fix: bug in KVstore and Join;

mess code in KVstore due to read after request() release the pointer;
lack of empty judgement in Join::pre_handler;

by zengli
This commit is contained in:
bookug 2017-07-12 20:33:44 +08:00
parent 09a30931d4
commit ae5e610acc
13 changed files with 159 additions and 66 deletions

View File

@ -732,18 +732,19 @@ Database::load_vstree(unsigned _vstree_size)
void
Database::check()
{
//string tstr;
string tstr;
//unsigned pid = this->kvstore->getIDByPredicate("<http://www.w3.org/1999/02/22-rdf-syntax-ns#type>");
//cout<<"check: pre "<<pid<<endl;
//this->stringindex->randomAccess(pid, &tstr, false);
//cout<<"string index: "<<tstr<<endl;
////cout<<"kvstore: "<<this->kvstore->getPredicateByID(pid)<<endl;
//unsigned sid = this->kvstore->getIDByEntity("<http://www.Department4.University1821.edu/Course50>");
//cout<<"check: sub "<<sid<<endl;
//this->stringindex->randomAccess(sid, &tstr, true);
//cout<<"string index: "<<tstr<<endl;
//cout<<"kvstore: "<<this->kvstore->getEntityByID(sid)<<endl;
cout<<"right pair: "<<62<<" "<<"<http://www.Department0.University0.edu/GraduateCourse11>"<<endl;
unsigned sid = this->kvstore->getIDByEntity("<http://www.Department0.University0.edu/GraduateCourse11>");
cout<<"check: sub "<<sid<<endl;
this->stringindex->randomAccess(sid, &tstr, true);
cout<<"string index: "<<tstr<<endl;
cout<<"kvstore: "<<this->kvstore->getEntityByID(sid)<<endl;
//unsigned oid = this->kvstore->getIDByString("<http://www.lehigh.edu/~zhp2/2004/0401/univ-bench.owl#Course>");
//cout<<"check: obj "<<oid<<endl;

View File

@ -196,8 +196,9 @@ Join::join_sparql(SPARQLquery& _sparql_query)
return true;
}
//TODO: consider a node with multiple same predicates(not pre var), use getXXX(...false) to do this
//TODO: consider a node with multiple same predicates(not pre var), use p2s(...false) to do this
//BETTER?: ?s-p-?o, use p2so instead of p2s and p2o to get candidates for ?s and ?o will be better??
//TODO: deal with predicate variables, maybe not ready like literals
bool
Join::pre_handler()
{
@ -214,12 +215,12 @@ Join::pre_handler()
for (int j = 0; j < var_degree; j++)
{
int neighbor_id = this->basic_query->getEdgeNeighborID(_var_i, j);
if (neighbor_id != -1) //TODO:what does it mean???variables in join not considered here
//-1: constant or variable not in join; otherwise, variable in join
if (neighbor_id != -1)
{
continue;
}
//TODO set ready here?
this->basic_query->setReady(_var_i);
char edge_type = this->basic_query->getEdgeType(_var_i, j);
int triple_id = this->basic_query->getEdgeID(_var_i, j);
Triple triple = this->basic_query->getTriple(triple_id);
@ -234,8 +235,8 @@ Join::pre_handler()
neighbor_name = triple.subject;
}
//TODO : consider or not??
bool only_preid_filter = (this->basic_query->isOneDegreeNotJoinVar(neighbor_name));
//NOTICE: we only need to consider constants here
if(only_preid_filter)
{
continue;
@ -243,11 +244,11 @@ Join::pre_handler()
else
{
this->dealed_triple[triple_id] = true;
this->basic_query->setReady(_var_i);
}
TYPE_PREDICATE_ID pre_id = this->basic_query->getEdgePreID(_var_i, j);
//TODO : cancle the literal
TYPE_ENTITY_LITERAL_ID lit_id = (this->kvstore)->getIDByEntity(neighbor_name);
if (lit_id == INVALID_ENTITY_LITERAL_ID)
{
@ -284,19 +285,20 @@ Join::pre_handler()
id_list_len = 0;
}
//WARN: this may need to check, end directly
if (id_list_len == 0)
{
_list.clear();
delete[] id_list;
return false;
}
//TODO : correct ways to intersect
//updateList(_list, id_list, id_list_len);
if (_list.size() == 0)
_list.unionList(id_list,id_list_len);
else
_list.intersectList(id_list, id_list_len);
delete[] id_list;
if (_list.size() == 0)
{
return false;
@ -305,7 +307,9 @@ Join::pre_handler()
cout << "\t\t[" << _var_i << "] after constant filter, candidate size = " << _list.size() << endl << endl << endl;
}
cout << "pre var filter start here" << endl;
cout << "pre filter start here" << endl;
//TODO:use vector instead of set
for(int _var = 0; _var < this->var_num; _var++)
{
if(this->basic_query->isSatelliteInJoin(_var))
@ -316,7 +320,6 @@ Join::pre_handler()
unsigned size = this->basic_query->getCandidateSize(_var);
//result if already empty for non-literal variable
//TODO: whether use this
/*
if (size == 0)
{
@ -362,8 +365,6 @@ Join::pre_handler()
continue;
}
//TODO+BETTER: is any pre really used? do we need to losen the restrictions?
//size:m<n; time:mlgn < n-m
//The former time is computed because the m should be small if we select this p, tending to use binary-search
//when doing intersectList operation(mlgn < m+n).
@ -393,7 +394,6 @@ Join::pre_handler()
{
continue;
}
//TODO : set ready here ?
this->basic_query->setReady(_var);
//NOTICE:use p2s here, use s2p in only_pre_filter_after_join because pres there are not efficient
set<TYPE_PREDICATE_ID>::iterator it;
@ -407,6 +407,10 @@ Join::pre_handler()
else
cans.intersectList(list, len);
delete[] list;
if(cans.size() == 0)
{
return false;
}
}
if(in_edge_pre_id.size() != 0 && cans.size() == 0)
@ -423,6 +427,10 @@ Join::pre_handler()
else
cans.intersectList(list, len);
delete[] list;
if(cans.size() == 0)
{
return false;
}
}
@ -432,8 +440,8 @@ Join::pre_handler()
return false;
}
cout << "\t\t[" << _var << "] after pre var filter, candidate size = " << cans.size() << endl << endl << endl;
}
return true;
}

View File

@ -129,7 +129,20 @@ ISTree::search(unsigned _key, char*& _str, unsigned& _len)
_str = val->getStr();
_len = val->getLen();
char* debug = new char[_len];
memcpy(debug, _str, _len);
// debug[_len] = '\0';
_str = debug;
//if(_key==62)
//{
//cout<<"check in search: "<<string(_str, _len)<<endl;
//}
this->TSM->request(request);
//if(_key==62)
//{
//cout<<"check in search: "<<string(_str, _len)<<endl;
//}
return true;
}

View File

@ -20,8 +20,8 @@ ISHeap::ISHeap(unsigned _size)
{
this->length = 0;
this->size = _size;
//this->heap = (Node**)malloc(this->size * sizeof(Node*)); //not use 4 or 8
this->heap = new ISNode*[this->size];
this->heap = (ISNode**)malloc(this->size * sizeof(ISNode*)); //not use 4 or 8
//this->heap = new ISNode*[this->size];
if (this->heap == NULL)
{
this->print("error in ISHeap: Allocation fail!");
@ -69,6 +69,10 @@ ISHeap::insert(ISNode* _np)
{
if (this->length == this->size) //when full, reallocate
{
cout<<"check: double the heap"<<endl;
//WARN: realloc should be combined with malloc instead of new
//http://bbs.csdn.net/topics/320148799
//For new, use placement new to enlarge space: http://blog.csdn.net/vangoals/article/details/4252833
this->heap = (ISNode**)realloc(this->heap, 2 * this->size * sizeof(ISNode*));
if (this->heap == NULL)
{
@ -173,7 +177,8 @@ ISHeap::modify(ISNode* _np, bool _flag) //control direction
ISHeap::~ISHeap()
{
delete[] this->heap;
//delete[] this->heap;
free(this->heap);
this->heap = NULL;
this->length = this->size = 0;
}
@ -183,4 +188,4 @@ ISHeap::print(string s)
{
#ifdef DEBUG_KVSTORE
#endif
}
}

View File

@ -223,6 +223,7 @@ ISLeafNode::split(ISNode* _father, int _index)
p->addNum();
}
unsigned tp = this->keys[MIN_KEY_NUM];
//cout<<"split: "<<tp<<" "<<this->values[MIN_KEY_NUM].getStr()<<endl;
this->setNum(MIN_KEY_NUM);
_father->addKey(tp, _index);
_father->addChild(p, _index + 1); //DEBUG(check the index)

View File

@ -137,6 +137,13 @@ IVTree::search(unsigned _key, char*& _str, unsigned& _len)
//_str = this->transfer[0].getStr();
//_len = this->transfer[0].getLen();
if (!VList::isLongList(_len))
{
char* debug = new char [_len];
memcpy(debug, _str, _len);
_str = debug;
}
this->TSM->request(request);
return true;
}

View File

@ -20,8 +20,8 @@ IVHeap::IVHeap(unsigned _size)
{
this->length = 0;
this->size = _size;
//this->heap = (Node**)malloc(this->size * sizeof(Node*)); //not use 4 or 8
this->heap = new IVNode*[this->size];
this->heap = (IVNode**)malloc(this->size * sizeof(IVNode*)); //not use 4 or 8
//this->heap = new IVNode*[this->size];
if (this->heap == NULL)
{
this->print("error in IVHeap: Allocation fail!");
@ -69,6 +69,7 @@ IVHeap::insert(IVNode* _np)
{
if (this->length == this->size) //when full, reallocate
{
cout<<"check: double the heap"<<endl;
this->heap = (IVNode**)realloc(this->heap, 2 * this->size * sizeof(IVNode*));
if (this->heap == NULL)
{
@ -173,7 +174,8 @@ IVHeap::modify(IVNode* _np, bool _flag) //control direction
IVHeap::~IVHeap()
{
delete[] this->heap;
//delete[] this->heap;
free(this->heap);
this->heap = NULL;
this->length = this->size = 0;
}

View File

@ -342,7 +342,7 @@ IVLeafNode::split(IVNode* _father, int _index)
p->addValue(this->values + i, k);
p->addNum();
}
int tp = this->keys[MIN_KEY_NUM];
unsigned tp = this->keys[MIN_KEY_NUM];
this->setNum(MIN_KEY_NUM);
_father->addKey(tp, _index);
_father->addChild(p, _index + 1); //DEBUG(check the index)

View File

@ -1154,6 +1154,38 @@ KVstore::getEntityByID(TYPE_ENTITY_LITERAL_ID _id) const
//NOTICE: no need to add \0 at last if we indicate the length
string _ret = string(_tmp, _len);
delete[] _tmp;
//CONSIDER: represent "abc" by 4 chars with length=4, and construct with string(str) directly
//if(_id == 62)
//{
//cout<<"check 62: "<<_ret<<endl;
//char* nstr = new char[_len+1];
//memcpy(nstr, _tmp, _len);
//nstr[_len] = '\0';
//cout<<"check 62: "<<nstr<<endl;
//for(int k = 0; k < _len; ++k)
//{
//cout<<_tmp[k]<<" ";
//}cout<<endl;
//delete[] nstr;
//}
//NOTICE: if there is messy code, then int(char) will < 0, otherwise not
//bool invalid = false;
//for(unsigned i = 0; i < _len; ++i)
//{
//if(int(_ret[i])<0)
//{
//invalid = true;
//break;
//}
//}
//if(invalid)
//{
//cout<<"error in kvstore: "<<_id<<" "<<_ret<<endl;
//cout<<"length: "<<_len<<" "<<_tmp<<" "<<_tmp[0]<<endl;
//}
return _ret;
}
@ -1167,6 +1199,8 @@ KVstore::setEntityByID(TYPE_ENTITY_LITERAL_ID _id, string _entity)
char* str = new char[len];
memcpy(str, _entity.c_str(), len);
//cout<<"check 62: "<<this->getEntityByID(62)<<endl;
return this->addValueByKey(this->id2entity, _id, str, len);
}
@ -1287,6 +1321,7 @@ KVstore::getPredicateByID(TYPE_PREDICATE_ID _id) const
return "";
}
string _ret = string(_tmp, _len);
delete[] _tmp;
return _ret;
}
@ -1422,6 +1457,21 @@ KVstore::getLiteralByID(TYPE_ENTITY_LITERAL_ID _id) const
return "";
}
string _ret = string(_tmp, _len);
delete[] _tmp;
//bool invalid = false;
//for(unsigned i = 0; i < _len; ++i)
//{
//if(int(_ret[i])<0)
//{
//invalid = true;
//break;
//}
//}
//if(invalid)
//{
//cout<<"error in kvstore: "<<_id<<" "<<_ret<<endl;
//}
return _ret;
}
@ -1596,11 +1646,11 @@ KVstore::getpreIDlistBysubID(TYPE_ENTITY_LITERAL_ID _subid, unsigned*& _preidlis
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1634,11 +1684,11 @@ KVstore::getobjIDlistBysubID(TYPE_ENTITY_LITERAL_ID _subid, unsigned*& _objidlis
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1684,11 +1734,11 @@ KVstore::getobjIDlistBysubIDpreID(TYPE_ENTITY_LITERAL_ID _subid, TYPE_PREDICATE_
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1732,11 +1782,11 @@ KVstore::getpreIDobjIDlistBysubID(TYPE_ENTITY_LITERAL_ID _subid, unsigned*& _pre
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1872,11 +1922,11 @@ KVstore::getpreIDlistByobjID(TYPE_ENTITY_LITERAL_ID _objid, unsigned*& _preidlis
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1904,11 +1954,11 @@ KVstore::getsubIDlistByobjID(TYPE_ENTITY_LITERAL_ID _objid, unsigned*& _subidlis
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1948,11 +1998,11 @@ KVstore::getsubIDlistByobjIDpreID(TYPE_ENTITY_LITERAL_ID _objid, TYPE_PREDICATE_
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -1989,11 +2039,11 @@ KVstore::getpreIDsubIDlistByobjID(TYPE_ENTITY_LITERAL_ID _objid, unsigned*& _pre
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -2109,11 +2159,11 @@ KVstore::getsubIDlistBypreID(TYPE_PREDICATE_ID _preid, unsigned*& _subidlist, un
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -2140,11 +2190,11 @@ KVstore::getobjIDlistBypreID(TYPE_PREDICATE_ID _preid, unsigned*& _objidlist, un
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -2180,14 +2230,14 @@ KVstore::getsubIDobjIDlistBypreID(TYPE_PREDICATE_ID _preid, unsigned*& _subid_ob
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
#ifdef DEBUG_KVSTORE
cout<<"this is a vlist"<<endl;
#endif
delete[] _tmp;
//_tmp = NULL;
}
//}
return true;
}
@ -2254,11 +2304,11 @@ KVstore::getpreIDlistBysubIDobjID(TYPE_ENTITY_LITERAL_ID _subid, TYPE_ENTITY_LIT
//if this is a long list, then we should remove itself after copying
//otherwise, we should not free the list memory
if(VList::listNeedDelete(_len))
{
//if(VList::listNeedDelete(_len))
//{
delete[] _tmp;
//_tmp = NULL;
}
//}
if (_list_len == 0) {
_preidlist = NULL;

View File

@ -20,8 +20,8 @@ SIHeap::SIHeap(unsigned _size)
{
this->length = 0;
this->size = _size;
//this->heap = (Node**)malloc(this->size * sizeof(Node*)); //not use 4 or 8
this->heap = new SINode*[this->size];
this->heap = (SINode**)malloc(this->size * sizeof(SINode*)); //not use 4 or 8
//this->heap = new SINode*[this->size];
if (this->heap == NULL)
{
this->print("error in SIHeap: Allocation fail!");
@ -69,6 +69,7 @@ SIHeap::insert(SINode* _np)
{
if (this->length == this->size) //when full, reallocate
{
cout<<"check: double the heap"<<endl;
this->heap = (SINode**)realloc(this->heap, 2 * this->size * sizeof(SINode*));
if (this->heap == NULL)
{
@ -173,7 +174,8 @@ SIHeap::modify(SINode* _np, bool _flag) //control direction
SIHeap::~SIHeap()
{
delete[] this->heap;
//delete[] this->heap;
free(this->heap);
this->heap = NULL;
this->length = this->size = 0;
}

View File

@ -249,6 +249,7 @@ int initialize() {
string format = "json";
cout<<"HTTP: this is query"<<endl;
string db_query=request->path_match[1];
cout<<"check: "<<db_query<<endl;
db_query = UrlDecode(db_query);
cout<<"check: "<<db_query<<endl;
string str = db_query;

View File

@ -63,6 +63,7 @@ in the sparql query can point to the same node in data graph)
#include <algorithm>
#include <functional>
#include <utility>
#include <new>
//NOTICE:below are libraries need to link
#include <thread> //only for c++11 or greater versions

View File

@ -132,6 +132,8 @@ public class GstoreConnector {
return false;
}
//TODO: also use encode to support spaces?
//Consider change format into ?name=DBname
String cmd = "build/" + _db_name + "/" + _rdf_file_path;
String msg = this.sendGet(cmd);