I correct my logic, please read again my final post...

[Ramine]

I correct my logic, please read again my final post...

Hello...

Here is my final solution that i think works efficiently...

First it is better to do this:

I think that C++ and ADA and Object Pascal must make the data types
objects, and make datastructures objects and make IO objects etc. and
wrap those objects with a scalable reader-writer lock or transactional
memory because it is easy to do it this way, and the compiler designers
and implementors must test those objects correctly to ensure that they
work correctly, this will ensure a level of security that avoids race
conditions that is tolerable, so no need for pure functions and Mvars of
Haskel and Lisp.

And second you have to use set() and get() wrapped with a
scalable reader-writer lock or transactional memory to access the
properties of the object because it is easy to do it this way.

and third you can add also this:

About C++ and my following enhancement to it...

C++ must add properties that you can access only from the constructor
and can not be accessed from the methods of the object.

This way you will be able to implement more cleanly my following
implementation:

First you have to call a constructor that will put the global variables
as pointer that are shared by threads in there respective queues..

And after that each method that wants to access a shared global variable
will take the shared global variable from the queue and copy it in a
local variable and work with it locally, and after that it will put it
back in its global queue , and other threads will block waiting
for the variables on there respective queues, it's like message passing
and it's like pure functions with Mvars in Haskel and this mechanism is
good and will avoid race conditions in Object oriented programming.

If you say to me that pure functions of Lisp and Haskel are pure
functions , but with my previous implementation you still can
by error forget to put your global variables that is
shared by threads in the properties that can not be accessed
by the object or in there respective global queues, i will say that
my implementation easy the job for us, because defining what is shared
and putting it in the properties that can not be accessed from the
object is easy and is as easy as making an error on sequential
programming, so in my opinion this level of security can be tolerated in
Object oriented programming without using Functional programming, other
than that we also have Transactional memory that is composable and easy
to use to solve our parallel programming problems in Object oriented
programming.


For Deadlocks use this:

Use Lock Hierarchies to Avoid Deadlock

http://www.drdobbs.com/parallel/use-lock-hierarchies-to-avoid-deadlock/204801163


Thank you,
Amine Moulay Ramdane.