AhoCorasick: Track generation nodes are added

Author: Joseph-Edwards

include/libsemigroups/detail/aho-corasick-impl.hpp

@@ -65,6 +65,7 @@ namespace libsemigroups {
         // Private data
         ////////////////////////////////////////////////////////////////////////
        private:
+        mutable size_t                 _last_checked;
         uint32_t                       _height;
         index_type                     _link;
         index_type                     _parent;
@@ -97,6 +98,10 @@ namespace libsemigroups {
         // Getters - public
         ////////////////////////////////////////////////////////////////////////
 
+        [[nodiscard]] size_t last_checked() const noexcept {
+          return _last_checked;
+        }
+
         [[nodiscard]] size_t height() const noexcept {
           return _height;
         }
@@ -126,12 +131,22 @@ namespace libsemigroups {
           return _value;
         }
 
+        ////////////////////////////////////////////////////////////////////////
+        // Setters - public
+        ////////////////////////////////////////////////////////////////////////
+
+        Node const& last_checked(size_t val) const noexcept {
+          _last_checked = val;
+          return *this;
+        }
+
        private:
         ////////////////////////////////////////////////////////////////////////
         // Setters - private
         ////////////////////////////////////////////////////////////////////////
 
-        // All setters are private to avoid corrupting the objects.
+        // All setters of non-mutable members are private to avoid corrupting
+        // the objects.
 
         Node const& height(size_t val) noexcept {
           _height = val;
@@ -162,6 +177,7 @@ namespace libsemigroups {
       std::vector<index_type>           _inactive_nodes_index;
       std::vector<index_type>           _node_indices_to_update;
       std::unordered_set<index_type>    _terminal_nodes_index;
+      mutable size_t                    _generation;
 
       // TODO(1): it seems likely that the positions of the active nodes in
       // _all_nodes will become scattered and disordered over time, and so it'd
@@ -187,6 +203,14 @@ namespace libsemigroups {
 
       ~AhoCorasickImpl();
 
+      void increment_generation() const noexcept {
+        ++_generation;
+      }
+
+      size_t generation() const noexcept {
+        return _generation;
+      }
+
       size_t alphabet_size() const noexcept {
         return _children.number_of_cols();
       }
@@ -407,7 +431,7 @@ namespace libsemigroups {
                                               Word const&            w);
 
     }  // namespace aho_corasick_impl
-  }  // namespace detail
+  }    // namespace detail
 }  // namespace libsemigroups
 
 #include "aho-corasick-impl.tpp"

include/libsemigroups/detail/aho-corasick-impl.tpp

@@ -29,6 +29,7 @@ namespace libsemigroups {
       LIBSEMIGROUPS_ASSERT(val != nullptr);
       index_type current = root;
       for (auto it = first; it != last; ++it) {
+        _all_nodes[current].last_checked(generation());
         index_type next = _children.get(current, *it);
         if (next == UNDEFINED) {
           // index of next node added
@@ -40,7 +41,7 @@ namespace libsemigroups {
       if (inserted) {
         _all_nodes[current].value(val);
       }
-
+      _all_nodes[current].last_checked(generation());
       return {current, inserted};
     }
 
@@ -309,5 +310,5 @@ namespace libsemigroups {
       }
 
     }  // namespace aho_corasick_impl
-  }  // namespace detail
+  }    // namespace detail
 }  // namespace libsemigroups

src/detail/aho-corasick-impl.cpp

@@ -39,7 +39,8 @@ namespace libsemigroups {
     ////////////////////////////////////////////////////////////////////////
 
     AhoCorasickImpl::Node::Node(index_type parent, letter_type a)
-        : _height(),
+        : _last_checked(0),
+          _height(),
           _link(),
           _parent(),
           _parent_letter(),
@@ -50,7 +51,8 @@ namespace libsemigroups {
 
     typename AhoCorasickImpl::Node&
     AhoCorasickImpl::Node::init(index_type parent, letter_type a) noexcept {
-      _height = parent == UNDEFINED ? 0 : UNDEFINED;
+      _height       = parent == UNDEFINED ? 0 : UNDEFINED;
+      _last_checked = 0;
       if (_parent == root || _parent == UNDEFINED) {
         _link = root;
       } else {
@@ -77,7 +79,8 @@ namespace libsemigroups {
           _active_nodes_index({root}),
           _inactive_nodes_index(),
           _node_indices_to_update(),
-          _terminal_nodes_index() {}
+          _terminal_nodes_index(),
+          _generation(0) {}
 
     AhoCorasickImpl& AhoCorasickImpl::init() {
       init(0);
@@ -99,7 +102,8 @@ namespace libsemigroups {
           _active_nodes_index({root}),
           _inactive_nodes_index(),
           _node_indices_to_update(),
-          _terminal_nodes_index() {}
+          _terminal_nodes_index(),
+          _generation(0){};
 
     AhoCorasickImpl& AhoCorasickImpl::init(size_t num_letters) {
       LIBSEMIGROUPS_ASSERT(!_all_nodes.empty());
@@ -124,6 +128,7 @@ namespace libsemigroups {
                                + _inactive_nodes_index.size()
                            == _all_nodes.size());
       LIBSEMIGROUPS_ASSERT(_children.number_of_rows() == _all_nodes.size());
+      _generation = 0;
 
       return *this;
     }