“继承"关系的关系另一个(1:N)关系 [英] relationship that "inherit" another (1:N) relationship
问题描述
我想要一个支持这些特定的1:N关系的数据结构:-
1#. Human
加薪 0-N Human
2#. Human
有 0-N Dog
3#. Human
耕种 0-N Tree
4#. Dog
是 0-N Parasites
的房子.
I want a data-structure that supports these specific 1:N relations :-
1#. Human
raise 0-N Human
2#. Human
has 0-N Dog
3#. Human
cultivate 0-N Tree
4#. Dog
is a house of 0-N Parasites
.
注意:
-这些关系中的国家都是暂时的,例如Human1
可能提高 Human2
,但是一年之后,Human1
可能会放弃Human2
.
-所有对象均继承自BaseObject
,并且具有唯一的int ID.
Note:
- State in these relations are all temporary e.g. Human1
may raise Human2
, but after a year, Human1
may abandon Human2
.
- All objects are inherited from BaseObject
and has unique int ID.
在以上所有关系中,我希望能够支持以下功能:-
F1.添加关系例如human_dog->addRelation(Human* a,Dog* b)
F2.删除关系human_dog->removeRelation(Human* a,Dog* b)
F3.查询所有孩子human_dog->getAllChildren(Human*)
F4.查询所有父项human_dog->getAllParents(Dog*)
F5.检查父母是否有> = 1个孩子
F6.检查孩子是否有> = 1个父母
F7.删除父母的所有孩子
F8.删除一个孩子的所有父母
In all of the above relation, I want to be able to support these features :-
F1. add relation e.g. human_dog->addRelation(Human* a,Dog* b)
F2. remove relation e.g. human_dog->removeRelation(Human* a,Dog* b)
F3. query all children e.g. human_dog->getAllChildren(Human*)
F4. query all parent e.g. human_dog->getAllParents(Dog*)
F5. check whether a parent has >=1 child
F6. check whether a child has >=1 parent
F7. remove all children for a parent
F8. remove all parent for a child
这可以通过std::unordered_map
或更容易定制的方式来实现.
This can be implemented by std::unordered_map
or something more customized quite easily.
我想将关系1#,2#,3#(即所有实线)标记为 Feed .
它必须以 聚集 样式支持功能F3-F8.
I want to mark relation 1#,2#,3# (i.e. all solid lines) as Feed.
It has to support feature F3-F8 in an aggregating style.
例如:-
-
feed->getAllChildren(BaseObject* b)
:
如果b
是人类,则必须返回b
的所有抚养,有和耕种的孩子. -
feed->removeAllParent(BaseObject* b)
:
如果b
是狗,它的作用类似于cultivate->removeAllParent(b)
.
feed->getAllChildren(BaseObject* b)
:
Ifb
is human, it must return all children of raise,has and cultivate of theb
.feed->removeAllParent(BaseObject* b)
:
Ifb
is a dog, it will effect likecultivate->removeAllParent(b)
.
总而言之,我希望能够轻松注入这样的 聚合 .
前任.调用:-
In summary, I want to be able to easily inject such aggregation.
Ex. It is useful to call :-
void BaseObject::declareForFreedom(){
feed->removeAllParent(this);
}
上面的示例仅显示4个关系和1个间接级别.
在我的真实情况下,这种继承/间接关系有8-10个关系和3-4个级别.
The above example shows only 4 relations and 1 level of indirection.
In my real case, there are 8-10 relations and 3-4 levels of such inherit/indirection.
什么是适合这种情况的数据结构/设计模式?
What is a data-structure/design-pattern that suitable for this case?
我目前为1#-4#创建一个自定义的1:N关系,并对每个 feed 函数进行硬编码.这很乏味.
我已经猛烈抨击了几个月,但没有发现任何看起来优雅的实现.
I currently create a custom 1:N relation for 1#-4#, and hard-code every feed's function. It is tedious.
I have banged by head for a few months, but not found any implementation that look elegant.
http://coliru.stacked-crooked.com/a/1f2decd7a8d96e3c
基本类型:-
#include <iostream>
#include <map>
#include <vector>
enum class Type{
HUMAN,DOG,TREE,PARASITE,ERROR
}; //for simplicity
class BaseObject{public: Type type=Type::ERROR; };
class Human : public BaseObject{
public: Human(){ type=Type::HUMAN; }
};
class Dog : public BaseObject{
public: Dog(){ type=Type::DOG; }
};
class Tree : public BaseObject{
public: Tree(){ type=Type::TREE; }
};
class Parasite : public BaseObject{
public: Parasite(){ type=Type::PARASITE; }
};
基本的1:N地图
template<class A,class B> class MapSimple{
std::multimap<A*, B*> aToB;
std::multimap<B*, A*> bToA;
public: void addRelation(A* b1,B* b2){
aToB.insert ( std::pair<A*,B*>(b1,b2) );
bToA.insert ( std::pair<B*,A*>(b2,b1) );
}
public: std::vector<B*> queryAllChildren(A* b1){
auto ret = aToB.equal_range(b1);
auto result=std::vector<B*>();
for (auto it=ret.first; it!=ret.second; ++it){
result.push_back(it->second);
}
return result;
}
public: void removeAllParent(B* b){
if(bToA.count(b)==0)return;
A* a=bToA.find(b)->second;
bToA.erase(b);
auto iterpair = aToB.equal_range(a);
auto it = iterpair.first;
for (; it != iterpair.second; ++it) {
if (it->second == b) {
aToB.erase(it);
break;
}
}
}
//.. other functions
};
这是数据库实例和聚合:-
Here is the database instance and the aggregation :-
MapSimple<Human,Human> raise;
MapSimple<Human,Dog> has;
MapSimple<Human,Tree> cultivate;
MapSimple<Dog,Parasite> isHouseOf;
class Feed{
public: void removeAllParent(BaseObject* b1){
if(b1->type==Type::HUMAN){
raise.removeAllParent(static_cast<Human*>(b1));
}
if(b1->type==Type::DOG){
has.removeAllParent(static_cast<Dog*>(b1));
}
//.... some other condition (I have to hard code them - tedious) ...
}
//other function
};
Feed feed;
用法
int main(){
Human h1;
Dog d1,d2;
has.addRelation(&h1,&d1);
has.addRelation(&h1,&d2);
auto result=has.queryAllChildren(&h1);
std::cout<<result.size(); //print 2
feed.removeAllParent(&d1);
result=has.queryAllChildren(&h1);
std::cout<<result.size(); //print 1
}
推荐答案
编辑
Jarod42 建议使用更好的代码
EDIT
Better code was suggested by Jarod42 in this topic. C++17 style:
#include <algorithm>
#include <tuple>
#include <vector>
class BaseObject {
public:
virtual ~BaseObject() = default;
virtual std::vector<BaseObject*> getAllParents() const = 0;
virtual std::vector<BaseObject*> getAllChildren() const = 0;
virtual void removeAllParents() = 0;
virtual void removeAllChildren() = 0;
};
template<typename TParentTuple, typename TChilderenTuple>
class Obj;
template<typename... ParentTags,
typename... ChildTags>
class Obj<std::tuple<ParentTags...>, std::tuple<ChildTags...>> : public BaseObject
{
std::tuple<std::vector<typename ParentTags::obj_type*>...> parents;
std::tuple<std::vector<typename ChildTags::obj_type*>...> children;
public:
template <typename T>
void addParent(T* parent) { std::get<std::vector<T*>>(parents).push_back(parent); }
template <typename T>
void removeParent(const T* parent) {
auto& v = std::get<std::vector<T*>>(parents);
auto it = std::find(std::cbegin(v), std::cend(v), parent);
if (it != std::cend(v)) { v.erase(it); }
}
template <typename T>
void addChild(T* child) { std::get<std::vector<T*>>(children).push_back(child); }
template <typename T>
void removeChild(const T* child) {
auto& v = std::get<std::vector<T*>>(children);
auto it = std::find(std::cbegin(v), std::cend(v), child);
if (it != std::cend(v)) { v.erase(it); }
}
std::vector<BaseObject*> getAllParents() const override {
std::vector<BaseObject*> res;
std::apply([&](auto&... v){ (res.insert(res.end(), v.begin(), v.end()), ...); },
parents);
return res;
}
std::vector<BaseObject*> getAllChildren() const override {
std::vector<BaseObject*> res;
std::apply([&](auto&... v){ (res.insert(res.end(), v.begin(), v.end()), ...); },
children);
return res;
}
void removeAllParents() override {
std::apply(
[this](auto&... v)
{
[[maybe_unused]] auto clean = [this](auto& v) {
for (auto* parent : v) {
parent->removeChild(this);
}
v.clear();
};
(clean(v), ...);
},
parents);
}
void removeAllChildren() override {
std::apply(
[this](auto&... v)
{
[[maybe_unused]] auto clean = [this](auto& v) {
for (auto* child : v) {
child->removeParent(this);
}
v.clear();
};
( clean(v), ...);
},
children);
}
};
struct Human_tag;
struct Tree_tag;
struct Dog_tag;
struct Parasite_tag;
using Human = Obj<std::tuple<>, std::tuple<Tree_tag, Dog_tag>>;
using Tree = Obj<std::tuple<Human_tag>, std::tuple<>>;
using Dog = Obj<std::tuple<Human_tag>, std::tuple<Parasite_tag>>;
using Parasite = Obj<std::tuple<Dog_tag>, std::tuple<>>;
struct Human_tag { using obj_type = Human; };
struct Tree_tag { using obj_type = Tree; };
struct Dog_tag { using obj_type = Dog; };
struct Parasite_tag { using obj_type = Parasite; };
template<class A, class B>
void addRelation(A* a, B* b)
{
a->addChild(b);
b->addParent(a);
}
#include <iostream>
int main() {
Human h1;
Dog d1, d2;
addRelation(&h1, &d1);
addRelation(&h1, &d2);
auto result = h1.getAllChildren();
std::cout << result.size() << "\n"; //print 2
d1.removeAllParents();
result = h1.getAllChildren();
std::cout << result.size() << "\n"; //print 1
}
旧代码:(我的尝试)
好的,因为您不想重复的代码,所以我一直将此项目作为元编程/可变模板的初次经验.这就是我得到的:
Old code: (my attempt)
OK, since you did not want duplicated code, I've been using this project as my first experience with metaprogramming/variadic templating. So this is what I got:
#include <tuple>
#include <vector>
#include <algorithm>
template<class T>
using prtVector = std::vector<T*>;
// Interface, as required by assignment
class BaseObject {
public:
virtual ~BaseObject() {}
virtual prtVector<BaseObject> getAllParents() const = 0;
virtual prtVector<BaseObject> getAllChildren() const = 0;
virtual void removeAllParents() = 0;
virtual void removeAllChildren() = 0;
};
// base prototype
template<typename TOwnTag, typename TParentTagsTuple, typename TChildTagsTuple>
class Obj;
// Parent-type deduction
template<typename TOwnTag, typename TParentTag, typename... TParentTags, typename... TChildTags>
class Obj<TOwnTag, std::tuple<TParentTag, TParentTags...>, std::tuple<TChildTags...>>
: public Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>
{
// local types
using TOwn = typename TOwnTag::obj_type;
using TParent = typename TParentTag::obj_type;
// container
prtVector<TParent> parentsPtrs;
//befriend types
friend class Obj;
template<class A, class B>
friend void addRelation(A* const a, B* const b);
protected:
// prevent base function hiding with 'using'-declaration
using Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::addParent;
using Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::removeParent;
// add and remove element functions
void addParent(TParent* const parentPtr) { parentsPtrs.push_back(parentPtr); }
void removeParent(TParent const* const parentPtr) {
auto it = std::find(std::cbegin(parentsPtrs), std::cend(parentsPtrs), parentPtr);
if (it != std::cend(parentsPtrs)) parentsPtrs.erase(it);
}
public:
virtual ~Obj() {}
virtual prtVector<BaseObject> getAllParents() const override {
auto result = Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::getAllParents();
result.insert(std::begin(result), std::cbegin(parentsPtrs), std::cend(parentsPtrs));
return result;
}
virtual prtVector<BaseObject> getAllChildren() const override {
return Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::getAllChildren();
}
virtual void removeAllParents() override {
Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::removeAllParents();
for (auto&& parent : parentsPtrs) parent->removeChild(reinterpret_cast<TOwn* const>(this));
}
virtual void removeAllChildren() override {
Obj<TOwnTag, std::tuple<TParentTags...>, std::tuple<TChildTags...>>::removeAllChildren();
}
};
// Child-type deduction
template<typename TOwnTag, typename TChildTag, typename... TChildTags>
class Obj<TOwnTag, std::tuple<>, std::tuple<TChildTag, TChildTags...>>
: public Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>
{
// local types
using TOwn = typename TOwnTag::obj_type;
using TChild = typename TChildTag::obj_type;
// container
prtVector<TChild> childrenPtrs;
//befriend types
friend class Obj;
template<class A, class B>
friend void addRelation(A* const a, B* const b);
protected:
// empty functions required for 'using'-declaration
void addParent() {}
void removeParent() {}
// prevent base function hiding with 'using'-declaration
using Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>::addChild;
using Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>::removeChild;
// add and remove element functions
void addChild(TChild* const childPtr) { childrenPtrs.push_back(childPtr); }
void removeChild(TChild const* const childPtr) {
auto it = std::find(std::cbegin(childrenPtrs), std::cend(childrenPtrs), childPtr);
if (it != std::cend(childrenPtrs)) childrenPtrs.erase(it);
}
public:
virtual ~Obj() {}
virtual prtVector<BaseObject> getAllParents() const override {
return Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>::getAllParents();
}
virtual prtVector<BaseObject> getAllChildren() const override {
auto result = Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>::getAllChildren();
result.insert(std::begin(result), std::cbegin(childrenPtrs), std::cend(childrenPtrs));
return result;
}
virtual void removeAllParents() override {}
virtual void removeAllChildren() override {
Obj<TOwnTag, std::tuple<>, std::tuple<TChildTags...>>::removeAllChildren();
for (auto&& child : childrenPtrs) child->removeParent(reinterpret_cast<TOwn* const>(this));
}
};
// terminator
template<typename TOwnTag>
class Obj<TOwnTag, std::tuple<>, std::tuple<>> : public BaseObject {
protected:
// empty functions required for 'using'-declaration
void addChild() {}
void removeChild() {}
void addParent() {}
void removeParent() {}
public:
virtual ~Obj() {}
virtual prtVector<BaseObject> getAllParents() const override {
return prtVector<BaseObject>();
}
virtual prtVector<BaseObject> getAllChildren() const override {
return prtVector<BaseObject>();
}
virtual void removeAllParents() override {}
virtual void removeAllChildren() override {}
};
//prototype class tags
struct Human_tag;
struct Tree_tag;
struct Dog_tag;
struct Parasite_tag;
//define class types
using Human = Obj<Human_tag, std::tuple<>, std::tuple<Tree_tag, Dog_tag>>;
using Tree = Obj<Tree_tag, std::tuple<Human_tag>, std::tuple<>>;
using Dog = Obj<Dog_tag, std::tuple<Human_tag>, std::tuple<Parasite_tag>>;
using Parasite = Obj<Parasite_tag, std::tuple<Dog_tag>, std::tuple<>>;
//couple tags to classes
struct Human_tag { using obj_type = Human; };
struct Tree_tag { using obj_type = Tree; };
struct Dog_tag { using obj_type = Dog; };
struct Parasite_tag { using obj_type = Parasite; };
//(befriend)helper function
// maybe could do somehting with std::enable_if
// i.e. "enable if type B is in child tuple of A and
// type A is in parent tuple of B"
// that way the parser will already detect a relation is not possible
template<class A, class B>
void addRelation(A* const a, B* const b)
{
a->addChild(b);
b->addParent(a);
}
// now for some testing
#include <iostream>
int main() {
Human h1;
Dog d1, d2;
Parasite p1;
addRelation(&h1, &d1);
addRelation(&h1, &d2);
addRelation(&d1, &p1);
//addRelation(&h1, &p1); // compiler error
auto result = h1.getAllChildren();
std::cout << result.size() << "\n"; //print 2
d1.removeAllParents();
result = h1.getAllChildren();
std::cout << result.size() << "\n"; //print 1
std::cin.ignore();
}
请问任何不清楚的问题,因为在过去的24小时里我一直在学习很多新东西,所以我不知道从哪里开始我的解释.
Please ask questions about anything that is unclear, because I've been learning so much new stuff over the past 24 hours, that I don't know where to begin with my explanation.
这篇关于“继承"关系的关系另一个(1:N)关系的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!