memgraph/poc/astar.cpp
2016-08-29 01:01:42 +01:00

301 lines
8.2 KiB
C++

#include <chrono>
#include <cstring>
#include <ctime>
#include <fstream>
#include <iostream>
#include <queue>
#include <regex>
#include <sstream>
#include <string>
#include <vector>
#include "data_structures/map/rh_hashmap.hpp"
#include "database/db.hpp"
#include "database/db_accessor.cpp"
#include "database/db_accessor.hpp"
#include "import/csv_import.hpp"
#include "storage/edge_x_vertex.hpp"
#include "storage/edges.cpp"
#include "storage/edges.hpp"
#include "storage/indexes/impl/nonunique_unordered_index.cpp"
#include "storage/model/properties/properties.cpp"
#include "storage/record_accessor.cpp"
// #include "storage/vertex_accessor.cpp"
#include "storage/vertex_accessor.hpp"
#include "storage/vertices.cpp"
#include "storage/vertices.hpp"
#include "utils/command_line/arguments.hpp"
#include "communication/bolt/v1/serialization/bolt_serializer.hpp"
const int max_score = 1000000;
using namespace std;
typedef VertexAccessor VertexAccessor;
void add_scores(Db &db);
class Node
{
public:
Node *parent = {nullptr};
type_key_t<TypeGroupVertex, Double> tkey;
double cost;
int depth = {0};
VertexAccessor vacc;
Node(VertexAccessor vacc, double cost,
type_key_t<TypeGroupVertex, Double> tkey)
: cost(cost), vacc(vacc), tkey(tkey)
{
}
Node(VertexAccessor vacc, double cost, Node *parent,
type_key_t<TypeGroupVertex, Double> tkey)
: cost(cost), vacc(vacc), parent(parent), depth(parent->depth + 1),
tkey(tkey)
{
}
double sum_vertex_score()
{
auto now = this;
double sum = 0;
do {
sum += (now->vacc.at(tkey).get())->value;
now = now->parent;
} while (now != nullptr);
return sum;
}
};
class Score
{
public:
Score() : value(std::numeric_limits<double>::max()) {}
Score(double v) : value(v) {}
double value;
};
void found_result(Node *res)
{
double sum = res->sum_vertex_score();
std::cout << "{score: " << sum << endl;
auto bef = res;
while (bef != nullptr) {
std::cout << " " << *(bef->vacc.operator->()) << endl;
bef = bef->parent;
}
}
double calc_heuristic_cost_dummy(type_key_t<TypeGroupVertex, Double> tkey,
EdgeAccessor &edge, VertexAccessor &vertex)
{
assert(!vertex.empty());
return 1 - vertex.at(tkey).get()->value;
}
typedef bool (*EdgeFilter)(DbAccessor &t, EdgeAccessor &, Node *before);
typedef bool (*VertexFilter)(DbAccessor &t, VertexAccessor &, Node *before);
bool edge_filter_dummy(DbAccessor &t, EdgeAccessor &e, Node *before)
{
return true;
}
bool vertex_filter_dummy(DbAccessor &t, VertexAccessor &va, Node *before)
{
return va.fill();
}
bool vertex_filter_contained_dummy(DbAccessor &t, VertexAccessor &v,
Node *before)
{
if (v.fill()) {
bool found;
do {
found = false;
before = before->parent;
if (before == nullptr) {
return true;
}
auto it = before->vacc.out();
for (auto e = it.next(); e.is_present(); e = it.next()) {
VertexAccessor va = e.get().to();
if (va == v) {
found = true;
break;
}
}
} while (found);
}
return false;
}
bool vertex_filter_contained(DbAccessor &t, VertexAccessor &v, Node *before)
{
if (v.fill()) {
bool found;
do {
found = false;
before = before->parent;
if (before == nullptr) {
return true;
}
} while (v.in_contains(before->vacc));
}
return false;
}
// Vertex filter ima max_depth funkcija te edge filter ima max_depth funkcija.
// Jedan za svaku dubinu.
// Filtri vracaju true ako element zadovoljava uvjete.
auto a_star(
Db &db, int64_t sys_id_start, uint max_depth, EdgeFilter e_filter[],
VertexFilter v_filter[],
double (*calc_heuristic_cost)(type_key_t<TypeGroupVertex, Double> tkey,
EdgeAccessor &edge, VertexAccessor &vertex),
int limit)
{
DbAccessor t(db);
type_key_t<TypeGroupVertex, Double> tkey =
t.vertex_property_family_get("score")
.get(Flags::Double)
.type_key<Double>();
auto best_found = new std::map<Id, Score>[max_depth];
std::vector<Node *> best;
auto cmp = [](Node *left, Node *right) { return left->cost > right->cost; };
std::priority_queue<Node *, std::vector<Node *>, decltype(cmp)> queue(cmp);
auto start_vr = t.vertex_find(sys_id_start);
assert(start_vr);
start_vr.get().fill();
Node *start = new Node(start_vr.take(), 0, tkey);
queue.push(start);
int count = 0;
do {
auto now = queue.top();
queue.pop();
// if(!visited.insert(now)){
// continue;
// }
if (max_depth <= now->depth) {
best.push_back(now);
count++;
if (count >= limit) {
return best;
}
continue;
}
// { // FOUND FILTER
// Score &bef = best_found[now->depth][now->vacc.id()];
// if (bef.value <= now->cost) {
// continue;
// }
// bef.value = now->cost;
// }
iter::for_all(now->vacc.out(), [&](auto edge) {
if (e_filter[now->depth](t, edge, now)) {
VertexAccessor va = edge.to();
if (v_filter[now->depth](t, va, now)) {
auto cost = calc_heuristic_cost(tkey, edge, va);
Node *n = new Node(va, now->cost + cost, now, tkey);
queue.push(n);
}
}
});
} while (!queue.empty());
// TODO: GUBI SE MEMORIJA JER SE NODOVI NEBRISU
t.commit();
return best;
}
int main(int argc, char **argv)
{
auto para = all_arguments(argc, argv);
Db db;
auto loaded = import_csv_from_arguments(db, para);
add_scores(db);
EdgeFilter e_filters[] = {&edge_filter_dummy, &edge_filter_dummy,
&edge_filter_dummy, &edge_filter_dummy};
VertexFilter f_filters[] = {
&vertex_filter_contained, &vertex_filter_contained,
&vertex_filter_contained, &vertex_filter_contained};
// CONF
std::srand(time(0));
auto best_n = 10;
auto bench_n = 1000;
auto best_print_n = 10;
bool pick_best_found =
strcmp(get_argument(para, "-p", "true").c_str(), "true") == 0;
double sum = 0;
std::vector<Node *> best;
for (int i = 0; i < bench_n; i++) {
auto start_vertex_index = std::rand() % loaded.first;
auto begin = clock();
auto found = a_star(db, start_vertex_index, 3, e_filters, f_filters,
&calc_heuristic_cost_dummy, best_n);
clock_t end = clock();
double elapsed_ms = (double(end - begin) / CLOCKS_PER_SEC) * 1000;
sum += elapsed_ms;
if ((best.size() < best_print_n && found.size() > best.size()) ||
(pick_best_found && found.size() > 0 &&
found.front()->sum_vertex_score() >
best.front()->sum_vertex_score())) {
best = found;
}
// Just to be safe
if (i + 1 == bench_n && best.size() == 0) {
bench_n++;
}
}
std::cout << "\nSearch for best " << best_n
<< " results has runing time of:\n avg: " << sum / bench_n
<< " [ms]\n";
std::cout << "\nExample of best result:\n";
for (int i = 0; i < best_print_n && best.size() > 0; i++) {
found_result(best.front());
best.erase(best.begin());
}
return 0;
}
// Adds property score to all vertices.
void add_scores(Db &db)
{
DbAccessor t(db);
auto key_score =
t.vertex_property_family_get("score").get(Flags::Double).family_key();
int i = 1;
iter::for_all(t.vertex_access(), [&](auto v) {
if (v.fill()) {
// from Kruno's head :) (could be ALMOST anything else)
std::srand(i ^ 0x7482616);
v.set(key_score,
std::make_shared<Double>((std::rand() % max_score) /
(max_score + 0.0)));
i++;
}
});
t.commit();
}