El 19/09/2016 a las 7:54, Ram escribió:

Hi All ,

I have created a boost::multi_index container for my use and it compiles successfully.
[...] The container I defined is ,

typedef boost::multi_index_ClassX<

  ClassX*,

  boost::multi_index::indexed_by<

boost::multi_index::hashed_unique<

   boost::multi_index::const_mem_fun<ParentOfClassX, Int64, &ParentOfClassX::getKey1>,

boost::multi_index::const_mem_fun<ParentOfClassX, Int64, &ParentOfClassX::getKey2>,

>

>

> MultiIndexedClassX;

I'm afraid this is seriously wrong (leaving aside the translation typos such as
"boost::multi_index_ClassX" instead of "boost::multi_index_container" etc.)
The structure of a definition of a multi_index_container is

multi_index_container<
    element_type,
    indexed_by<
      index_specifier1<index1 args>,
      index_specifier2<index2 args>,
      ...
    >
>

where each index_specifier is one of "hashed_unique", "hashed_non_unique",
"ordered_unique", etc. I assume your intention is to have *two* hashed indices
(by key1 and key2, respectively), so rather than

indexed_by<
    hashed_unique<const_mem_fun1,const_mem_fun2>
>

(which is what you've written), the definition has to follow this structure:

indexed_by<
    hashed_unique<const_mem_fun1>,
    hashed_unique<const_mem_fun2>
>

See the difference? Your original definition is just plain wrong, and the fact that it
seems to compile is sheer luck (or lack of it): once you start trying to use it,
compile-time errors will pop up. (if you're curious, Boost.MultiIndex is trying to use
your second const_mem_fun as a hash function of the first const_mem_fun, which,
of course, makes no sense, but this is what you instructed the lib to do).

I've written a small test program reproducing your scenario that you can play
with at

http://coliru.stacked-crooked.com/a/456f8546e51a29e6

Note that, here, sample.insert(&a) works without problems.

Joaquín M López Muñoz