Data struct own some relation-ship or logics internally.  Usually,
they are described in code spreaded over the project.  The problem of
code is deattach relationships from the data struct.  You don't see
the relationship with definition of data struct alone.  You can also
not get the relationship from the code easily.  So, we need a way
to describe relationship in the definition of data struct to attach
the information with data struct.

Attaching relationship is not only easy to read, but also parse-able.
Relationships are attached a part of definition of data struct.  They
are not code, it means more easy to be parsed.  A lot of code can be
generated automatically by tools.

Persisted data (in secondary storage) is usually a tree with reference
between branches while internal data struct is usually a group of
interconnected objects.  We dont use persisted one directly because
they are more hard to be used.  You need extra-logic to travel
persisted data struct.  We need the code to convert persisted data
struct to internal one and in reversed, but it is tedious.  With a
language to describe data struct with attaching relationship, we can
generate code for conversion of data struct from/to persisted to/from
internal ones.

** Construction

 - types
 - relationships
   - reference
   - dependency
 - constrain
 - condition
 - path