haskell-igraph-0.8.0: igraph/include/bliss/graph.hh
#ifndef BLISS_GRAPH_HH
#define BLISS_GRAPH_HH
/*
Copyright (c) 2003-2015 Tommi Junttila
Released under the GNU Lesser General Public License version 3.
This file is part of bliss.
bliss is free software: you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation, version 3 of the License.
bliss 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 Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with bliss. If not, see <http://www.gnu.org/licenses/>.
*/
/**
* \namespace bliss
* The namespace bliss contains all the classes and functions of the bliss
* tool except for the C programming language API.
*/
namespace bliss {
class AbstractGraph;
}
#include <cstdio>
#include <vector>
#include "kstack.hh"
#include "kqueue.hh"
#include "heap.hh"
#include "orbit.hh"
#include "partition.hh"
#include "bignum.hh"
#include "uintseqhash.hh"
namespace bliss {
/**
* \brief Statistics returned by the bliss search algorithm.
*/
class Stats
{
friend class AbstractGraph;
public:
/** \internal The size of the automorphism group. */
BigNum group_size;
private:
/** \internal An approximation (due to possible overflows) of
* the size of the automorphism group. */
long double group_size_approx;
/** \internal The number of nodes in the search tree. */
long unsigned int nof_nodes;
/** \internal The number of leaf nodes in the search tree. */
long unsigned int nof_leaf_nodes;
/** \internal The number of bad nodes in the search tree. */
long unsigned int nof_bad_nodes;
/** \internal The number of canonical representative updates. */
long unsigned int nof_canupdates;
/** \internal The number of generator permutations. */
long unsigned int nof_generators;
/** \internal The maximal depth of the search tree. */
unsigned long int max_level;
/** */
void reset()
{
group_size.assign(1);
group_size_approx = 1.0;
nof_nodes = 0;
nof_leaf_nodes = 0;
nof_bad_nodes = 0;
nof_canupdates = 0;
nof_generators = 0;
max_level = 0;
}
public:
Stats() { reset(); }
/** Print the statistics. */
size_t print(FILE* const fp) const
{
size_t r = 0;
r += fprintf(fp, "Nodes: %lu\n", nof_nodes);
r += fprintf(fp, "Leaf nodes: %lu\n", nof_leaf_nodes);
r += fprintf(fp, "Bad nodes: %lu\n", nof_bad_nodes);
r += fprintf(fp, "Canrep updates: %lu\n", nof_canupdates);
r += fprintf(fp, "Generators: %lu\n", nof_generators);
r += fprintf(fp, "Max level: %lu\n", max_level);
r += fprintf(fp, "|Aut|: ")+group_size.print(fp)+fprintf(fp, "\n");
fflush(fp);
return r;
}
/** An approximation (due to possible overflows/rounding errors) of
* the size of the automorphism group. */
long double get_group_size_approx() const {return group_size_approx;}
/** The number of nodes in the search tree. */
long unsigned int get_nof_nodes() const {return nof_nodes;}
/** The number of leaf nodes in the search tree. */
long unsigned int get_nof_leaf_nodes() const {return nof_leaf_nodes;}
/** The number of bad nodes in the search tree. */
long unsigned int get_nof_bad_nodes() const {return nof_bad_nodes;}
/** The number of canonical representative updates. */
long unsigned int get_nof_canupdates() const {return nof_canupdates;}
/** The number of generator permutations. */
long unsigned int get_nof_generators() const {return nof_generators;}
/** The maximal depth of the search tree. */
unsigned long int get_max_level() const {return max_level;}
};
/**
* \brief An abstract base class for different types of graphs.
*/
class AbstractGraph
{
friend class Partition;
public:
AbstractGraph();
virtual ~AbstractGraph();
/**
* Set the verbose output level for the algorithms.
* \param level the level of verbose output, 0 means no verbose output
*/
void set_verbose_level(const unsigned int level);
/**
* Set the file stream for the verbose output.
* \param fp the file stream; if null, no verbose output is written
*/
void set_verbose_file(FILE * const fp);
/**
* Add a new vertex with color \a color in the graph and return its index.
*/
virtual unsigned int add_vertex(const unsigned int color = 0) = 0;
/**
* Add an edge between vertices \a source and \a target.
* Duplicate edges between vertices are ignored but try to avoid introducing
* them in the first place as they are not ignored immediately but will
* consume memory and computation resources for a while.
*/
virtual void add_edge(const unsigned int source, const unsigned int target) = 0;
/**
* Change the color of the vertex \a vertex to \a color.
*/
virtual void change_color(const unsigned int vertex, const unsigned int color) = 0;
/**
* Check whether \a perm is an automorphism of this graph.
* Unoptimized, mainly for debugging purposes.
*/
virtual bool is_automorphism(const std::vector<unsigned int>& perm) const;
/** Activate/deactivate failure recording.
* May not be called during the search, i.e. from an automorphism reporting
* hook function.
* \param active if true, activate failure recording, deactivate otherwise
*/
void set_failure_recording(const bool active) {assert(!in_search); opt_use_failure_recording = active;}
/** Activate/deactivate component recursion.
* The choice affects the computed canonical labelings;
* therefore, if you want to compare whether two graphs are isomorphic by
* computing and comparing (for equality) their canonical versions,
* be sure to use the same choice for both graphs.
* May not be called during the search, i.e. from an automorphism reporting
* hook function.
* \param active if true, activate component recursion, deactivate otherwise
*/
void set_component_recursion(const bool active) {assert(!in_search); opt_use_comprec = active;}
/**
* Return the number of vertices in the graph.
*/
virtual unsigned int get_nof_vertices() const = 0;
/**
* Return a new graph that is the result of applying the permutation \a perm
* to this graph. This graph is not modified.
* \a perm must contain N=this.get_nof_vertices() elements and be a bijection
* on {0,1,...,N-1}, otherwise the result is undefined or a segfault.
*/
virtual AbstractGraph* permute(const unsigned int* const perm) const = 0;
virtual AbstractGraph* permute(const std::vector<unsigned int>& perm) const = 0;
/**
* Find a set of generators for the automorphism group of the graph.
* The function \a hook (if non-null) is called each time a new generator
* for the automorphism group is found.
* The first argument \a user_param for the hook is the
* \a hook_user_param given below,
* the second argument \a n is the length of the automorphism (equal to
* get_nof_vertices()) and
* the third argument \a aut is the automorphism
* (a bijection on {0,...,get_nof_vertices()-1}).
* The memory for the automorphism \a aut will be invalidated immediately
* after the return from the hook function;
* if you want to use the automorphism later, you have to take a copy of it.
* Do not call any member functions in the hook.
* The search statistics are copied in \a stats.
*/
void find_automorphisms(Stats& stats,
void (*hook)(void* user_param,
unsigned int n,
const unsigned int* aut),
void* hook_user_param);
/**
* Otherwise the same as find_automorphisms() except that
* a canonical labeling of the graph (a bijection on
* {0,...,get_nof_vertices()-1}) is returned.
* The memory allocated for the returned canonical labeling will remain
* valid only until the next call to a member function with the exception
* that constant member functions (for example, bliss::Graph::permute()) can
* be called without invalidating the labeling.
* To compute the canonical version of an undirected graph, call this
* function and then bliss::Graph::permute() with the returned canonical
* labeling.
* Note that the computed canonical version may depend on the applied version
* of bliss as well as on some other options (for instance, the splitting
* heuristic selected with bliss::Graph::set_splitting_heuristic()).
*/
const unsigned int* canonical_form(Stats& stats,
void (*hook)(void* user_param,
unsigned int n,
const unsigned int* aut),
void* hook_user_param);
/**
* Write the graph to a file in a variant of the DIMACS format.
* See the <A href="http://www.tcs.hut.fi/Software/bliss/">bliss website</A>
* for the definition of the file format.
* Note that in the DIMACS file the vertices are numbered from 1 to N while
* in this C++ API they are from 0 to N-1.
* Thus the vertex n in the file corresponds to the vertex n-1 in the API.
* \param fp the file stream where the graph is written
*/
virtual void write_dimacs(FILE * const fp) = 0;
/**
* Write the graph to a file in the graphviz dotty format.
* \param fp the file stream where the graph is written
*/
virtual void write_dot(FILE * const fp) = 0;
/**
* Write the graph in a file in the graphviz dotty format.
* Do nothing if the file cannot be written.
* \param file_name the name of the file to which the graph is written
*/
virtual void write_dot(const char * const file_name) = 0;
/**
* Get a hash value for the graph.
* \return the hash value
*/
virtual unsigned int get_hash() = 0;
/**
* Disable/enable the "long prune" method.
* The choice affects the computed canonical labelings;
* therefore, if you want to compare whether two graphs are isomorphic by
* computing and comparing (for equality) their canonical versions,
* be sure to use the same choice for both graphs.
* May not be called during the search, i.e. from an automorphism reporting
* hook function.
* \param active if true, activate "long prune", deactivate otherwise
*/
void set_long_prune_activity(const bool active) {
assert(!in_search);
opt_use_long_prune = active;
}
protected:
/** \internal
* How much verbose output is produced (0 means none) */
unsigned int verbose_level;
/** \internal
* The output stream for verbose output. */
FILE *verbstr;
protected:
/** \internal
* The ordered partition used in the search algorithm. */
Partition p;
/** \internal
* Whether the search for automorphisms and a canonical labeling is
* in progress.
*/
bool in_search;
/** \internal
* Is failure recording in use?
*/
bool opt_use_failure_recording;
/* The "tree-specific" invariant value for the point when current path
* got different from the first path */
unsigned int failure_recording_fp_deviation;
/** \internal
* Is component recursion in use?
*/
bool opt_use_comprec;
unsigned int refine_current_path_certificate_index;
bool refine_compare_certificate;
bool refine_equal_to_first;
unsigned int refine_first_path_subcertificate_end;
int refine_cmp_to_best;
unsigned int refine_best_path_subcertificate_end;
static const unsigned int CERT_SPLIT = 0; //UINT_MAX;
static const unsigned int CERT_EDGE = 1; //UINT_MAX-1;
/** \internal
* Add a triple (v1,v2,v3) in the certificate.
* May modify refine_equal_to_first and refine_cmp_to_best.
* May also update eqref_hash and failure_recording_fp_deviation. */
void cert_add(const unsigned int v1,
const unsigned int v2,
const unsigned int v3);
/** \internal
* Add a redundant triple (v1,v2,v3) in the certificate.
* Can also just dicard the triple.
* May modify refine_equal_to_first and refine_cmp_to_best.
* May also update eqref_hash and failure_recording_fp_deviation. */
void cert_add_redundant(const unsigned int x,
const unsigned int y,
const unsigned int z);
/**\internal
* Is the long prune method in use?
*/
bool opt_use_long_prune;
/**\internal
* Maximum amount of memory (in megabytes) available for
* the long prune method
*/
static const unsigned int long_prune_options_max_mem = 50;
/**\internal
* Maximum amount of automorphisms stored for the long prune method;
* less than this is stored if the memory limit above is reached first
*/
static const unsigned int long_prune_options_max_stored_auts = 100;
unsigned int long_prune_max_stored_autss;
std::vector<std::vector<bool> *> long_prune_fixed;
std::vector<std::vector<bool> *> long_prune_mcrs;
std::vector<bool> long_prune_temp;
unsigned int long_prune_begin;
unsigned int long_prune_end;
/** \internal
* Initialize the "long prune" data structures.
*/
void long_prune_init();
/** \internal
* Release the memory allocated for "long prune" data structures.
*/
void long_prune_deallocate();
void long_prune_add_automorphism(const unsigned int *aut);
std::vector<bool>& long_prune_get_fixed(const unsigned int index);
std::vector<bool>& long_prune_allocget_fixed(const unsigned int index);
std::vector<bool>& long_prune_get_mcrs(const unsigned int index);
std::vector<bool>& long_prune_allocget_mcrs(const unsigned int index);
/** \internal
* Swap the i:th and j:th stored automorphism information;
* i and j must be "in window, i.e. in [long_prune_begin,long_prune_end[
*/
void long_prune_swap(const unsigned int i, const unsigned int j);
/*
* Data structures and routines for refining the partition p into equitable
*/
Heap neighbour_heap;
virtual bool split_neighbourhood_of_unit_cell(Partition::Cell *) = 0;
virtual bool split_neighbourhood_of_cell(Partition::Cell * const) = 0;
void refine_to_equitable();
void refine_to_equitable(Partition::Cell * const unit_cell);
void refine_to_equitable(Partition::Cell * const unit_cell1,
Partition::Cell * const unit_cell2);
/** \internal
* \return false if it was detected that the current certificate
* is different from the first and/or best (whether this is checked
* depends on in_search and refine_compare_certificate flags.
*/
bool do_refine_to_equitable();
unsigned int eqref_max_certificate_index;
/** \internal
* Whether eqref_hash is updated during equitable refinement process.
*/
bool compute_eqref_hash;
UintSeqHash eqref_hash;
/** \internal
* Check whether the current partition p is equitable.
* Performance: very slow, use only for debugging purposes.
*/
virtual bool is_equitable() const = 0;
unsigned int *first_path_labeling;
unsigned int *first_path_labeling_inv;
Orbit first_path_orbits;
unsigned int *first_path_automorphism;
unsigned int *best_path_labeling;
unsigned int *best_path_labeling_inv;
Orbit best_path_orbits;
unsigned int *best_path_automorphism;
void update_labeling(unsigned int * const lab);
void update_labeling_and_its_inverse(unsigned int * const lab,
unsigned int * const lab_inv);
void update_orbit_information(Orbit &o, const unsigned int *perm);
void reset_permutation(unsigned int *perm);
/* Mainly for debugging purposes */
virtual bool is_automorphism(unsigned int* const perm);
std::vector<unsigned int> certificate_current_path;
std::vector<unsigned int> certificate_first_path;
std::vector<unsigned int> certificate_best_path;
unsigned int certificate_index;
virtual void initialize_certificate() = 0;
virtual void remove_duplicate_edges() = 0;
virtual void make_initial_equitable_partition() = 0;
virtual Partition::Cell* find_next_cell_to_be_splitted(Partition::Cell *cell) = 0;
void search(const bool canonical, Stats &stats);
void (*report_hook)(void *user_param,
unsigned int n,
const unsigned int *aut);
void *report_user_param;
/*
*
* Nonuniform component recursion (NUCR)
*
*/
/** The currently traversed component */
unsigned int cr_level;
/** \internal
* The "Component End Point" data structure
*/
class CR_CEP {
public:
/** At which level in the search was this CEP created */
unsigned int creation_level;
/** The current component has been fully traversed when the partition has
* this many discrete cells left */
unsigned int discrete_cell_limit;
/** The component to be traversed after the current one */
unsigned int next_cr_level;
/** The next component end point */
unsigned int next_cep_index;
bool first_checked;
bool best_checked;
};
/** \internal
* A stack for storing Component End Points
*/
std::vector<CR_CEP> cr_cep_stack;
/** \internal
* Find the first non-uniformity component at the component recursion
* level \a level.
* The component is stored in \a cr_component.
* If no component is found, \a cr_component is empty.
* Returns false if all the cells in the component recursion level \a level
* were discrete.
* Modifies the max_ival and max_ival_count fields of Partition:Cell
* (assumes that they are 0 when called and
* quarantees that they are 0 when returned).
*/
virtual bool nucr_find_first_component(const unsigned int level) = 0;
virtual bool nucr_find_first_component(const unsigned int level,
std::vector<unsigned int>& component,
unsigned int& component_elements,
Partition::Cell*& sh_return) = 0;
/** \internal
* The non-uniformity component found by nucr_find_first_component()
* is stored here.
*/
std::vector<unsigned int> cr_component;
/** \internal
* The number of vertices in the component \a cr_component
*/
unsigned int cr_component_elements;
};
/**
* \brief The class for undirected, vertex colored graphs.
*
* Multiple edges between vertices are not allowed (i.e., are ignored).
*/
class Graph : public AbstractGraph
{
public:
/**
* The possible splitting heuristics.
* The selected splitting heuristics affects the computed canonical
* labelings; therefore, if you want to compare whether two graphs
* are isomorphic by computing and comparing (for equality) their
* canonical versions, be sure to use the same splitting heuristics
* for both graphs.
*/
typedef enum {
/** First non-unit cell.
* Very fast but may result in large search spaces on difficult graphs.
* Use for large but easy graphs. */
shs_f = 0,
/** First smallest non-unit cell.
* Fast, should usually produce smaller search spaces than shs_f. */
shs_fs,
/** First largest non-unit cell.
* Fast, should usually produce smaller search spaces than shs_f. */
shs_fl,
/** First maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_fm,
/** First smallest maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_fsm,
/** First largest maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_flm
} SplittingHeuristic;
protected:
class Vertex {
public:
Vertex();
~Vertex();
void add_edge(const unsigned int other_vertex);
void remove_duplicate_edges(std::vector<bool>& tmp);
void sort_edges();
unsigned int color;
std::vector<unsigned int> edges;
unsigned int nof_edges() const {return edges.size(); }
};
std::vector<Vertex> vertices;
void sort_edges();
void remove_duplicate_edges();
/** \internal
* Partition independent invariant.
* Returns the color of the vertex.
* Time complexity: O(1).
*/
static unsigned int vertex_color_invariant(const Graph* const g,
const unsigned int v);
/** \internal
* Partition independent invariant.
* Returns the degree of the vertex.
* DUPLICATE EDGES MUST HAVE BEEN REMOVED BEFORE.
* Time complexity: O(1).
*/
static unsigned int degree_invariant(const Graph* const g,
const unsigned int v);
/** \internal
* Partition independent invariant.
* Returns 1 if there is an edge from the vertex to itself, 0 if not.
* Time complexity: O(k), where k is the number of edges leaving the vertex.
*/
static unsigned int selfloop_invariant(const Graph* const g,
const unsigned int v);
bool refine_according_to_invariant(unsigned int (*inv)(const Graph* const g,
const unsigned int v));
/*
* Routines needed when refining the partition p into equitable
*/
bool split_neighbourhood_of_unit_cell(Partition::Cell *);
bool split_neighbourhood_of_cell(Partition::Cell * const);
/** \internal
* \copydoc AbstractGraph::is_equitable() const
*/
bool is_equitable() const;
/* Splitting heuristics, documented in more detail in graph.cc */
SplittingHeuristic sh;
Partition::Cell* find_next_cell_to_be_splitted(Partition::Cell *cell);
Partition::Cell* sh_first();
Partition::Cell* sh_first_smallest();
Partition::Cell* sh_first_largest();
Partition::Cell* sh_first_max_neighbours();
Partition::Cell* sh_first_smallest_max_neighbours();
Partition::Cell* sh_first_largest_max_neighbours();
void make_initial_equitable_partition();
void initialize_certificate();
bool is_automorphism(unsigned int* const perm);
bool nucr_find_first_component(const unsigned int level);
bool nucr_find_first_component(const unsigned int level,
std::vector<unsigned int>& component,
unsigned int& component_elements,
Partition::Cell*& sh_return);
public:
/**
* Create a new graph with \a N vertices and no edges.
*/
Graph(const unsigned int N = 0);
/**
* Destroy the graph.
*/
~Graph();
/**
* Read the graph from the file \a fp in a variant of the DIMACS format.
* See the <A href="http://www.tcs.hut.fi/Software/bliss/">bliss website</A>
* for the definition of the file format.
* Note that in the DIMACS file the vertices are numbered from 1 to N while
* in this C++ API they are from 0 to N-1.
* Thus the vertex n in the file corresponds to the vertex n-1 in the API.
*
* \param fp the file stream for the graph file
* \param errstr if non-null, the possible error messages are printed
* in this file stream
* \return a new Graph object or 0 if reading failed for some
* reason
*/
static Graph* read_dimacs(FILE* const fp, FILE* const errstr = stderr);
/**
* Write the graph to a file in a variant of the DIMACS format.
* See the <A href="http://www.tcs.hut.fi/Software/bliss/">bliss website</A>
* for the definition of the file format.
*/
void write_dimacs(FILE* const fp);
/**
* \copydoc AbstractGraph::write_dot(FILE * const fp)
*/
void write_dot(FILE* const fp);
/**
* \copydoc AbstractGraph::write_dot(const char * const file_name)
*/
void write_dot(const char* const file_name);
/**
* \copydoc AbstractGraph::is_automorphism(const std::vector<unsigned int>& perm) const
*/
bool is_automorphism(const std::vector<unsigned int>& perm) const;
/**
* \copydoc AbstractGraph::get_hash()
*/
virtual unsigned int get_hash();
/**
* Return the number of vertices in the graph.
*/
unsigned int get_nof_vertices() const {return vertices.size(); }
/**
* \copydoc AbstractGraph::permute(const unsigned int* const perm) const
*/
Graph* permute(const unsigned int* const perm) const;
Graph* permute(const std::vector<unsigned int>& perm) const;
/**
* Add a new vertex with color \a color in the graph and return its index.
*/
unsigned int add_vertex(const unsigned int color = 0);
/**
* Add an edge between vertices \a v1 and \a v2.
* Duplicate edges between vertices are ignored but try to avoid introducing
* them in the first place as they are not ignored immediately but will
* consume memory and computation resources for a while.
*/
void add_edge(const unsigned int v1, const unsigned int v2);
/**
* Change the color of the vertex \a vertex to \a color.
*/
void change_color(const unsigned int vertex, const unsigned int color);
/**
* Compare this graph with the graph \a other.
* Returns 0 if the graphs are equal, and a negative (positive) integer
* if this graph is "smaller than" ("greater than", resp.) than \a other.
*/
int cmp(Graph& other);
/**
* Set the splitting heuristic used by the automorphism and canonical
* labeling algorithm.
* The selected splitting heuristics affects the computed canonical
* labelings; therefore, if you want to compare whether two graphs
* are isomorphic by computing and comparing (for equality) their
* canonical versions, be sure to use the same splitting heuristics
* for both graphs.
*/
void set_splitting_heuristic(const SplittingHeuristic shs) {sh = shs; }
};
/**
* \brief The class for directed, vertex colored graphs.
*
* Multiple edges between vertices are not allowed (i.e., are ignored).
*/
class Digraph : public AbstractGraph
{
public:
/**
* The possible splitting heuristics.
* The selected splitting heuristics affects the computed canonical
* labelings; therefore, if you want to compare whether two graphs
* are isomorphic by computing and comparing (for equality) their
* canonical versions, be sure to use the same splitting heuristics
* for both graphs.
*/
typedef enum {
/** First non-unit cell.
* Very fast but may result in large search spaces on difficult graphs.
* Use for large but easy graphs. */
shs_f = 0,
/** First smallest non-unit cell.
* Fast, should usually produce smaller search spaces than shs_f. */
shs_fs,
/** First largest non-unit cell.
* Fast, should usually produce smaller search spaces than shs_f. */
shs_fl,
/** First maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_fm,
/** First smallest maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_fsm,
/** First largest maximally non-trivially connected non-unit cell.
* Not so fast, should usually produce smaller search spaces than shs_f,
* shs_fs, and shs_fl. */
shs_flm
} SplittingHeuristic;
protected:
class Vertex {
public:
Vertex();
~Vertex();
void add_edge_to(const unsigned int dest_vertex);
void add_edge_from(const unsigned int source_vertex);
void remove_duplicate_edges(std::vector<bool>& tmp);
void sort_edges();
unsigned int color;
std::vector<unsigned int> edges_out;
std::vector<unsigned int> edges_in;
unsigned int nof_edges_in() const {return edges_in.size(); }
unsigned int nof_edges_out() const {return edges_out.size(); }
};
std::vector<Vertex> vertices;
void remove_duplicate_edges();
/** \internal
* Partition independent invariant.
* Returns the color of the vertex.
* Time complexity: O(1).
*/
static unsigned int vertex_color_invariant(const Digraph* const g,
const unsigned int v);
/** \internal
* Partition independent invariant.
* Returns the indegree of the vertex.
* DUPLICATE EDGES MUST HAVE BEEN REMOVED BEFORE.
* Time complexity: O(1).
*/
static unsigned int indegree_invariant(const Digraph* const g,
const unsigned int v);
/** \internal
* Partition independent invariant.
* Returns the outdegree of the vertex.
* DUPLICATE EDGES MUST HAVE BEEN REMOVED BEFORE.
* Time complexity: O(1).
*/
static unsigned int outdegree_invariant(const Digraph* const g,
const unsigned int v);
/** \internal
* Partition independent invariant.
* Returns 1 if there is an edge from the vertex to itself, 0 if not.
* Time complexity: O(k), where k is the number of edges leaving the vertex.
*/
static unsigned int selfloop_invariant(const Digraph* const g,
const unsigned int v);
/** \internal
* Refine the partition \a p according to
* the partition independent invariant \a inv.
*/
bool refine_according_to_invariant(unsigned int (*inv)(const Digraph* const g,
const unsigned int v));
/*
* Routines needed when refining the partition p into equitable
*/
bool split_neighbourhood_of_unit_cell(Partition::Cell* const);
bool split_neighbourhood_of_cell(Partition::Cell* const);
/** \internal
* \copydoc AbstractGraph::is_equitable() const
*/
bool is_equitable() const;
/* Splitting heuristics, documented in more detail in the cc-file. */
SplittingHeuristic sh;
Partition::Cell* find_next_cell_to_be_splitted(Partition::Cell *cell);
Partition::Cell* sh_first();
Partition::Cell* sh_first_smallest();
Partition::Cell* sh_first_largest();
Partition::Cell* sh_first_max_neighbours();
Partition::Cell* sh_first_smallest_max_neighbours();
Partition::Cell* sh_first_largest_max_neighbours();
void make_initial_equitable_partition();
void initialize_certificate();
bool is_automorphism(unsigned int* const perm);
void sort_edges();
bool nucr_find_first_component(const unsigned int level);
bool nucr_find_first_component(const unsigned int level,
std::vector<unsigned int>& component,
unsigned int& component_elements,
Partition::Cell*& sh_return);
public:
/**
* Create a new directed graph with \a N vertices and no edges.
*/
Digraph(const unsigned int N = 0);
/**
* Destroy the graph.
*/
~Digraph();
/**
* Read the graph from the file \a fp in a variant of the DIMACS format.
* See the <A href="http://www.tcs.hut.fi/Software/bliss/">bliss website</A>
* for the definition of the file format.
* Note that in the DIMACS file the vertices are numbered from 1 to N while
* in this C++ API they are from 0 to N-1.
* Thus the vertex n in the file corresponds to the vertex n-1 in the API.
* \param fp the file stream for the graph file
* \param errstr if non-null, the possible error messages are printed
* in this file stream
* \return a new Digraph object or 0 if reading failed for some
* reason
*/
static Digraph* read_dimacs(FILE* const fp, FILE* const errstr = stderr);
/**
* \copydoc AbstractGraph::write_dimacs(FILE * const fp)
*/
void write_dimacs(FILE* const fp);
/**
* \copydoc AbstractGraph::write_dot(FILE *fp)
*/
void write_dot(FILE * const fp);
/**
* \copydoc AbstractGraph::write_dot(const char * const file_name)
*/
void write_dot(const char * const file_name);
/**
* \copydoc AbstractGraph::is_automorphism(const std::vector<unsigned int>& perm) const
*/
bool is_automorphism(const std::vector<unsigned int>& perm) const;
/**
* \copydoc AbstractGraph::get_hash()
*/
virtual unsigned int get_hash();
/**
* Return the number of vertices in the graph.
*/
unsigned int get_nof_vertices() const {return vertices.size(); }
/**
* Add a new vertex with color 'color' in the graph and return its index.
*/
unsigned int add_vertex(const unsigned int color = 0);
/**
* Add an edge from the vertex \a source to the vertex \a target.
* Duplicate edges are ignored but try to avoid introducing
* them in the first place as they are not ignored immediately but will
* consume memory and computation resources for a while.
*/
void add_edge(const unsigned int source, const unsigned int target);
/**
* Change the color of the vertex 'vertex' to 'color'.
*/
void change_color(const unsigned int vertex, const unsigned int color);
/**
* Compare this graph with the graph \a other.
* Returns 0 if the graphs are equal, and a negative (positive) integer
* if this graph is "smaller than" ("greater than", resp.) than \a other.
*/
int cmp(Digraph& other);
/**
* Set the splitting heuristic used by the automorphism and canonical
* labeling algorithm.
* The selected splitting heuristics affects the computed canonical
* labelings; therefore, if you want to compare whether two graphs
* are isomorphic by computing and comparing (for equality) their
* canonical versions, be sure to use the same splitting heuristics
* for both graphs.
*/
void set_splitting_heuristic(SplittingHeuristic shs) {sh = shs; }
/**
* \copydoc AbstractGraph::permute(const unsigned int* const perm) const
*/
Digraph* permute(const unsigned int* const perm) const;
Digraph* permute(const std::vector<unsigned int>& perm) const;
};
}
#endif