2
我是比較新的編程C++。我正在實現一個像索引樹一樣的樹，使用unorderd_map來實現樹數據結構來存儲子節點。由於即時通訊工作與結構樹一樣，構造一個搜索方法是遞歸的，我也存儲節點的指針，所以我懷疑我可能有一種處理不好的內存問題。我遇到了分段錯誤。接下來是我的代碼和它的輸出。在遞歸函數中使用unordered_map unordered_set進行分段錯誤
#include <memory>
#include <sstream>
#include <unordered_map>
#include <iostream>
#include <string>
#include <sqlite3.h>
#include "aux_functions.cpp"
#include <math.h>
using namespace std;
class TreeLikeIndex
{
public:
TreeLikeIndex(string attribute, string indices, int indices_count, short int is_leaf, unordered_map<string, TreeLikeIndex*> children);
TreeLikeIndex(string indices, int indices_count);
TreeLikeIndex();
string search(unordered_map<string, string> *);
private:
string indices;
int indices_count;
short int is_leaf;
string attribute;
unordered_map<string, TreeLikeIndex*> children;
};
string TreeLikeIndex::search(unordered_map<string, string> * _tuple)
{
if((*_tuple).empty() || this->is_leaf) return this->indices;
string att_val = (*_tuple)[this->attribute];
(*_tuple).erase(this->attribute);
TreeLikeIndex * child_with_that_value = this->children[att_val];
return (*child_with_that_value).search(_tuple);
}
class DecisionTreeLikeIndexer
{
public:
DecisionTreeLikeIndexer(string, string, string);
int rebuild_index();
TreeLikeIndex * get_index();
private:
TreeLikeIndex * build_index(unordered_set<string> attributes_list, int depth, string comma_separated_ids, int ids_list_count);
TreeLikeIndex * index;
string source_db_address;
string dest_folder_address;
time_t time_of_last_build;
unordered_set<string> columns_names;
string source_table_name;
unordered_set<string> temp_tables_names;
string id_column_name;
sqlite3 * source_db_connection;
int table_count;
};
int DecisionTreeLikeIndexer::rebuild_index()
{
this->index = this->build_index(this->columns_names, 0, "", 0);
this->time_of_last_build = time(NULL);
return 0;
}
TreeLikeIndex * DecisionTreeLikeIndexer::get_index()
{
return this->index;
}
DecisionTreeLikeIndexer::DecisionTreeLikeIndexer(string source_db_address, string table_name, string dest_folder_address)
{
this->source_db_address = source_db_address;
this->dest_folder_address = dest_folder_address;
this->columns_names = Aux::get_column_names(source_db_address, table_name);
this->source_table_name = table_name;
this->id_column_name = "rowid";
this->source_db_connection = Aux::get_db_connection(this->source_db_address);
// Getting count of this table
sqlite3_stmt* statement;
string query = "SELECT count(*) FROM " + this->source_table_name + ";";
if(sqlite3_prepare(this->source_db_connection, query.c_str(), -1, &statement, 0) == SQLITE_OK)
{
int res = sqlite3_step(statement);
const unsigned char * count_char = sqlite3_column_text(statement,0);
if(res == SQLITE_ROW)
{
stringstream _temp;
_temp << count_char;
_temp >> this->table_count;
}
sqlite3_finalize(statement);
}
else
{
cout << "Error initializating Indexer (Getting initial table count): " << sqlite3_errmsg(this->source_db_connection) << endl;
}
}
TreeLikeIndex * DecisionTreeLikeIndexer::build_index(unordered_set<string> attributes_list, int depth, string comma_separated_ids, int ids_list_count)
{
if(attributes_list.size() <=1 || (depth > 0 && ids_list_count <= 1))
{
Aux::tabs(depth);
cout << "Leaf at depth: " << depth << " Ids are: " << comma_separated_ids << " Ids count: " << ids_list_count << endl;
static TreeLikeIndex * node = new TreeLikeIndex((string)comma_separated_ids, (int)ids_list_count);
return node;
}
string source_table = this->source_table_name;
int count = this->table_count;
if(depth > 0)
{
while(1)
{
source_table = *Aux::get_random_list_of_strings(1).begin();
if(this->temp_tables_names.insert(source_table).second) break;
}
const string create_temp_table_stmnt = "CREATE TEMP TABLE " + source_table + " AS SELECT * FROM " + this->source_table_name + " WHERE " + this->id_column_name + " IN(" + comma_separated_ids + ")";
sqlite3_exec(this->source_db_connection, create_temp_table_stmnt.c_str(),Aux::sqlt_callback,0,NULL);
count = ids_list_count;
Aux::tabs(depth);
cout << "Not root node" << endl;
}
Aux::tabs(depth);
cout << "Source table is: " << source_table << " Table count is: " << count << endl;
Aux::tabs(depth);
cout << "Attributes list is: "; for_each(attributes_list.begin(), attributes_list.end(),[](string v){cout << v << " ";});
cout << endl;
const double E = log2(count) ;
Aux::tabs(depth);
cout << "Entropy of node: " << E << endl;
string best_attribute;
double best_gain;
unordered_set<string> best_attribute_values;
for(string attr: attributes_list)
{
Aux::tabs(depth+1);
cout << "Analysing attribute: " << attr << endl;
const string get_at_count_values_query = "SELECT " + attr + ", count(" + attr + ") FROM " + source_table + " GROUP BY " + attr + ";";
sqlite3_stmt * stmnt;
double weighted_entropy = 0;
unordered_set<string> this_att_values;
if(sqlite3_prepare(this->source_db_connection, get_at_count_values_query.c_str(), -1, &stmnt, 0) == SQLITE_OK)
{
for(;;)
{
int res = sqlite3_step(stmnt);
if(res == SQLITE_DONE || res==SQLITE_ERROR)
{
double gAti = E - weighted_entropy;
Aux::tabs(depth+1);
cout << "Finish computing WE for att: " << attr << " Gain is: " << gAti << endl;
if(gAti > best_gain)
{
Aux::tabs(depth+1);
cout << "Found attribute with better gain." << endl;
best_gain = gAti;
best_attribute = attr;
best_attribute_values.clear();
Aux::tabs(depth+1);
for(string v:this_att_values)
{
best_attribute_values.insert(v);
}
cout << endl;
this_att_values.clear();
}
sqlite3_finalize(stmnt);
//delete &res;
break;
}
if(res == SQLITE_ROW)
{
string val = std::string(reinterpret_cast<const char*>(sqlite3_column_text(stmnt,0)));
int vSize = sqlite3_column_int(stmnt,1);
Aux::tabs(depth+2);
this_att_values.insert(val);
double ratio = double(vSize)/double(count);
weighted_entropy += double(ratio) * double(log2(vSize));
Aux::tabs(depth+2);
cout << "Processing value: " << val << " With vSize: " << vSize << " Current WE is: " << weighted_entropy << endl;
}
}
}
}
Aux::tabs(depth);
cout << "Finish processing attributes list. Best attribute is: " << best_attribute << " Best gain is: " << best_gain << endl;
Aux::tabs(depth);
cout << "Best attribute values are: "; for_each(best_attribute_values.begin(), best_attribute_values.end(), [](string v){cout << v << ",";}); cout << endl;
unordered_map<string, TreeLikeIndex *> children;
for(string val: best_attribute_values)
{
const string get_ids_of_bestatt_val = "SELECT rowid FROM " + source_table + " WHERE " + best_attribute + " = " + val + ";";
int ids_count = 0;
sqlite3_stmt * stmnt;
string ids = "";
bool first = 1;
int next_depth = depth + 1;
unordered_set<string> next_attributes_set;
for(string attr: attributes_list) if(attr != best_attribute) next_attributes_set.insert(attr);
if(sqlite3_prepare(this->source_db_connection, get_ids_of_bestatt_val.c_str(), -1, &stmnt,0) == SQLITE_OK)
{
for(;;)
{
int res = sqlite3_step(stmnt);
if(res == SQLITE_ROW)
{
string id = std::string(reinterpret_cast<const char*>(sqlite3_column_text(stmnt,0)));
if(!first) ids += "," + id;
else ids += id;
ids_count++;
}
if(res == SQLITE_DONE || res == SQLITE_ERROR)
{
Aux::tabs(depth+1);
cout << "Adding branch for val: " << val << endl;
Aux::tabs(depth+1);
cout << " Next attributes are: "; for_each(next_attributes_set.begin(), next_attributes_set.end(), [](string v){cout << v << ",";});
cout << " Depth is: " << next_depth << " Ids are: " << ids << " Ids count: " << ids_count << endl;
sqlite3_finalize(stmnt);
static TreeLikeIndex * temp_child = this->build_index(next_attributes_set, next_depth, ids, ids_count);
pair<string, TreeLikeIndex*> child (val, temp_child);
children.insert(child);
}
}
}
}
Aux::tabs(depth);
cout << "Finish processing node, will return." << endl;
static TreeLikeIndex * no_leaf_node = new TreeLikeIndex(best_attribute, "all", count, 0, children);
return no_leaf_node;
}
}
TreeLikeIndex::TreeLikeIndex(std::string attribute, std::string indices, int indices_count, short int is_leaf, unordered_map<std::string, TreeLikeIndex*> children)
{
this->attribute = attribute;
this->indices = indices;
this->is_leaf = is_leaf;
this->children = children;
this->children.clear();
for(pair<string, TreeLikeIndex*> p: children) this->children.insert(p);
this->indices_count = indices_count;
}
TreeLikeIndex::TreeLikeIndex(string indices, int indices_count)
{
this->indices = indices;
this->indices_count = indices_count;
this->is_leaf = 1;
}
TreeLikeIndex::TreeLikeIndex()
{
this->indices = "";
this->indices_count = 0;
this->is_leaf = 1;
}
int main()
{
string source_db_address = "my_table";
string table_name = "b";
string dest_folder_address = ".";
DecisionTreeLikeIndexer indexer(source_db_address, table_name, dest_folder_address);
indexer.rebuild_index();
}
，輸出是：
Source table is: b Table count is: 9
Attributes list is: cant_n_dec cant_n_des cant_n_control
Entropy of node: 3.16993
Analysing attribute: cant_n_dec
Processing value: 1 With vSize: 1 Current WE is: 0
Processing value: 2 With vSize: 4 Current WE is: 0.888889
Processing value: 3 With vSize: 2 Current WE is: 1.11111
Processing value: 4 With vSize: 1 Current WE is: 1.11111
Processing value: 5 With vSize: 1 Current WE is: 1.11111
Finish computing WE for att: cant_n_dec Gain is: 2.05881
Found attribute with better gain.
Analysing attribute: cant_n_des
Processing value: 1 With vSize: 2 Current WE is: 0.222222
Processing value: 2 With vSize: 4 Current WE is: 1.11111
Processing value: 3 With vSize: 2 Current WE is: 1.33333
Processing value: 5 With vSize: 1 Current WE is: 1.33333
Finish computing WE for att: cant_n_des Gain is: 1.83659
Analysing attribute: cant_n_control
Processing value: 1 With vSize: 2 Current WE is: 0.222222
Processing value: 2 With vSize: 3 Current WE is: 0.750543
Processing value: 3 With vSize: 3 Current WE is: 1.27886
Processing value: 5 With vSize: 1 Current WE is: 1.27886
Finish computing WE for att: cant_n_control Gain is: 1.89106
Finish processing attributes list. Best attribute is: cant_n_dec Best gain is: 2.05881
Best attribute values are: 1,2,3,4,5,
Adding branch for val: 1
Next attributes are: cant_n_control,cant_n_des, Depth is: 1 Ids are: 3 Ids count: 1
Leaf at depth: 1 Ids are: 3 Ids count: 1
Segmentation fault
而且當您逐行調試您的代碼時，您的調試器會根據調試程序導致分段故障，以及堆棧跟蹤是什麼樣的？ –
我使用打印（cout <<）作爲調試技術。程序執行從第一個（初始）一步進入遞歸上下文，並返回一個指針（指向葉節點）。這個指針要在使用insert的第一個（初始）上下文中存儲到unordered_map中。就像這樣：T * obj = recursive_call（）; some_pair（「key」，obj）; uMap.insert（some_pair）;有崩潰的地方。 –
把額外的打印語句完全不適用於所有，但最簡單的錯誤。學習和了解如何使用調試器是每個C++開發人員必備的技能。調試器提供的信息比基本的打印語句有用得多。放下這個程序，花一些時間學習使用調試器。 –