/* Copyright (c) 2021, 2024, Oracle and/or its affiliates. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License, version 2.0, as published by the Free Software Foundation. This program is designed to work with certain software (including but not limited to OpenSSL) that is licensed under separate terms, as designated in a particular file or component or in included license documentation. The authors of MySQL hereby grant you an additional permission to link the program and your derivative works with the separately licensed software that they have either included with the program or referenced in the documentation. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License, version 2.0, for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ #ifndef SQL_JOIN_OPTIMIZER_ONLINE_CYCLE_FINDER_H_ #define SQL_JOIN_OPTIMIZER_ONLINE_CYCLE_FINDER_H_ #include "map_helpers.h" #include "sql/mem_root_array.h" #include "sql/sql_array.h" struct MEM_ROOT; /** A fast online cycle finder, based on [Pea03]. It keeps a DAG in memory, built up incrementally, and is able to reject adding edges that would create cycles (or, equivalently, test if adding an edge would create a cycle). The amortized cost of checking ϴ(E) insertions is O(V). The basic working of the algorithm is to keep a list of all vertices, topologically sorted given the order so far. When inserting a new edge, we can quickly identify any vertices that would need to be moved in the topological sort (they are the ones stored between the two endpoints), run a DFS, and see if moving them would cause a contradiction (and thus, a cycle). See EdgeWouldCreateCycle() or the paper for more details. Note that confusingly enough, when used from the graph simplification algorithm, the vertices in this graph represent hyperedges (joins) in the join hypergraph, _not_ the vertices (tables) themselves. The edges in this graph are happens-before relations between those joins. [Pea03] Pearce et al: “Online Cycle Detection and Difference Propagation for Pointer Analysis”, section 3.2. */ class OnlineCycleFinder { public: OnlineCycleFinder(MEM_ROOT *mem_root, int num_vertices); // Returns true iff this would create a cycle. bool EdgeWouldCreateCycle(int a_idx, int b_idx); // Adds edge A -> B (A must be before B). // Returns true iff this would create a cycle. bool AddEdge(int a_idx, int b_idx); // Remove edge A -> B. The edge must have been added earlier with AddEdge // (or we will assert-fail). void DeleteEdge(int a_idx, int b_idx); // Returns a topological sort, respecting the added edges. // Note that the ordering is entirely arbitrary except for that, // and can be changed by e.g. EdgeWouldCreateCycle() calls. Bounds_checked_array order() const { return m_order; } private: bool DepthFirstSearch(int node_idx, int upper_bound, int node_idx_to_avoid); void MoveAllMarked(int start_pos, int new_pos); void Allocate(int node_idx, int index_in_order) { m_order[index_in_order] = node_idx; m_position_of_node[node_idx] = index_in_order; } // List of nodes, in topological order. Called i2n in the paper. Bounds_checked_array m_order; // For each node index, where in m_order is it? // Called n2i in the paper. Bounds_checked_array m_position_of_node; // For each node, was it seen during this search or not? Bounds_checked_array m_visited; // Used as a temporary during MoveAllMarked(). Mem_root_array m_to_shift; // All edges that have been added, keyed by index of the from-node. mem_root_unordered_multimap m_edges; }; #endif // SQL_JOIN_OPTIMIZER_ONLINE_CYCLE_FINDER_H_