我是否需要将类型设为 POD 以使用内存映射文件持久化它？

Question

Pointers cannot be persisted directly to file, because they point to absolute addresses. To address this issue I wrote a relative_ptr template that holds an offset instead of an absolute address.

Based on the fact that only trivially copyable types can be safely copied bit-by-bit, I made the assumption that this type needed to be trivially copyable to be safely persisted in a memory-mapped file and retrieved later on.

This restriction turned out to be a bit problematic, because the compiler generated copy constructor does not behave in a meaningful way. I found nothing that forbid me from defaulting the copy constructor and making it private, so I made it private to avoid accidental copies that would lead to undefined behaviour.

Later on, I found boost::interprocess::offset_ptr whose creation was driven by the same needs. However, it turns out that offset_ptr is not trivially copyable because it implements its own custom copy constructor.

Is my assumption that the smart pointer needs to be trivially copyable to be persisted safely wrong?

If there's no such restriction, I wonder if I can safely do the following as well. If not, exactly what are the requirements a type must fulfill to be usable in the scenario I described above?

struct base {
    int x;
    virtual void f() = 0;
    virtual ~base() {} // virtual members!
};

struct derived : virtual base {
    int x;
    void f() { std::cout << x; }
};

using namespace boost::interprocess;

void persist() {
    file_mapping file("blah");
    mapped_region region(file, read_write, 128, sizeof(derived));
    // create object on a memory-mapped file
    derived* d = new (region.get_address()) derived();
    d.x = 42;
    d->f();
    region.flush();
}

void retrieve() {
    file_mapping file("blah");
    mapped_region region(file, read_write, 128, sizeof(derived));
    derived* d = region.get_address();
    d->f();
}

int main() {
    persist();
    retrieve();
}

---

Thanks to all those that provided alternatives. It's unlikely that I will be using something else any time soon, because as I explained, I already have a working solution. And as you can see from the use of question marks above, I'm really interested in knowing why Boost can get away without a trivially copyable type, and how far can you go with it: it's quite obvious that classes with virtual members will not work, but where do you draw the line?

Answer 1

To avoid confusion let me restate the problem.

You want to create an object in mapped memory in such a way that after the application is closed and reopened the file can be mapped once again and object used without further deserialization.

POD is kind of a red herring for what you are trying to do. You don't need to be binary copyable (what POD means); you need to be address-independent.

Address-independence requires you to:

• avoid all absolute pointers.
• only use offset pointers to addresses within the mapped memory.

There are a few correlaries that follow from these rules.

• You can't use virtual anything. C++ virtual functions are implemented with a hidden vtable pointer in the class instance. The vtable pointer is an absolute pointer over which you don't have any control.
• You need to be very careful about the other C++ objects your address-independent objects use. Basically everything in the standard library may break if you use them. Even if they don't use new they may use virtual functions internally, or just store the address of a pointer.
• You can't store references in the address-independent objects. Reference members are just syntactic sugar over absolute pointers.

Inheritance is still possible but of limited usefulness since virtual is outlawed.

Any and all constructors / destructors are fine as long as the above rules are followed.

Even Boost.Interprocess isn't a perfect fit for what you're trying to do. Boost.Interprocess also needs to manage shared access to the objects, whereas you can assume that you're only one messing with the memory.

In the end it may be simpler / saner to just use Google Protobufs and conventional serialization.

Answer 2

Yes, but for reasons other than the ones that seem to concern you.

You've got virtual functions and a virtual base class. These lead to a host of pointers created behind your back by the compiler. You can't turn them into offsets or anything else.

If you want to do this style of persistence, you need to eschew 'virtual'. After that, it's all a matter of the semantics. Really, just pretend you were doing this in C.

Answer 3

Even PoD has pitfalls if you are interested in interoperating across different systems or across time.

You might look at Google Protocol Buffers for a way to do this in a portable fashion.

Answer 4

Not as much an answer as a comment that grew too big:

I think it's going to depend on how much safety you're willing to trade for speed/ease of usage. In the case where you have a struct like this:

struct S { char c; double d; };

You have to consider padding and the fact that some architectures might not allow you to access a double unless it is aligned on a proper memory address. Adding accessor functions and fixing the padding tackles this and the structure is still memcpy -able, but now we're entering territory where we're not really gaining much of a benefit from using a memory mapped file.

Since it seems like you'll only be using this locally and in a fixed setup, relaxing the requirements a little seems OK, so we're back to using the above struct normally. Now does the function have to be trivially copyable? I don't necessarily think so, consider this (probably broken) class:

   1 #include <iostream>
   2 #include <utility>
   3 
   4 enum Endian { LittleEndian, BigEndian };
   5 template<typename T, Endian e> struct PV {
   6         union {
   7                 unsigned char b[sizeof(T)];
   8                 T x;
   9         } val;  
  10         
  11         template<Endian oe> PV& operator=(const PV<T,oe>& rhs) {
  12                 val.x = rhs.val.x;
  13                 if (e != oe) {
  14                         for(size_t b = 0; b < sizeof(T) / 2; b++) {
  15                                 std::swap(val.b[sizeof(T)-1-b], val.b[b]);
  16                         }       
  17                 }       
  18                 return *this;
  19         }       
  20 };      

It's not trivially copyable and you can't just use memcpy to move it around in general, but I don't see anything immediately wrong with using a class like this in the context of a memory mapped file (especially not if the file matches the native byte order).

Update:
Where do you draw the line?

I think a decent rule of thumb is: if the equivalent C code is acceptable and C++ is just being used as a convenience, to enforce type-safety, or proper access it should be fine.

That would make boost::interprocess::offset_ptr OK since it's just a helpful wrapper around a ptrdiff_t with special semantic rules. In the same vein struct PV above would be OK as it's just meant to byte swap automatically, though like in C you have to be careful to keep track of the byte order and assume that the structure can be trivially copied. Virtual functions wouldn't be OK as the C equivalent, function pointers in the structure, wouldn't work. However something like the following (untested) code would again be OK:

struct Foo { 
    unsigned char obj_type;
    void vfunc1(int arg0) { vtables[obj_type].vfunc1(this, arg0); }
};

Answer 5

That is not going to work. Your class Derived is not a POD, therefore it depends on the compiler how it compiles your code. In another words - do not do it.

by the way, where are you releasing your objects? I see are creaing in-place your objects, but you are not calling destructor.

Answer 6

Absolutely not. Serialisation is a well established functionality that is used in numerous of situations, and certainly does not require PODs. What it does require is that you specify a well defined serialisation binary interface (SBI).

Serialisation is needed anytime your objects leave the runtime environment, including shared memory, pipes, sockets, files, and many other persistence and communication mechanisms.

Where PODs help is where you know you are not leaving the processor architecture. If you will never be changing versions between writers of the object (serialisers) and readers (deserialisers) and you have no need for dynamically-sized data, then PODs allow easy memcpy based serialisers.

Commonly, though, you need to store things like strings. Then, you need a way to store and retrieve the dynamic information. Sometimes, 0 terminated strings are used, but that is pretty specific to strings, and doesn't work for vectors, maps, arrays, lists, etc. You will often see strings and other dynamic elements serialized as [size][element 1][element 2]… this is the Pascal array format. Additionally, when dealing with cross machine communications, the SBI must define integral formats to deal with potential endianness issues.

Now, pointers are usually implemented by IDs, not offsets. Each object that needs to be serialise can be given an incrementing number as an ID, and that can be the first field in the SBI. The reason you usually don't use offsets is because you may not be able to easily calculate future offsets without going through a sizing step or a second pass. IDs can be calculated inside the serialisation routine on first pass.

Additional ways to serialize include text based serialisers using some syntax like XML or JSON. These are parsed using standard textual tools that are used to reconstruct the object. These keep the SBI simple at the cost of pessimising performance and bandwidth.

In the end, you typically build an architecture where you build serialisation streams that take your objects and translate them member by member to the format of your SBI. In the case of shared memory, it typically pushes the members directly on to the memory after acquiring the shared mutex.

This often looks like

void MyClass::Serialise(SerialisationStream & stream)
{
  stream & member1;
  stream & member2;
  stream & member3;
  // ...
}

where the & operator is overloaded for your different types. You may take a look at boost.serialize for more examples.