I have written code that allows one to traverse mapped data in the order it was entered.
The solution I coded a couple of times was:
Given a keytype, K, and and data type, D,
std::map
std::vector
When one wanted to randomly find a data entry, use map.find(K). When one wanted to traverse the map in entry order, use std::vector::iterator (begin(), end()].
This was fine, but as an exercise, I wanted to write this 'OrderedMap' as an STL compliant container. I also have (stripped down to this discussion):
template <typename K, typename D>
class OrderedMapValue
{
private:
K first_ref;
std::map<K,size_t>& m;
std::vector<D>& v;
public:
const K& first
D& second
assignment operator=(const D& data)
{
std::map<K,size_t>::const_iterator iter = m.find(first_ref);
v[iter.second] = data; // error checking of iter stripped
}
};
Further assuming
template <typename K, typename D>
class OrderedMap
{
public:
typename OrderedMapValue<K,D>& OrderedMap<K,D>::operator[](const K&);
// snip...
};
class MyClass
{
public:
MyClass(std::string s) : _my_data(s) {}
private:
std::string _my_data;
};
The following code works:
OrderedMap<std::string,MyClass*> omap;
omap["MyKey"] = new MyClass("dummy");
However, this code does not:
OrderedMap::iterator iter = omap.find("MyKey");
MyClass * obj = iter->second;
delete obj;
iter->second = new MyClass("dummy");
Assuming I have done something
a) Structurally silly or
b) Unnecessarily complex, how should this be done?
I realize that I'm likely reinventing the wheel here, but again, this effort is mainly to increase my knowledge of STL containers, their design patterns and proper use.
Thanks in advance for any insights,
I don't have a compiler right now to test this, so there could be errors, but I think you want it more like:
template <typename K, typename D>
class OrderedMap
{
private:
std::map<K,size_t> &m;
std::vector<D> &v;
public:
typename pair<K,D> TYPE;
TYPE& operator[](const K &k)
{
return v[ m[ k ]];
}
TYPE& operator[](size_t idx)
{
return v[ idx ];
}
pair<iterator,bool> insert( const TYPE& pair )
{
map<K, size_t>::const_iterator iter;
iter = m.find( pair.first );
if( iter != m.end() )
return make_pair( v[ iter.second], false );
m.insert( make_pair( pair->first, v.size() ));
v.push_back( pair->second );
return make_pair( v.last() , inserted );
}
iterator &begin()
{
return v.begin();
}
// etc
};
In OrderedMapValue::operator=, you have:
std::map<K,size_t>::const_iterator iter = m.find(first_ref);
What is first_ref? The code doesn't reference it (no pun intended) elsewhere. It looks to me like it might be a vestige from an older implementation, replaced elsewhere by the public member
const K& first.
Could this be the problem?
EDIT from the comments: The code doesn't show that first_ref is initialized anywhere; so for all I can tell, the call to m.find(first_ref) is searching for an empty string, rather than the key for the OrderedMapValue.
Related
use of deleted function in class operator=
old version worked with old compiler but not with new versions
I need this "operator=" overloading for container operation.
#include <boost/interprocess/ipc/message_queue.hpp>
#include <iostream>
using namespace boost::interprocess;
class X {
public:
size_t m_len;
shared_memory_object m_shm;
const char* m_ptr;
X():
m_len(0),
m_shm(shared_memory_object()),
m_ptr(nullptr){}
X(size_t t, const char* n):
m_len(t),
m_shm(shared_memory_object()),
m_ptr(nullptr){
shared_memory_object::remove(n);
m_shm = shared_memory_object(open_or_create,n, read_write);
m_shm.truncate (m_len);
mapped_region region(m_shm, read_write);
m_ptr = static_cast<char*>(region.get_address());
}
X(const X&& x){
m_len = x.m_len;
m_shm = x.m_shm; //error use deleted function
m_ptr = x.m_ptr;
}
virtual ~X(){}
X& operator = (const X&& a) {
if(&a == this) return *this;
m_len = a.m_len;
m_ptr = a.m_ptr;
m_shm = a.m_shm; //error use deleted function
return (*this);
}
const char* get_name(){
return m_shm.get_name();
}
};
int main ()
{
X a = X(22, "test");
X b = a; //Error
return 0;
};
The above class will be used in std::vector and operator= is needed.
boost shared_memory_object has member:
shared_memory_object(shared_memory_object &&);
shared_memory_object& operator=(shared_memory_object &&);
Move operations can only work on non-const objects. Why? Because it steals resources from the source instance, then it must be non-const to be modifiable. So your move operations should take non-const objects:
now should be
----------------------------------------------------------
X(const X&& x) ==> X(X&& x)
X& operator = (const X&& a) { ==> X& operator = (X&& a) {
Named variable is treated as L-value, it implies copy operations are called. You need to use std::move to cast source to R-value reference, then move operations can be called:
m_shm = std::move(x.m_shm); // in move ctor
m_shm = std::move(a.m_shm); // in move assignment operator
and finally to call move ctor:
X b = std::move(a);
I would like to pass to a function a reference to a class method.
For example:
#include <functional>
struct Foo
{
int f(const int& val) const
{
return val+2;
}
};
int caller(const std::function<int(const int&)>& f)
{
return f(1);
}
int main()
{
caller([](const int& val){return val+2;}); // OK
Foo foo;
caller(foo.f); // WRONG
return 0;
}
How can I fix the second call of caller() (NB: Foo:f() is not static)?
In your case, function f doesn't use any member of Foo so it can be declared static
static int f(const int& val)
and passed as:
caller(&Foo::f);
But suppose that f cannot be declared static and you want to pass "reference" to member function of particular object.
You can use lambda in that case:
Foo foo;
caller(
[&foo](const int& val){
return foo.f(val);
}
);
foo object is captured in square brackets (in this case by reference) so that you can call f member function on that particular object.
Although it is not part of your question, I should add that it is not really useful to pass int by const reference, as you will not gain any performance improvement that way. Actually, your code will run slower than if you pass int by value.
/* There is something wrong with the function delete_back(); I think something wrong with the remove function 3 parts.
Also remove_ele() I do not how to do it, thanks.
why I use the same method to delete element does not work
*/
#include <iostream>
using namespace std;
template<class T>
class doulinked
{
private:
doulinked *head;
doulinked *tail;
doulinked *prev;
doulinked *next;
T data;
public:
doulinked()
{
head=tail=prev=next=NULL;
T data;
}
void Inlist (doulinked *head);
void add(T d);
void insert_node();
void remove(doulinked* v);
void push_tail(T d);
void delete_front();
void delete_back();
void remove_ele (T d);
template <class U>
friend ostream & operator<<(ostream & os, const doulinked<U> & dll);
};
template<class U>
ostream & operator<<(ostream & os,const doulinked<U> & dll)
{
doulinked<U> * tmp = dll.head;
while (tmp)
{
os << tmp->data << " ";
tmp = tmp->next;
}
return os;
}
template<class T>
void doulinked<T>::add(T d)
{
doulinked *n = new doulinked;
n->data=d;
if( head == NULL)
{
head = n;
tail = n;
}
else
{
head->prev = n;
n->next = head;
head = n;
}
}
template<class T>
void doulinked<T>::push_tail(T d)
{
doulinked *n = new doulinked;
n->data=d;
if( tail == NULL)
{
head = n;
tail = n;
}
else
{
tail->next = n;
n->prev = tail;
tail = n;
}
}
template <class T>
void doulinked<T>::delete_front()
{
remove(head);
}
template <class T>
void doulinked<T>::delete_back()
{
remove(tail);
}
template <class T>
void doulinked<T>::remove(doulinked* v)
{
if(v->prev!=NULL && v->next!=NULL)
{
doulinked* p = v->prev;
doulinked* n = v->next;
p->next = n;
n->prev = p;
delete v;
}
else if(v->prev==NULL && v->next!=NULL)
{
doulinked* n =v->next;
head->next = n;
n->prev = head;
delete head;
head=n;
}
else if(v->prev!=NULL && v->next==NULL) // have some wrong with this loop;
{
doulinked* p=v->prev;
p->next=tail;
tail->prev=p;
delete tail;
tail=p;
}
}
template <class T>
void doulinked<T>::remove_ele(T d) // have some wrong with this loop
{
if(head->data==d)
{
remove(head);
head=head->next;
}
else
head=head->next;
}
int main()
{
doulinked<int> dll;
dll.add(5123);
dll.add(1227);
dll.add(127);
dll.push_tail(1235);
dll.push_tail(834);
dll.push_tail(1595);
dll.delete_front();
//dll.delete_back();
//dll.remove_ele(834);
cout<<dll<<endl;
system("pause");
}
Your design is a little confused.
The traditional C++ way to design a linked list (like std::list) has separate node and list classes, instead of a single class that acts as both:
template <typename T> struct node {
node *prev, *next;
};
template <typename T> struct list {
node *head, *tail;
};
If you want to just pass around node pointers, that's fine—but then you have to pass around node pointers, not node objects. And the mutator functions have to return a pointer as well—if you call delete_front on the head node, you've now got a reference to a deleted node; you need its next or you've lost any reference to the list. Since the constructor has to return a pointer, you can't use a real public constructor; you want a static factory method instead. And so on.
You also have to be consistent about whether there's a "sentinel node" before the head (and after the tail) or not. If you're creating a sentinel in your constructor—as you are doing—new nodes inserted at the end(s) need to point at the sentinel(s)—which you aren't doing.
Also, the whole head/tail notion you're using is wrong for a node API. (Also, it's incredibly confusing to mix and match names from different styles—you've got add matching delete_front and push_tail matching delete_back…) To have a push_tail method, you either have to walk the entire list (making it O(N)), or you have to have every node hold the tail pointer (making any list change O(N)), or you have to make the head hold a tail pointer and the tail hold a head pointer.
The last one works (it wastes a couple of pointers for every node when only one node needs each, but that rarely matters). But it gets confusing to think about.
It's actually a lot simpler to just create a circular list, where the head's prev points at the tail (or sentinel) instead of 0, and the tail's next points at the head (or sentinel) instead of 0. This gets you all the advantages of a separate list class, without needing that class—if you have a pointer to the head, that's all you need to refer to the entire list (because node is the head and node->prev is the tail, or or similarly if you have a sentinel).
Also, your constructor doesn't make much sense:
doulinked()
{
head=tail=prev=next=NULL;
T data;
}
This creates a local default-constructed T variable named data, and then… does nothing with it. You probably wanted to set data to something. And you probably wanted to use initializers for this. And in that case, you don't need to do anything, because that's already the default.
And I'm not sure what Inlist is even supposed to do.
As for remove_ele(T d), presumably you want to remove the first element whose data == d, right? If you write a find method first, then it's trivial: remove(find(d)). (I'm assuming that find throws an exception; if you want find to return null or the sentinel or something else instead, and remove_ele to return true or false, obviously you need one more line to check whether the find worked.)
If you don't know how to write a find method… well, that's kind of the whole point of a linked list, there's a trivial recursive definition for all traversal functions, including find:
node *node::find(T d) {
if (data == d) { return this; }
if (next) { return next->find(d); }
return 0;
}
Anyway, I think rather than try to bang on your code until it works, you should look at existing implementations of the various designs until you understand the differences, then pick the design you want and try to implement that.
I have a stl::map which key type is a custom struct. I want to know if this map already has a key with a specific string as component (noted as "id" below), whatever the value of its other components. Inspired by this answer and this one also, I try to use stl::find_if with a custom functor:
map<myStruct, vector<size_t> > myMap;
struct myStruct
{
string a, b, c, id;
};
struct checkId : unary_function<pair<myStruct, vector<size_t> >, bool>
{
private:
string _exp;
public:
checkId (myStruct x) : _exp(x.id) {}
bool operator() (const pair<myStruct, vector<size_t> > & p) const
{
return p.first.id.compare(_exp) == 0;
}
};
map<myStruct, vector<size_t> >::iterator it;
myStruct newS; // to be initialized, but not shown here
it_mP2P = find_if(myMap.begin(), myMap.end(), checkId(newS));
When I compile this, gcc returns me:
/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../include/c++/4.1.2/bits/stl_function.h: In member function ‘bool std::less<_Tp>::operator()(const _Tp&, const _Tp&) const [with _Tp = myStruct]’:
/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../include/c++/4.1.2/bits/stl_map.h:347: instantiated from ‘_Tp& std::map<_Key,_Tp, _Compare, _Alloc>::operator[](const _Key&) [with _Key = myStruct, _Tp = std::vector<long unsigned int, std::allocator<long unsigned int> >, _Compare = std::less<myStruct>, _Alloc = std::allocator<std::pair<const myStruct, std::vector<long unsigned int, std::allocator<long unsigned int> > > >]’
myprogram.cpp:386: instantiated from here
/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../include/c++/4.1.2/bits/stl_function.h:227: error: no match for ‘operator<’ in ‘__x < __y’
Does this mean that I have to overload the operator "<" to work with my custom struct if I want to use my functor "checkId"? How can I do this? I am not a C++ expert,so thanks in advance for any piece of code.
http://codepad.org/VYNqdZeF
struct myStruct
{
std::string a, b, c, id;
bool operator<(const myStruct& rhs) const //HERES THE MAGIC
{return id < rhs.id;} //WHEEEEEEEEEEEE!
};
This is required to make std::map<myStruct, stuff> work, without passing a custom comparison functor.
It's got nothing to do with checkId. The following demonstrates the same problem:
map<myStruct, vector<size_t> > myMap;
struct myStruct
{
string a, b, c, id;
};
std::map is a sorted map (it's usually implemented as a binary search tree)... the key type (myStruct in this case) must have operator< implemented simply to construct the map.
If you don't need fast lookups, you can just use a vector< pair< myStruct, vector<size_t> > >; if you need fast lookups, you have to add some kind of structure to myStruct. For a std::map this must be a comparison operator; alternatively you can use a std::unordered_map (in C++11; this is usually implemented as a hash table) -- in this case you will need to implement a hash function.
See the documentation of std::map and std::unordered_map for more details.
This has nothing to do with your use of find_if.
For a type to be a key of a map, it needs to be comparable, either by implementing operator< or by providing a comparator as a template parameter of the map.
Mooing Duck's answer is the easiest to do, but not the most flexible
if you go and write
struct myStruct
{
std::string a, b, c, id;
bool operator<(const myStruct& rhs) const
{return id < rhs.id;}
};
then all your maps that have myStruct as a key, will have to be sorted
by id.
Suppose you have one map that needs to be compared by a, and another one by b. If you put operator<() inside your myStruct, then you will have coupled the struct with its users, tightly, which is not good programming practice.
Instead, you could set a compare function to each map:
struct myStruct
{
std::string a, b, c, id; // keep your struct unchanged, and independent of clients
};
bool Compare_by_a(const myStruct &s1, const myStruct& s2) {
return s1.a < s2.a;
}
bool Compare_by_b(const myStruct &s1, const myStruct& s2) {
return s1.b < s2.b;
}
bool Compare_by_id(const myStruct &s1, const myStruct& s2) {
return s1.id < s2.id;
}
map<myStruct, vector<size_t>, Compare_by_a > map1;
map<myStruct, vector<size_t>, Compare_by_b > map2;
map<myStruct, vector<size_t>, Compare_by_id > map3;
I have the following:
typedef std::function<void(const EventArgs&)> event_type;
class Event : boost::noncopyable
{
private:
typedef std::vector<event_type> EventVector;
typedef EventVector::const_iterator EventVector_cit;
EventVector m_Events;
public:
Event()
{
}; // eo ctor
Event(Event&& _rhs) : m_Events(std::move(_rhs.m_Events))
{
}; // eo mtor
// operators
Event& operator += (const event_type& _ev)
{
assert(std::find(m_Events.begin(), m_Events.end(), _ev) == m_Events.end());
m_Events.push_back(_ev);
return *this;
}; // eo +=
Event& operator -= (const event_type& _ev)
{
EventVector_cit cit(std::find(m_Events.begin(), m_Events.end(), _ev));
assert(cit != m_Events.end());
m_Events.erase(cit);
return *this;
}; // eo -=
}; // eo class Event
And during compilation:
1>c:\program files (x86)\microsoft visual studio 10.0\vc\include\algorithm(41): error C2451: conditional expression of type 'void' is illegal
1> Expressions of type void cannot be converted to other types
Now, I understand this is because of what is being stored in the vector and the operator ==. Is there another way to store std::function in an STL container? Do I need to wrap it up in something else?
You can store boost::function in the vector, provided you don't use std::find. Since you seem to need this, wrapping the function in its own class with equality would be probably the best.
class EventFun
{
int id_;
boost::function<...> f_;
public:
...
bool operator==(const EventFun& o) const { return id_==o.id_; } // you get it...
};
Note that this requires you maintain the id_ in a sane way (eg. two different EventFuns will have different id_s, etc.).
Another possibility would be to store boost::functions with a tag the client would remember and use to identify the particular function on deleting it.