cluster_linearize.h raw

   1  // Copyright (c) The Limenka developers
   2  // Distributed under the MIT software license, see the accompanying
   3  // file COPYING or http://www.opensource.org/licenses/mit-license.php.
   4  
   5  #ifndef LIMENKA_TEST_UTIL_CLUSTER_LINEARIZE_H
   6  #define LIMENKA_TEST_UTIL_CLUSTER_LINEARIZE_H
   7  
   8  #include <cluster_linearize.h>
   9  #include <serialize.h>
  10  #include <span.h>
  11  #include <streams.h>
  12  #include <util/bitset.h>
  13  #include <util/feefrac.h>
  14  
  15  #include <stdint.h>
  16  #include <numeric>
  17  #include <vector>
  18  #include <utility>
  19  
  20  namespace {
  21  
  22  using namespace cluster_linearize;
  23  
  24  using TestBitSet = BitSet<32>;
  25  
  26  /** Check if a graph is acyclic. */
  27  template<typename SetType>
  28  bool IsAcyclic(const DepGraph<SetType>& depgraph) noexcept
  29  {
  30      for (ClusterIndex i : depgraph.Positions()) {
  31          if ((depgraph.Ancestors(i) & depgraph.Descendants(i)) != SetType::Singleton(i)) {
  32              return false;
  33          }
  34      }
  35      return true;
  36  }
  37  
  38  /** A formatter for a bespoke serialization for acyclic DepGraph objects.
  39   *
  40   * The serialization format outputs information about transactions in a topological order (parents
  41   * before children), together with position information so transactions can be moved back to their
  42   * correct position on deserialization.
  43   *
  44   * - For each transaction t in the DepGraph (in some topological order);
  45   *   - The size: VARINT(t.size), which cannot be 0.
  46   *   - The fee: VARINT(SignedToUnsigned(t.fee)), see below for SignedToUnsigned.
  47   *   - For each direct dependency:
  48   *     - VARINT(skip)
  49   *   - The position of t in the cluster: VARINT(skip)
  50   * - The end of the graph: VARINT(0)
  51   *
  52   * The list of skip values encodes the dependencies of t, as well as its position in the cluster.
  53   * Each skip value is the number of possibilities that were available, but were not taken. These
  54   * possibilities are, in order:
  55   * - For each previous transaction in the graph, in reverse serialization order, whether it is a
  56   *   direct parent of t (but excluding transactions which are already implied to be dependencies
  57   *   by parent relations that were serialized before it).
  58   * - The various insertion positions in the cluster, from the very end of the cluster, to the
  59   *   front.
  60   * - The appending of 1, 2, 3, ... holes at the end of the cluster, followed by appending the new
  61   *   transaction.
  62   *
  63   * Let's say you have a 7-transaction cluster, consisting of transactions F,A,C,B,_,G,E,_,D
  64   * (where _ represent holes; unused positions within the DepGraph) but serialized in order
  65   * A,B,C,D,E,F,G, because that happens to be a topological ordering. By the time G gets serialized,
  66   * what has been serialized already represents the cluster F,A,C,B,_,E,_,D (in that order). G has B
  67   * and E as direct parents, and E depends on C.
  68   *
  69   * In this case, the possibilities are, in order:
  70   * - [ ] the dependency G->F
  71   * - [X] the dependency G->E
  72   * - [ ] the dependency G->D
  73   * - [X] the dependency G->B
  74   * - [ ] the dependency G->A
  75   * - [ ] put G at the end of the cluster
  76   * - [ ] put G before D
  77   * - [ ] put G before the hole before D
  78   * - [X] put G before E
  79   * - [ ] put G before the hole before E
  80   * - [ ] put G before B
  81   * - [ ] put G before C
  82   * - [ ] put G before A
  83   * - [ ] put G before F
  84   * - [ ] add 1 hole at the end of the cluster, followed by G
  85   * - [ ] add 2 holes at the end of the cluster, followed by G
  86   * - [ ] add ...
  87   *
  88   * The skip values in this case are 1 (G->F), 1 (G->D), 4 (G->A, G at end, G before D, G before
  89   * hole). No skip after 4 is needed (or permitted), because there can only be one position for G.
  90   * Also note that G->C is not included in the list of possibilities, as it is implied by the
  91   * included G->E and E->C that came before it. On deserialization, if the last skip value was 8 or
  92   * larger (putting G before the beginning of the cluster), it is interpreted as wrapping around
  93   * back to the end.
  94   *
  95   *
  96   * Rationale:
  97   * - Why VARINTs? They are flexible enough to represent large numbers where needed, but more
  98   *   compact for smaller numbers. The serialization format is designed so that simple structures
  99   *   involve smaller numbers, so smaller size maps to simpler graphs.
 100   * - Why use SignedToUnsigned? It results in small unsigned values for signed values with small
 101   *   absolute value. This way we can encode negative fees in graphs, but still let small negative
 102   *   numbers have small encodings.
 103   * - Why are the parents emitted in reverse order compared to the transactions themselves? This
 104   *   naturally lets us skip parents-of-parents, as they will be reflected as implied dependencies.
 105   * - Why encode skip values and not a bitmask to convey the list positions? It turns out that the
 106   *   most complex graphs (in terms of linearization complexity) are ones with ~1 dependency per
 107   *   transaction. The current encoding uses ~1 byte per transaction for dependencies in this case,
 108   *   while a bitmask would require ~N/2 bits per transaction.
 109   */
 110  
 111  struct DepGraphFormatter
 112  {
 113      /** Convert x>=0 to 2x (even), x<0 to -2x-1 (odd). */
 114      [[maybe_unused]] static uint64_t SignedToUnsigned(int64_t x) noexcept
 115      {
 116          if (x < 0) {
 117              return 2 * uint64_t(-(x + 1)) + 1;
 118          } else {
 119              return 2 * uint64_t(x);
 120          }
 121      }
 122  
 123      /** Convert even x to x/2 (>=0), odd x to -(x/2)-1 (<0). */
 124      [[maybe_unused]] static int64_t UnsignedToSigned(uint64_t x) noexcept
 125      {
 126          if (x & 1) {
 127              return -int64_t(x / 2) - 1;
 128          } else {
 129              return int64_t(x / 2);
 130          }
 131      }
 132  
 133      template <typename Stream, typename SetType>
 134      static void Ser(Stream& s, const DepGraph<SetType>& depgraph)
 135      {
 136          /** Construct a topological order to serialize the transactions in. */
 137          std::vector<ClusterIndex> topo_order;
 138          topo_order.reserve(depgraph.TxCount());
 139          for (auto i : depgraph.Positions()) topo_order.push_back(i);
 140          std::sort(topo_order.begin(), topo_order.end(), [&](ClusterIndex a, ClusterIndex b) {
 141              auto anc_a = depgraph.Ancestors(a).Count(), anc_b = depgraph.Ancestors(b).Count();
 142              if (anc_a != anc_b) return anc_a < anc_b;
 143              return a < b;
 144          });
 145  
 146          /** Which positions (incl. holes) the deserializer already knows when it has deserialized
 147           *  what has been serialized here so far. */
 148          SetType done;
 149  
 150          // Loop over the transactions in topological order.
 151          for (ClusterIndex topo_idx = 0; topo_idx < topo_order.size(); ++topo_idx) {
 152              /** Which depgraph index we are currently writing. */
 153              ClusterIndex idx = topo_order[topo_idx];
 154              // Write size, which must be larger than 0.
 155              s << VARINT_MODE(depgraph.FeeRate(idx).size, VarIntMode::NONNEGATIVE_SIGNED);
 156              // Write fee, encoded as an unsigned varint (odd=negative, even=non-negative).
 157              s << VARINT(SignedToUnsigned(depgraph.FeeRate(idx).fee));
 158              // Write dependency information.
 159              SetType written_parents;
 160              uint64_t diff = 0; //!< How many potential parent/child relations we have skipped over.
 161              for (ClusterIndex dep_dist = 0; dep_dist < topo_idx; ++dep_dist) {
 162                  /** Which depgraph index we are currently considering as parent of idx. */
 163                  ClusterIndex dep_idx = topo_order[topo_idx - 1 - dep_dist];
 164                  // Ignore transactions which are already known to be ancestors.
 165                  if (depgraph.Descendants(dep_idx).Overlaps(written_parents)) continue;
 166                  if (depgraph.Ancestors(idx)[dep_idx]) {
 167                      // When an actual parent is encountered, encode how many non-parents were skipped
 168                      // before it.
 169                      s << VARINT(diff);
 170                      diff = 0;
 171                      written_parents.Set(dep_idx);
 172                  } else {
 173                      // When a non-parent is encountered, increment the skip counter.
 174                      ++diff;
 175                  }
 176              }
 177              // Write position information.
 178              auto add_holes = SetType::Fill(idx) - done - depgraph.Positions();
 179              if (add_holes.None()) {
 180                  // The new transaction is to be inserted N positions back from the end of the
 181                  // cluster. Emit N to indicate that that many insertion choices are skipped.
 182                  auto skips = (done - SetType::Fill(idx)).Count();
 183                  s << VARINT(diff + skips);
 184              } else {
 185                  // The new transaction is to be appended at the end of the cluster, after N holes.
 186                  // Emit current_cluster_size + N, to indicate all insertion choices are skipped,
 187                  // plus N possibilities for the number of holes.
 188                  s << VARINT(diff + done.Count() + add_holes.Count());
 189                  done |= add_holes;
 190              }
 191              done.Set(idx);
 192          }
 193  
 194          // Output a final 0 to denote the end of the graph.
 195          s << uint8_t{0};
 196      }
 197  
 198      template <typename Stream, typename SetType>
 199      void Unser(Stream& s, DepGraph<SetType>& depgraph)
 200      {
 201          /** The dependency graph which we deserialize into first, with transactions in
 202           *  topological serialization order, not original cluster order. */
 203          DepGraph<SetType> topo_depgraph;
 204          /** Mapping from serialization order to cluster order, used later to reconstruct the
 205           *  cluster order. */
 206          std::vector<ClusterIndex> reordering;
 207          /** How big the entries vector in the reconstructed depgraph will be (including holes). */
 208          ClusterIndex total_size{0};
 209  
 210          // Read transactions in topological order.
 211          while (true) {
 212              FeeFrac new_feerate; //!< The new transaction's fee and size.
 213              SetType new_ancestors; //!< The new transaction's ancestors (excluding itself).
 214              uint64_t diff{0}; //!< How many potential parents/insertions we have to skip.
 215              bool read_error{false};
 216              try {
 217                  // Read size. Size 0 signifies the end of the DepGraph.
 218                  int32_t size;
 219                  s >> VARINT_MODE(size, VarIntMode::NONNEGATIVE_SIGNED);
 220                  size &= 0x3FFFFF; // Enough for size up to 4M.
 221                  static_assert(0x3FFFFF >= 4000000);
 222                  if (size == 0 || topo_depgraph.TxCount() == SetType::Size()) break;
 223                  // Read fee, encoded as an unsigned varint (odd=negative, even=non-negative).
 224                  uint64_t coded_fee;
 225                  s >> VARINT(coded_fee);
 226                  coded_fee &= 0xFFFFFFFFFFFFF; // Enough for fee between -21M...21M BTC.
 227                  static_assert(0xFFFFFFFFFFFFF > uint64_t{2} * 21000000 * 100000000);
 228                  new_feerate = {UnsignedToSigned(coded_fee), size};
 229                  // Read dependency information.
 230                  auto topo_idx = reordering.size();
 231                  s >> VARINT(diff);
 232                  for (ClusterIndex dep_dist = 0; dep_dist < topo_idx; ++dep_dist) {
 233                      /** Which topo_depgraph index we are currently considering as parent of topo_idx. */
 234                      ClusterIndex dep_topo_idx = topo_idx - 1 - dep_dist;
 235                      // Ignore transactions which are already known ancestors of topo_idx.
 236                      if (new_ancestors[dep_topo_idx]) continue;
 237                      if (diff == 0) {
 238                          // When the skip counter has reached 0, add an actual dependency.
 239                          new_ancestors |= topo_depgraph.Ancestors(dep_topo_idx);
 240                          // And read the number of skips after it.
 241                          s >> VARINT(diff);
 242                      } else {
 243                          // Otherwise, dep_topo_idx is not a parent. Decrement and continue.
 244                          --diff;
 245                      }
 246                  }
 247              } catch (const std::ios_base::failure&) {
 248                  // Continue even if a read error was encountered.
 249                  read_error = true;
 250              }
 251              // Construct a new transaction whenever we made it past the new_feerate construction.
 252              if (new_feerate.IsEmpty()) break;
 253              assert(reordering.size() < SetType::Size());
 254              auto topo_idx = topo_depgraph.AddTransaction(new_feerate);
 255              topo_depgraph.AddDependencies(new_ancestors, topo_idx);
 256              if (total_size < SetType::Size()) {
 257                  // Normal case.
 258                  diff %= SetType::Size();
 259                  if (diff <= total_size) {
 260                      // Insert the new transaction at distance diff back from the end.
 261                      for (auto& pos : reordering) {
 262                          pos += (pos >= total_size - diff);
 263                      }
 264                      reordering.push_back(total_size++ - diff);
 265                  } else {
 266                      // Append diff - total_size holes at the end, plus the new transaction.
 267                      total_size = diff;
 268                      reordering.push_back(total_size++);
 269                  }
 270              } else {
 271                  // In case total_size == SetType::Size, it is not possible to insert the new
 272                  // transaction without exceeding SetType's size. Instead, interpret diff as an
 273                  // index into the holes, and overwrite a position there. This branch is never used
 274                  // when deserializing the output of the serializer, but gives meaning to otherwise
 275                  // invalid input.
 276                  diff %= (SetType::Size() - reordering.size());
 277                  SetType holes = SetType::Fill(SetType::Size());
 278                  for (auto pos : reordering) holes.Reset(pos);
 279                  for (auto pos : holes) {
 280                      if (diff == 0) {
 281                          reordering.push_back(pos);
 282                          break;
 283                      }
 284                      --diff;
 285                  }
 286              }
 287              // Stop if a read error was encountered during deserialization.
 288              if (read_error) break;
 289          }
 290  
 291          // Construct the original cluster order depgraph.
 292          depgraph = DepGraph(topo_depgraph, reordering, total_size);
 293      }
 294  };
 295  
 296  /** Perform a sanity/consistency check on a DepGraph. */
 297  template<typename SetType>
 298  void SanityCheck(const DepGraph<SetType>& depgraph)
 299  {
 300      // Verify Positions and PositionRange consistency.
 301      ClusterIndex num_positions{0};
 302      ClusterIndex position_range{0};
 303      for (ClusterIndex i : depgraph.Positions()) {
 304          ++num_positions;
 305          position_range = i + 1;
 306      }
 307      assert(num_positions == depgraph.TxCount());
 308      assert(position_range == depgraph.PositionRange());
 309      assert(position_range >= num_positions);
 310      assert(position_range <= SetType::Size());
 311      // Consistency check between ancestors internally.
 312      for (ClusterIndex i : depgraph.Positions()) {
 313          // Transactions include themselves as ancestors.
 314          assert(depgraph.Ancestors(i)[i]);
 315          // If a is an ancestor of b, then b's ancestors must include all of a's ancestors.
 316          for (auto a : depgraph.Ancestors(i)) {
 317              assert(depgraph.Ancestors(i).IsSupersetOf(depgraph.Ancestors(a)));
 318          }
 319      }
 320      // Consistency check between ancestors and descendants.
 321      for (ClusterIndex i : depgraph.Positions()) {
 322          for (ClusterIndex j : depgraph.Positions()) {
 323              assert(depgraph.Ancestors(i)[j] == depgraph.Descendants(j)[i]);
 324          }
 325          // No transaction is a parent or child of itself.
 326          auto parents = depgraph.GetReducedParents(i);
 327          auto children = depgraph.GetReducedChildren(i);
 328          assert(!parents[i]);
 329          assert(!children[i]);
 330          // Parents of a transaction do not have ancestors inside those parents (except itself).
 331          // Note that even the transaction itself may be missing (if it is part of a cycle).
 332          for (auto parent : parents) {
 333              assert((depgraph.Ancestors(parent) & parents).IsSubsetOf(SetType::Singleton(parent)));
 334          }
 335          // Similar for children and descendants.
 336          for (auto child : children) {
 337              assert((depgraph.Descendants(child) & children).IsSubsetOf(SetType::Singleton(child)));
 338          }
 339      }
 340      if (IsAcyclic(depgraph)) {
 341          // If DepGraph is acyclic, serialize + deserialize must roundtrip.
 342          std::vector<unsigned char> ser;
 343          VectorWriter writer(ser, 0);
 344          writer << Using<DepGraphFormatter>(depgraph);
 345          SpanReader reader(ser);
 346          DepGraph<TestBitSet> decoded_depgraph;
 347          reader >> Using<DepGraphFormatter>(decoded_depgraph);
 348          assert(depgraph == decoded_depgraph);
 349          assert(reader.empty());
 350          // It must also deserialize correctly without the terminal 0 byte (as the deserializer
 351          // will upon EOF still return what it read so far).
 352          assert(ser.size() >= 1 && ser.back() == 0);
 353          ser.pop_back();
 354          reader = SpanReader{ser};
 355          decoded_depgraph = {};
 356          reader >> Using<DepGraphFormatter>(decoded_depgraph);
 357          assert(depgraph == decoded_depgraph);
 358          assert(reader.empty());
 359  
 360          // In acyclic graphs, the union of parents with parents of parents etc. yields the
 361          // full ancestor set (and similar for children and descendants).
 362          std::vector<SetType> parents(depgraph.PositionRange()), children(depgraph.PositionRange());
 363          for (ClusterIndex i : depgraph.Positions()) {
 364              parents[i] = depgraph.GetReducedParents(i);
 365              children[i] = depgraph.GetReducedChildren(i);
 366          }
 367          for (auto i : depgraph.Positions()) {
 368              // Initialize the set of ancestors with just the current transaction itself.
 369              SetType ancestors = SetType::Singleton(i);
 370              // Iteratively add parents of all transactions in the ancestor set to itself.
 371              while (true) {
 372                  const auto old_ancestors = ancestors;
 373                  for (auto j : ancestors) ancestors |= parents[j];
 374                  // Stop when no more changes are being made.
 375                  if (old_ancestors == ancestors) break;
 376              }
 377              assert(ancestors == depgraph.Ancestors(i));
 378  
 379              // Initialize the set of descendants with just the current transaction itself.
 380              SetType descendants = SetType::Singleton(i);
 381              // Iteratively add children of all transactions in the descendant set to itself.
 382              while (true) {
 383                  const auto old_descendants = descendants;
 384                  for (auto j : descendants) descendants |= children[j];
 385                  // Stop when no more changes are being made.
 386                  if (old_descendants == descendants) break;
 387              }
 388              assert(descendants == depgraph.Descendants(i));
 389          }
 390      }
 391  }
 392  
 393  /** Perform a sanity check on a linearization. */
 394  template<typename SetType>
 395  void SanityCheck(const DepGraph<SetType>& depgraph, Span<const ClusterIndex> linearization)
 396  {
 397      // Check completeness.
 398      assert(linearization.size() == depgraph.TxCount());
 399      TestBitSet done;
 400      for (auto i : linearization) {
 401          // Check transaction position is in range.
 402          assert(depgraph.Positions()[i]);
 403          // Check topology and lack of duplicates.
 404          assert((depgraph.Ancestors(i) - done) == TestBitSet::Singleton(i));
 405          done.Set(i);
 406      }
 407  }
 408  
 409  } // namespace
 410  
 411  #endif // LIMENKA_TEST_UTIL_CLUSTER_LINEARIZE_H
 412