment::AND_V: return BuildScript(std::move(subs[0]), subs[1]); + case Fragment::AND_B: return BuildScript(std::move(subs[0]), subs[1], OP_BOOLAND); + case Fragment::OR_B: return BuildScript(std::move(subs[0]), subs[1], OP_BOOLOR); + case Fragment::OR_D: return BuildScript(std::move(subs[0]), OP_IFDUP, OP_NOTIF, subs[1], OP_ENDIF); + case Fragment::OR_C: return BuildScript(std::move(subs[0]), OP_NOTIF, subs[1], OP_ENDIF); + case Fragment::OR_I: return BuildScript(OP_IF, subs[0], OP_ELSE, subs[1], OP_ENDIF); + case Fragment::ANDOR: return BuildScript(std::move(subs[0]), OP_NOTIF, subs[2], OP_ELSE, subs[1], OP_ENDIF); + case Fragment::MULTI: { + CScript script = BuildScript(node.k); + for (const auto& key : node.keys) { + script = BuildScript(std::move(script), ctx.ToPKBytes(key)); + } + return BuildScript(std::move(script), node.keys.size(), verify ? OP_CHECKMULTISIGVERIFY : OP_CHECKMULTISIG); + } + case Fragment::THRESH: { + CScript script = std::move(subs[0]); + for (size_t i = 1; i < subs.size(); ++i) { + script = BuildScript(std::move(script), subs[i], OP_ADD); + } + return BuildScript(std::move(script), node.k, verify ? OP_EQUALVERIFY : OP_EQUAL); + } + } + assert(false); + return {}; + }; + return TreeEval<CScript>(false, downfn, upfn); + } + + template<typename CTx> + bool ToString(const CTx& ctx, std::string& ret) const { + // To construct the std::string representation for a Miniscript object, we use + // the TreeEvalMaybe algorithm. The State is a boolean: whether the parent node is a + // wrapper. If so, non-wrapper expressions must be prefixed with a ":". + auto downfn = [](bool, const Node& node, size_t) { + return (node.nodetype == Fragment::WRAP_A || node.nodetype == Fragment::WRAP_S || + node.nodetype == Fragment::WRAP_D || node.nodetype == Fragment::WRAP_V || + node.nodetype == Fragment::WRAP_J || node.nodetype == Fragment::WRAP_N || + node.nodetype == Fragment::WRAP_C || + (node.nodetype == Fragment::AND_V && node.subs[1]->nodetype == Fragment::JUST_1) || + (node.nodetype == Fragment::OR_I && node.subs[0]->nodetype == Fragment::JUST_0) || + (node.nodetype == Fragment::OR_I && node.subs[1]->nodetype == Fragment::JUST_0)); + }; + // The upward function computes for a node, given whether its parent is a wrapper, + // and the string representations of its child nodes, the string representation of the node. + auto upfn = [&ctx](bool wrapped, const Node& node, Span<std::string> subs) -> std::optional<std::string> { + std::string ret = wrapped ? ":" : ""; + + switch (node.nodetype) { + case Fragment::WRAP_A: return "a" + std::move(subs[0]); + case Fragment::WRAP_S: return "s" + std::move(subs[0]); + case Fragment::WRAP_C: + if (node.subs[0]->nodetype == Fragment::PK_K) { + // pk(K) is syntactic sugar for c:pk_k(K) + std::string key_str; + if (!ctx.ToString(node.subs[0]->keys[0], key_str)) return {}; + return std::move(ret) + "pk(" + std::move(key_str) + ")"; + } + if (node.subs[0]->nodetype == Fragment::PK_H) { + // pkh(K) is syntactic sugar for c:pk_h(K) + std::string key_str; + if (!ctx.ToString(node.subs[0]->keys[0], key_str)) return {}; + return std::move(ret) + "pkh(" + std::move(key_str) + ")"; + } + return "c" + std::move(subs[0]); + case Fragment::WRAP_D: return "d" + std::move(subs[0]); + case Fragment::WRAP_V: return "v" + std::move(subs[0]); + case Fragment::WRAP_J: return "j" + std::move(subs[0]); + case Fragment::WRAP_N: return "n" + std::move(subs[0]); + case Fragment::AND_V: + // t:X is syntactic sugar for and_v(X,1). + if (node.subs[1]->nodetype == Fragment::JUST_1) return "t" + std::move(subs[0]); + break; + case Fragment::OR_I: + if (node.subs[0]->nodetype == Fragment::JUST_0) return "l" + std::move(subs[1]); + if (node.subs[1]->nodetype == Fragment::JUST_0) return "u" + std::move(subs[0]); + break; + default: break; + } + switch (node.nodetype) { + case Fragment::PK_K: { + std::string key_str; + if (!ctx.ToString(node.keys[0], key_str)) return {}; + return std::move(ret) + "pk_k(" + std::move(key_str) + ")"; + } + case Fragment::PK_H: { + std::string key_str; + if (!ctx.ToString(node.keys[0], key_str)) return {}; + return std::move(ret) + "pk_h(" + std::move(key_str) + ")"; + } + case Fragment::AFTER: return std::move(ret) + "after(" + ::ToString(node.k) + ")"; + case Fragment::OLDER: return std::move(ret) + "older(" + ::ToString(node.k) + ")"; + case Fragment::HASH256: return std::move(ret) + "hash256(" + HexStr(node.data) + ")"; + case Fragment::HASH160: return std::move(ret) + "hash160(" + HexStr(node.data) + ")"; + case Fragment::SHA256: return std::move(ret) + "sha256(" + HexStr(node.data) + ")"; + case Fragment::RIPEMD160: return std::move(ret) + "ripemd160(" + HexStr(node.data) + ")"; + case Fragment::JUST_1: return std::move(ret) + "1"; + case Fragment::JUST_0: return std::move(ret) + "0"; + case Fragment::AND_V: return std::move(ret) + "and_v(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::AND_B: return std::move(ret) + "and_b(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::OR_B: return std::move(ret) + "or_b(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::OR_D: return std::move(ret) + "or_d(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::OR_C: return std::move(ret) + "or_c(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::OR_I: return std::move(ret) + "or_i(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + case Fragment::ANDOR: + // and_n(X,Y) is syntactic sugar for andor(X,Y,0). + if (node.subs[2]->nodetype == Fragment::JUST_0) return std::move(ret) + "and_n(" + std::move(subs[0]) + "," + std::move(subs[1]) + ")"; + return std::move(ret) + "andor(" + std::move(subs[0]) + "," + std::move(subs[1]) + "," + std::move(subs[2]) + ")"; + case Fragment::MULTI: { + auto str = std::move(ret) + "multi(" + ::ToString(node.k); + for (const auto& key : node.keys) { + std::string key_str; + if (!ctx.ToString(key, key_str)) return {}; + str += "," + std::move(key_str); + } + return std::move(str) + ")"; + } + case Fragment::THRESH: { + auto str = std::move(ret) + "thresh(" + ::ToString(node.k); + for (auto& sub : subs) { + str += "," + std::move(sub); + } + return std::move(str) + ")"; + } + default: assert(false); + } + return ""; // Should never be reached. + }; + + auto res = TreeEvalMaybe<std::string>(false, downfn, upfn); + if (res.has_value()) ret = std::move(*res); + return res.has_value(); + } + +public: + //! Return the size of the script for this expression (faster than ToScript().size()). + size_t ScriptSize() const { return scriptlen; } + + //! Return the expression type. + Type GetType() const { return typ; } + + //! Check whether this node is valid at all. + bool IsValid() const { return !(GetType() == ""_mst) && ScriptSize() <= MAX_STANDARD_P2WSH_SCRIPT_SIZE; } + + //! Check whether this node is valid as a script on its own. + bool IsValidTopLevel() const { return IsValid() && GetType() << "B"_mst; } + + //! Check whether this script can always be satisfied in a non-malleable way. + bool IsNonMalleable() const { return GetType() << "m"_mst; } + + //! Check whether this script always needs a signature. + bool NeedsSignature() const { return GetType() << "s"_mst; } + + //! Do all sanity checks. + bool IsSane() const { return IsValid() && GetType() << "mk"_mst; } + + //! Check whether this node is safe as a script on its own. + bool IsSaneTopLevel() const { return IsValidTopLevel() && IsSane() && NeedsSignature(); } + + //! Equality testing. + bool operator==(const Node<Key>& arg) const + { + if (nodetype != arg.nodetype) return false; + if (k != arg.k) return false; + if (data != arg.data) return false; + if (keys != arg.keys) return false; + if (subs.size() != arg.subs.size()) return false; + for (size_t i = 0; i < subs.size(); ++i) { + if (!(*subs[i] == *arg.subs[i])) return false; + } + assert(scriptlen == arg.scriptlen); + assert(typ == arg.typ); + return true; + } + + // Constructors with various argument combinations. + Node(Fragment nt, std::vector<NodeRef<Key>> sub, std::vector<unsigned char> arg, uint32_t val = 0) : nodetype(nt), k(val), data(std::move(arg)), subs(std::move(sub)), typ(CalcType()), scriptlen(CalcScriptLen()) {} + Node(Fragment nt, std::vector<unsigned char> arg, uint32_t val = 0) : nodetype(nt), k(val), data(std::move(arg)), typ(CalcType()), scriptlen(CalcScriptLen()) {} + Node(Fragment nt, std::vector<NodeRef<Key>> sub, std::vector<Key> key, uint32_t val = 0) : nodetype(nt), k(val), keys(std::move(key)), subs(std::move(sub)), typ(CalcType()), scriptlen(CalcScriptLen()) {} + Node(Fragment nt, std::vector<Key> key, uint32_t val = 0) : nodetype(nt), k(val), keys(std::move(key)), typ(CalcType()), scriptlen(CalcScriptLen()) {} + Node(Fragment nt, std::vector<NodeRef<Key>> sub, uint32_t val = 0) : nodetype(nt), k(val), subs(std::move(sub)), typ(CalcType()), scriptlen(CalcScriptLen()) {} + Node(Fragment nt, uint32_t val = 0) : nodetype(nt), k(val), typ(CalcType()), scriptlen(CalcScriptLen()) {} +}; + +namespace internal { + +enum class ParseContext { + /** An expression which may be begin with wrappers followed by a colon. */ + WRAPPED_EXPR, + /** A miniscript expression which does not begin with wrappers. */ + EXPR, + + /** SWAP wraps the top constructed node with s: */ + SWAP, + /** ALT wraps the top constructed node with a: */ + ALT, + /** CHECK wraps the top constructed node with c: */ + CHECK, + /** DUP_IF wraps the top constructed node with d: */ + DUP_IF, + /** VERIFY wraps the top constructed node with v: */ + VERIFY, + /** NON_ZERO wraps the top constructed node with j: */ + NON_ZERO, + /** ZERO_NOTEQUAL wraps the top constructed node with n: */ + ZERO_NOTEQUAL, + /** WRAP_U will construct an or_i(X,0) node from the top constructed node. */ + WRAP_U, + /** WRAP_T will construct an and_v(X,1) node from the top constructed node. */ + WRAP_T, + + /** AND_N will construct an andor(X,Y,0) node from the last two constructed nodes. */ + AND_N, + /** AND_V will construct an and_v node from the last two constructed nodes. */ + AND_V, + /** AND_B will construct an and_b node from the last two constructed nodes. */ + AND_B, + /** ANDOR will construct an andor node from the last three constructed nodes. */ + ANDOR, + /** OR_B will construct an or_b node from the last two constructed nodes. */ + OR_B, + /** OR_C will construct an or_c node from the last two constructed nodes. */ + OR_C, + /** OR_D will construct an or_d node from the last two constructed nodes. */ + OR_D, + /** OR_I will construct an or_i node from the last two constructed nodes. */ + OR_I, + + /** THRESH will read a wrapped expression, and then look for a COMMA. If + * no comma follows, it will construct a thresh node from the appropriate + * number of constructed children. Otherwise, it will recurse with another + * THRESH. */ + THRESH, + + /** COMMA expects the next element to be ',' and fails if not. */ + COMMA, + /** CLOSE_BRACKET expects the next element to be ')' and fails if not. */ + CLOSE_BRACKET, +}; + +int FindNextChar(Span<const char> in, const char m); + +/** Parse a key string ending with a ')' or ','. */ +template<typename Key, typename Ctx> +std::optional<std::pair<Key, int>> ParseKeyEnd(Span<const char> in, const Ctx& ctx) +{ + Key key; + int key_size = FindNextChar(in, ')'); + if (key_size < 1) return {}; + if (!ctx.FromString(in.begin(), in.begin() + key_size, key)) return {}; + return {{std::move(key), key_size}}; +} + +/** Parse a hex string ending at the end of the fragment's text representation. */ +template<typename Ctx> +std::optional<std::pair<std::vector<unsigned char>, int>> ParseHexStrEnd(Span<const char> in, const size_t expected_size, + const Ctx& ctx) +{ + int hash_size = FindNextChar(in, ')'); + if (hash_size < 1) return {}; + std::string val = std::string(in.begin(), in.begin() + hash_size); + if (!IsHex(val)) return {}; + auto hash = ParseHex(val); + if (hash.size() != expected_size) return {}; + return {{std::move(hash), hash_size}}; +} + +/** BuildBack pops the last two elements off `constructed` and wraps them in the specified Fragment */ +template<typename Key> +void BuildBack(Fragment nt, std::vector<NodeRef<Key>>& constructed, const bool reverse = false) +{ + NodeRef<Key> child = std::move(constructed.back()); + constructed.pop_back(); + if (reverse) { + constructed.back() = MakeNodeRef<Key>(nt, Vector(std::move(child), std::move(constructed.back()))); + } else { + constructed.back() = MakeNodeRef<Key>(nt, Vector(std::move(constructed.back()), std::move(child))); + } +} + +//! Parse a miniscript from its textual descriptor form. +template<typename Key, typename Ctx> +inline NodeRef<Key> Parse(Span<const char> in, const Ctx& ctx) +{ + using namespace spanparsing; + + // The two integers are used to hold state for thresh() + std::vector<std::tuple<ParseContext, int64_t, int64_t>> to_parse; + std::vector<NodeRef<Key>> constructed; + + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + + while (!to_parse.empty()) { + // Get the current context we are decoding within + auto [cur_context, n, k] = to_parse.back(); + to_parse.pop_back(); + + switch (cur_context) { + case ParseContext::WRAPPED_EXPR: { + int colon_index = -1; + for (int i = 1; i < (int)in.size(); ++i) { + if (in[i] == ':') { + colon_index = i; + break; + } + if (in[i] < 'a' || in[i] > 'z') break; + } + // If there is no colon, this loop won't execute + for (int j = 0; j < colon_index; ++j) { + if (in[j] == 'a') { + to_parse.emplace_back(ParseContext::ALT, -1, -1); + } else if (in[j] == 's') { + to_parse.emplace_back(ParseContext::SWAP, -1, -1); + } else if (in[j] == 'c') { + to_parse.emplace_back(ParseContext::CHECK, -1, -1); + } else if (in[j] == 'd') { + to_parse.emplace_back(ParseContext::DUP_IF, -1, -1); + } else if (in[j] == 'j') { + to_parse.emplace_back(ParseContext::NON_ZERO, -1, -1); + } else if (in[j] == 'n') { + to_parse.emplace_back(ParseContext::ZERO_NOTEQUAL, -1, -1); + } else if (in[j] == 'v') { + to_parse.emplace_back(ParseContext::VERIFY, -1, -1); + } else if (in[j] == 'u') { + to_parse.emplace_back(ParseContext::WRAP_U, -1, -1); + } else if (in[j] == 't') { + to_parse.emplace_back(ParseContext::WRAP_T, -1, -1); + } else if (in[j] == 'l') { + // The l: wrapper is equivalent to or_i(0,X) + constructed.push_back(MakeNodeRef<Key>(Fragment::JUST_0)); + to_parse.emplace_back(ParseContext::OR_I, -1, -1); + } else { + return {}; + } + } + to_parse.emplace_back(ParseContext::EXPR, -1, -1); + in = in.subspan(colon_index + 1); + break; + } + case ParseContext::EXPR: { + if (Const("0", in)) { + constructed.push_back(MakeNodeRef<Key>(Fragment::JUST_0)); + } else if (Const("1", in)) { + constructed.push_back(MakeNodeRef<Key>(Fragment::JUST_1)); + } else if (Const("pk(", in)) { + auto res = ParseKeyEnd<Key, Ctx>(in, ctx); + if (!res) return {}; + auto& [key, key_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::WRAP_C, Vector(MakeNodeRef<Key>(Fragment::PK_K, Vector(std::move(key)))))); + in = in.subspan(key_size + 1); + } else if (Const("pkh(", in)) { + auto res = ParseKeyEnd<Key>(in, ctx); + if (!res) return {}; + auto& [key, key_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::WRAP_C, Vector(MakeNodeRef<Key>(Fragment::PK_H, Vector(std::move(key)))))); + in = in.subspan(key_size + 1); + } else if (Const("pk_k(", in)) { + auto res = ParseKeyEnd<Key>(in, ctx); + if (!res) return {}; + auto& [key, key_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::PK_K, Vector(std::move(key)))); + in = in.subspan(key_size + 1); + } else if (Const("pk_h(", in)) { + auto res = ParseKeyEnd<Key>(in, ctx); + if (!res) return {}; + auto& [key, key_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::PK_H, Vector(std::move(key)))); + in = in.subspan(key_size + 1); + } else if (Const("sha256(", in)) { + auto res = ParseHexStrEnd(in, 32, ctx); + if (!res) return {}; + auto& [hash, hash_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::SHA256, std::move(hash))); + in = in.subspan(hash_size + 1); + } else if (Const("ripemd160(", in)) { + auto res = ParseHexStrEnd(in, 20, ctx); + if (!res) return {}; + auto& [hash, hash_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::RIPEMD160, std::move(hash))); + in = in.subspan(hash_size + 1); + } else if (Const("hash256(", in)) { + auto res = ParseHexStrEnd(in, 32, ctx); + if (!res) return {}; + auto& [hash, hash_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::HASH256, std::move(hash))); + in = in.subspan(hash_size + 1); + } else if (Const("hash160(", in)) { + auto res = ParseHexStrEnd(in, 20, ctx); + if (!res) return {}; + auto& [hash, hash_size] = *res; + constructed.push_back(MakeNodeRef<Key>(Fragment::HASH160, std::move(hash))); + in = in.subspan(hash_size + 1); + } else if (Const("after(", in)) { + int arg_size = FindNextChar(in, ')'); + if (arg_size < 1) return {}; + int64_t num; + if (!ParseInt64(std::string(in.begin(), in.begin() + arg_size), &num)) return {}; + if (num < 1 || num >= 0x80000000L) return {}; + constructed.push_back(MakeNodeRef<Key>(Fragment::AFTER, num)); + in = in.subspan(arg_size + 1); + } else if (Const("older(", in)) { + int arg_size = FindNextChar(in, ')'); + if (arg_size < 1) return {}; + int64_t num; + if (!ParseInt64(std::string(in.begin(), in.begin() + arg_size), &num)) return {}; + if (num < 1 || num >= 0x80000000L) return {}; + constructed.push_back(MakeNodeRef<Key>(Fragment::OLDER, num)); + in = in.subspan(arg_size + 1); + } else if (Const("multi(", in)) { + // Get threshold + int next_comma = FindNextChar(in, ','); + if (next_comma < 1) return {}; + if (!ParseInt64(std::string(in.begin(), in.begin() + next_comma), &k)) return {}; + in = in.subspan(next_comma + 1); + // Get keys + std::vector<Key> keys; + while (next_comma != -1) { + Key key; + next_comma = FindNextChar(in, ','); + int key_length = (next_comma == -1) ? FindNextChar(in, ')') : next_comma; + if (key_length < 1) return {}; + if (!ctx.FromString(in.begin(), in.begin() + key_length, key)) return {}; + keys.push_back(std::move(key)); + in = in.subspan(key_length + 1); + } + if (keys.size() < 1 || keys.size() > 20) return {}; + if (k < 1 || k > (int64_t)keys.size()) return {}; + constructed.push_back(MakeNodeRef<Key>(Fragment::MULTI, std::move(keys), k)); + } else if (Const("thresh(", in)) { + int next_comma = FindNextChar(in, ','); + if (next_comma < 1) return {}; + if (!ParseInt64(std::string(in.begin(), in.begin() + next_comma), &k)) return {}; + if (k < 1) return {}; + in = in.subspan(next_comma + 1); + // n = 1 here because we read the first WRAPPED_EXPR before reaching THRESH + to_parse.emplace_back(ParseContext::THRESH, 1, k); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + } else if (Const("andor(", in)) { + to_parse.emplace_back(ParseContext::ANDOR, -1, -1); + to_parse.emplace_back(ParseContext::CLOSE_BRACKET, -1, -1); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + to_parse.emplace_back(ParseContext::COMMA, -1, -1); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + to_parse.emplace_back(ParseContext::COMMA, -1, -1); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + } else { + if (Const("and_n(", in)) { + to_parse.emplace_back(ParseContext::AND_N, -1, -1); + } else if (Const("and_b(", in)) { + to_parse.emplace_back(ParseContext::AND_B, -1, -1); + } else if (Const("and_v(", in)) { + to_parse.emplace_back(ParseContext::AND_V, -1, -1); + } else if (Const("or_b(", in)) { + to_parse.emplace_back(ParseContext::OR_B, -1, -1); + } else if (Const("or_c(", in)) { + to_parse.emplace_back(ParseContext::OR_C, -1, -1); + } else if (Const("or_d(", in)) { + to_parse.emplace_back(ParseContext::OR_D, -1, -1); + } else if (Const("or_i(", in)) { + to_parse.emplace_back(ParseContext::OR_I, -1, -1); + } else { + return {}; + } + to_parse.emplace_back(ParseContext::CLOSE_BRACKET, -1, -1); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + to_parse.emplace_back(ParseContext::COMMA, -1, -1); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + } + break; + } + case ParseContext::ALT: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_A, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::SWAP: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_S, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::CHECK: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_C, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::DUP_IF: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_D, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::NON_ZERO: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_J, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::ZERO_NOTEQUAL: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_N, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::VERIFY: { + constructed.back() = MakeNodeRef<Key>(Fragment::WRAP_V, Vector(std::move(constructed.back()))); + break; + } + case ParseContext::WRAP_U: { + constructed.back() = MakeNodeRef<Key>(Fragment::OR_I, Vector(std::move(constructed.back()), MakeNodeRef<Key>(Fragment::JUST_0))); + break; + } + case ParseContext::WRAP_T: { + constructed.back() = MakeNodeRef<Key>(Fragment::AND_V, Vector(std::move(constructed.back()), MakeNodeRef<Key>(Fragment::JUST_1))); + break; + } + case ParseContext::AND_B: { + BuildBack(Fragment::AND_B, constructed); + break; + } + case ParseContext::AND_N: { + auto mid = std::move(constructed.back()); + constructed.pop_back(); + constructed.back() = MakeNodeRef<Key>(Fragment::ANDOR, Vector(std::move(constructed.back()), std::move(mid), MakeNodeRef<Key>(Fragment::JUST_0))); + break; + } + case ParseContext::AND_V: { + BuildBack(Fragment::AND_V, constructed); + break; + } + case ParseContext::OR_B: { + BuildBack(Fragment::OR_B, constructed); + break; + } + case ParseContext::OR_C: { + BuildBack(Fragment::OR_C, constructed); + break; + } + case ParseContext::OR_D: { + BuildBack(Fragment::OR_D, constructed); + break; + } + case ParseContext::OR_I: { + BuildBack(Fragment::OR_I, constructed); + break; + } + case ParseContext::ANDOR: { + auto right = std::move(constructed.back()); + constructed.pop_back(); + auto mid = std::move(constructed.back()); + constructed.pop_back(); + constructed.back() = MakeNodeRef<Key>(Fragment::ANDOR, Vector(std::move(constructed.back()), std::move(mid), std::move(right))); + break; + } + case ParseContext::THRESH: { + if (in.size() < 1) return {}; + if (in[0] == ',') { + in = in.subspan(1); + to_parse.emplace_back(ParseContext::THRESH, n+1, k); + to_parse.emplace_back(ParseContext::WRAPPED_EXPR, -1, -1); + } else if (in[0] == ')') { + if (k > n) return {}; + in = in.subspan(1); + // Children are constructed in reverse order, so iterate from end to beginning + std::vector<NodeRef<Key>> subs; + for (int i = 0; i < n; ++i) { + subs.push_back(std::move(constructed.back())); + constructed.pop_back(); + } + std::reverse(subs.begin(), subs.end()); + constructed.push_back(MakeNodeRef<Key>(Fragment::THRESH, std::move(subs), k)); + } else { + return {}; + } + break; + } + case ParseContext::COMMA: { + if (in.size() < 1 || in[0] != ',') return {}; + in = in.subspan(1); + break; + } + case ParseContext::CLOSE_BRACKET: { + if (in.size() < 1 || in[0] != ')') return {}; + in = in.subspan(1); + break; + } + } + } + + // Sanity checks on the produced miniscript + assert(constructed.size() == 1); + if (in.size() > 0) return {}; + const NodeRef<Key> tl_node = std::move(constructed.front()); + if (!tl_node->IsValidTopLevel()) return {}; + return tl_node; +} + +} // namespace internal + +template<typename Ctx> +inline NodeRef<typename Ctx::Key> FromString(const std::string& str, const Ctx& ctx) { + return internal::Parse<typename Ctx::Key>(str, ctx); +} + +} // namespace miniscript + +#endif // BITCOIN_SCRIPT_MINISCRIPT_H |