Menu

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

Object-Oriented Programming 4 replies

Become Your Best Trader: Process-oriented Trading Goals 0 replies

MLQ4 check if object selected 2 replies

user selection of Object color 2 replies

Left align object descriptions in MT4? 0 replies

Object Oriented Programming

  • #41
  • Quote
  • May 5, 2010 8:36am May 5, 2010 8:36am
Well said 7bit - regarding OOP. By the way, I saw that you succeeded embed python into MT. I am interested in this too and possibly many others. Do you think is worth to open a thread dedicated to Python integration? Just to have all the info centralized.

Cheers
Trading = a mirror to your human flaws. Fix them or be fixed.
  • #42
  • Quote
  • Edited 10:22am May 5, 2010 10:11am | Edited 10:22am
Quoting 7bit
It is no surprise to hear this from a C guy, you C programmers have not understood OOP in any way. These examples are only minor syntactical differences, they don't make programming any different, they don't change the way the program flow would be organized, they just make the code look slightly different.

The main advantage of OOP comes from the ability of late binding, overriding methods when you inherit from objects that already have some functionality. You can design libraries that can not only be called by your application, these...
I have programmed in OO lang for that last 6-7 years so I understand it fully. I just thinks its overused, especially inhertiance which I have never found the need to use. Any programmer I ever respected thinks the same, I went through the stage you are at thinking OO was amazing and the only way, but once you get some expereince you will realise its over hyped.

Ill just quote someone to clear things up

"OOP is to writing a program, what going through airport security is to flying"

"The problem with object-oriented languages is they've got all this implicit environment that they carry around with them. You wanted a banana but what you got was a gorilla holding the banana and the entire jungle."
Economists have forecast 9 out of the last 5 recessions
  • #43
  • Quote
  • Edited 10:51am May 5, 2010 10:38am | Edited 10:51am
Quoting markmm
I have programmed in OO lang for that last 6-7 years so I understand it fully. I just thinks its overused, especially inhertiance which I have never found the need to use.
Like I said, you haven't yet understood it. And you even admit it with this last sentence.

Maybe another 6 or 7 years programming will help. I have been where you are now (not understanding the concepts of OOP) when I was younger, only knowing assembly language and procedural TurboPascal about 20 years ago. It was around 1990 when the light bulb went on when I finally understood what a virtual method is and what it is used for.
  • #44
  • Quote
  • May 5, 2010 10:57am May 5, 2010 10:57am
Quoting 7bit
Like I said, you haven't yet understood it. And you even admit it with this last sentence.

Maybe another 6 or 7 years programming will help. I have been where you are now (not understanding the concepts of OOP) when I was younger, only knowing assembly language and procedural TurboPascal some 20 years ago.
You couldnt hack it using a lower lever language so you now use like the rest of the mediocre programmers out there a toy language that protects them from the complexity of a computer.

Thats fine, but while your programs is loading the runtime and checking the bounds over every array, mine will have completed.
Economists have forecast 9 out of the last 5 recessions
  • #45
  • Quote
  • May 5, 2010 11:10am May 5, 2010 11:10am
Quoting markmm
Thats fine, but while your programs is loading the runtime and checking the bounds over every array, mine will have completed.
There is no point in discussing this further. Nobody here was talking about bounds checking for arrays or runtime libraries. It was about late binding and callback functions and a nice syntax to abstract them. Even in C you would need to implement method tables with function pointers and modern OO language compilers do exactly the same on the machine level and don't come with any more overhead than the same implementation in C.

It seems you not only don't understand the problem and the solution, you don't even want to understand it. Fortunately this is entirely your problem, not mine.
  • #46
  • Quote
  • May 5, 2010 11:25am May 5, 2010 11:25am
Quoting 7bit
There is no point in discussing this further. Nobody here was talking about bounds checking for arrays or runtime libraries. It was about late binding and callback functions and a nice syntax to abstract them. Even in C you would need to implement method tables with function pointers and modern OO language compilers do exactly the same on the machine level and don't come with any more overhead than the same implementation in C.

It seems you not only don't understand the problem and the solution, you don't even want to understand it. Fortunately...
My issue is you are saying Cand C++ have done a lot of damage somehow, that is like blaming knives for knife crime. Just because you couldnt program in a powerful language that mose OS's, Games, compiler runtimes are written doesnt mean they are bad, all it means is you are below average or havent had a decent education in Computer Science. I see many people in the business like you and it is no wonder projects are late, code is over engineered garbage. They read a few books on OO and think they can architect complex systems.
Economists have forecast 9 out of the last 5 recessions
  • #47
  • Quote
  • May 5, 2010 11:27am May 5, 2010 11:27am
Ok guys, enough is enough , let's go back to useful discussions. Pleeease?
Trading = a mirror to your human flaws. Fix them or be fixed.
  • #48
  • Quote
  • May 5, 2010 12:07pm May 5, 2010 12:07pm
I am not saying OO is usless, but dogmatic adhernce to it is a mistake. Using interfaces to decouple components of a system can be done easily in C for eaxample.
Economists have forecast 9 out of the last 5 recessions
  • #49
  • Quote
  • May 5, 2010 12:48pm May 5, 2010 12:48pm
In terms of Metatrader, how can the principles of OOP be applied?

I spoke with a few CSE majors in school, and they're basically telling me that I can write my trading logic once, and then instantiate that trading logic several times with different parameters (Pair, timeframe, settings, etc).

I'm obviously quite clueless about OOP and I'm trying to figure out how to navigate this learning curve effectively.
  • #50
  • Quote
  • May 5, 2010 3:14pm May 5, 2010 3:14pm
Quoting Ronald Raygun
In terms of Metatrader, how can the principles of OOP be applied?
.
OOP is really effective in large & very large applications. EAs are far away from the concept of large applications, at least the EAs I know. I'd say thay if you don't know OOP don't bother learn it just because MT guys say is cool. I'm yet to understand why they've undertaken such big changes to MQL.

P.S. Language zealots, please don't shoot me!
Trading = a mirror to your human flaws. Fix them or be fixed.
  • #51
  • Quote
  • May 5, 2010 4:02pm May 5, 2010 4:02pm
Well, what sort of uses does OOP have now? (The very large applications which you speak of)

I'm thinking finally being able to combine my set of EAs into one super EA. Definitely would save me time maintaining each EA. Either that, or maybe a more flexible template than the one I currently use.
  • #52
  • Quote
  • May 5, 2010 6:08pm May 5, 2010 6:08pm
Quoting Ronald Raygun
Well, what sort of uses does OOP have now? (The very large applications which you speak of)

I'm thinking finally being able to combine my set of EAs into one super EA. Definitely would save me time maintaining each EA. Either that, or maybe a more flexible template than the one I currently use.
It's hard to explain without showing. Let's just say that it's much easier writing a large application like MetaTrader or Microsoft Word in C++ than in C. OOP allows you to do the same stuff in less lines of code, and in a standard way. You can simulate classes and polymorphism in C, but then if you are new to the code you must check that it actually does what you think it does. With a standard OOP syntax like C++, it's obvious what "virtual int computeSize()" does (ie: compute some size in a polymorphic way), you don't have to check how that particular project decided to simulate classes. This allows you to concentrate less on the mechanics and more on the meaning.

OOP would allow you for example to write a few money-management classes, which you could reuse instead of copy-pasting the same code over and over again. You could have an AggressivePyramiding MM class which you could use together with a particular strategy. This is code reuse. Now some theory: there are two ways of doing this - inheriting and composition. In general composition is better because it reduces coupling between code modules, making it easier to modify one without affecting the other: http://stackoverflow.com/questions/4...er-inheritance

As you can see, this stuff goes quite deep. Learning to use OOP effectively is the kind of stuff which takes years of practice, not months.

So don't expect fast progress. Your first OOP designs will be very bad, possibly worse than your current ones. And you won't realize that they are bad. When I look at my first Python programs, they are just C++ ones written with Python syntax. Because I was still thinking like a C++ programmer. It took a few years to understand how stuff is properly done in Python and how to think in new patterns. Now when I write code in C++, I use a few patterns I learned in Python. So in a way learning Python made me a better C++ programmer.

In the same way, learning OOP will make you a better programmer, even if you will not use OOP further down the line.
  • #53
  • Quote
  • May 5, 2010 6:16pm May 5, 2010 6:16pm
Quoting soso
I'm yet to understand why they've undertaken such big changes to MQL.
Maybe to attract more professional coders, which see current MQL as a toy language? But it will still remain an easy to use but very limited proprietary language.
  • #54
  • Quote
  • Edited 7:06am May 6, 2010 5:39am | Edited 7:06am
Quoting Ronald Raygun
In terms of Metatrader, how can the principles of OOP be applied?
I will give you an example. I recently had the following problem. I had a trading strategy that was like:

Inserted Code
extern int param1;
extern int param2;

int start(){
   int signal = getSignal(param1, param2, 0);
   if (signal>0){
      stayLong();
   }else{
      stayShort();
   }
}
and the interesting part, the core of my strategy was defined in the function getSignal:
Inserted Code
int getSignal(int p1, int p2, int shift){
   [... some complicated things involving a couple of indicators]
   if (some condition){
      return(1);
   }else{
      return(-1);
   }
}
Until here it is nothing special, Now fasten your seatbelts: I wanted to optimize this thing (the two parameters p1 and p2) during runtime, based on the last thousand bars. Here a quick first hack to test the concept. I would call this function once every few bars and it would calculate me the best set of parameters p1 and p2 that would have given the most profit during the last 1000 bars:

Inserted Code
void runtimeOptimizer(int &p1, int &p2){
   int bar;
   int signal;
   int p1_, p2_;
   double profit, profit_best;
   // search the whole parameter space
   for (p1_=0; p1_<100; p1_++){
      for (p2_=0; p2_<100; p2_++){
         profit = 0;
         for(bar = 1000; bar >0; bar--){
            [color=Red]signal = getSignal(p1_, p2_, bar);

[/color]            [...] // simulate the trading like in start() and count the profit
         }
         if (profit > profit_best){
            profit_best = profit;
            p1 = p1_;
            p2 = p2_;
         }
      }
   }
}
This gave promising result. I thought wow, this might be a useful idea. Lets include all kinds of bells and whistles in this runtime optimizer and make it a generic thing that i can reuse.

No way! I have reached the end of MQL4!

There is NO way in mql4 to make the above a function (or something else) that can accept arbitrary strategies. I can not put it into a library or an include file because the line marked in red must be changed and the whole thing recompiled for every different strategy i want to use this thing with. In the following I will refer to this dilemma as "The Problem". Every time you find yourself copy&pasting big snippets of code from one EA to the next and then just have to change 1 line where the logic is slightly different or it will call a different function but all the rest stays the same you might have run across yet another instance of The Problem. If you carefully think about it then you will find that you run across this problem all the time and an elegant solution to this problem is urgently needed.

Unfortunalely mql4 is a very poor language. I can not just declare my optimizer function like this:
Inserted Code
// ********** declared in my library optimizer.mqh *******

void runtimeOptimizer(int &p1, int &p2, [color=Red]funtion signalfunc[/color]){
   [...]
   signal = [color=Red]signalfunc[/color](...)
   [...]
}




// ********** and used in my EA ********
#include <optimizer.mqh>

int p1, p2;

int [color=Blue]getSignal[/color](int p1, int p2, int shift){
   // my holy grail strategy here
}

int start(){
   if(onceAnHour()){
      // optimize once an hour
      runtimeOptimizer(p1, p2, [color=Blue]getSignal[/color]);
   }
   int signal = getSignal(p1, p2, 0);
   // trade as above in the first example
}
and I also cannot just make the optimizer a class in mql4 like this (pseudo syntax, long time no java anymore):
Inserted Code
// ********** declared in my library optimizer.mqh *******

class runtimeOptimizer{
   void optimize(int &p1, int &p2){
      [...]
      signal = signalfunc(...)
      [...]
   }
   
   int signalfunc(int p1, int p2, int shift){
      // dummy function. override it.
   }
}




// ********** and used in in my EA ********
#include <optimizer.mqh>

int p1, p2;

int [color=Blue]getSignal[/color](int p1, int p2, int shift){
   // my holy grail strategy here
}

class myOptimizer [color=Red]extends[/color] runtimeOptimizer{
   
   // by overriding this method we "inject" our own callback here
   [color=Red]@Override[/color]
   int signalfunc(int p1, int p2, int shift){
      return([color=Blue]getSignal[/color](p1, p2, shift));
   }
}

int start(){
   if(onceAnHour()){
      // optimize once an hour
      myOptimizer o = new myOptimizer();
      o.optimize(p1, p2);
   }
   int signal = getSignal(p1, p2, 0);
   // trade as above in the first exmple
}
As you can see, the functional approach with passing a function to the function is even more elegant than the OOP one. There are OOP languages around with much less syntactic noise, so this example might not be the best, but it is the first one that came to my mind because I just recently had to fight with exactly this problem.

Note that there is absolutely no proper solution to this problem in plain mql4.

PS: Please don't suggest splitting up the optimizer function body into two incomplete halves in separate files that must be included immediately before and after the call to getSignal(). This would probably be the way a php programmer who has never heard about template inheritance would approach this problem but it would create an unmaintainable mess. I don't consider this a solution to the problem.
  • #55
  • Quote
  • May 6, 2010 10:00am May 6, 2010 10:00am
If you are serious about building ATS's then you would never use MT4/5. Try searching for FIX, you want to escape the shackles and contraints and roll your own solution. You can then use any language you want, C, C++, Java , C#, Python, Ruby, Scheme, CLisp, Haskell, Erlang, Ocaml etc, or mix and match. I currenly implement the socket based stuff, strategy anaylsis (monte carlo etc) in C and UI in C++ with some Python for non critical stuff.
Economists have forecast 9 out of the last 5 recessions
  • #56
  • Quote
  • May 6, 2010 10:22am May 6, 2010 10:22am
Quoting markmm
If you are serious about building ATS's then you would never use MT4/5. Try searching for FIX [...] You can then use any language you want, C, C++, Java , C#, Python, Ruby, Scheme, CLisp, Haskell, Erlang, Ocaml
This thread is about how OOP would be useful in writing code for Metatrader (or any other platform).

You stated in your previous postings that OOP is not useful for anything, so why are you now advertising all the "Toy-Languages" [your own words] C++, Java, C#, Python, Ruby, Scheme, LISP, Haskell, Erlang, Ocaml, etc.? I thought naked procedural C was the holy grail for all programming problems, no matter how complex they might be? What was your point again?
  • #57
  • Quote
  • May 6, 2010 10:30am May 6, 2010 10:30am
Quoting 7bit
This thread is about how OOP would be useful in writing code for Metatrader (or any other platform).

You stated in your previous postings that OOP is not useful for anything, so why are you now advertising all the "Toy-Languages" [your own words] C++, Java, C#, Python, Ruby, Scheme, LISP, Haskell, Erlang, Ocaml, etc.? I thought naked procedural C was the holy grail for all programming problems, no matter how complex they might be? What was your point again?
I may have overstated things, I aplogise if I was over the top, its clear you are an expericned programmer. I think there are many useful paradigms in programming and they all can up the level of abstraction to varying degrees.

All I was saying is don't denouce procedural programming using C as dangerous, it has it places. OO programming is useful specially for UI work but its also not a holy grail. I like to have the option of using anthing that fits best, not being forced into wrapping everying in an object. Python, C++ are a bit more open like this. Also as you say functional prgoramming can make thing eleagant and can be powerful.
Economists have forecast 9 out of the last 5 recessions
  • #58
  • Quote
  • May 6, 2010 10:57am May 6, 2010 10:57am
Quoting markmm
All I was saying is don't denouce procedural programming using C as dangerous, it has it places.
C is not dangerous because of procedural programming, C/C++ are dangerous because although they are static languages they don't provide any sort of type safety, not even remotely. Programmers are encouraged to cast things around and use pointers to naked memory locations for every little thing. This language doesn't even support strings, you have to emulate them on your own with arrays! Programming novices start messing around with pointers already in their very first hello world program and the compiler will allow things that would not even compile in any other sane language. I'm only talking about compile time checks here, not runtime, in case you are concerned about performance! And many useful (optional) runtime checks are not even possible because the compiler has no clue what the program is meant to do and what to check for and the main reason for this is the very limited expressive power of the language itself.
  • #59
  • Quote
  • Edited 11:20am May 6, 2010 11:09am | Edited 11:20am
Quoting 7bit
C is not dangerous because of procedural programming, C/C++ are dangerous because although they are static languages they don't provide any sort of type safety, not even remotely. Programmers are encouraged to cast things around and use pointers to naked memory locations for every little thing. This language doesn't even support strings. Programming novices start messing around with pointers already in their very first hello world program and the compiler will allow things that would not even compile in any other sane language. I'm only talking...
That is true, it is potentially dangerous if you dont know what you are doing and you are right the compiler lets you away with murder, but there are tools like valgrind that can analyse your program for memory leaks and memory vialotions, hopefully after a lot of experince this woluld be a quick last check and the fixes would be minimal.

I think its good for programmers to understand what is happinign behind the scenes, and a lot of these checks become crutches and stop the prorammer from thinking about what they are coding. Hack and fix style developement leads to bad code.
Economists have forecast 9 out of the last 5 recessions
  • #60
  • Quote
  • May 6, 2010 11:18am May 6, 2010 11:18am
Quoting 7bit
C is not dangerous because of procedural programming, C/C++ are dangerous because although they are static languages they don't provide any sort of type safety, not even remotely. Programmers are encouraged to cast things around and use pointers to naked memory locations for every little thing. This language doesn't even support strings, you have to emulate them on your own with arrays! Programming novices start messing around with pointers already in their very first hello world program and the compiler will allow things that would not even...
a few things. the point of C is that it's a half-way point between ASM and higher languages like VB and Pascal etc. you have the power AND responsibilities to do things correctly. if you do it in such a way that might be problematic then that is your right as well, but also your own fault. it's playing with knives, and if you don't have the skills to manage the risks then it's your own fault if you get cut. if you have the skills then you can perform some amazing feats that would be difficult or bloated in other languages.

a string IS an array... in any language on any computer. it's not always apparent or used like an array, but it is an array nonetheless, so refer to previous paragraph.

binary / asm is like being god. C is like controlling strings attached to the hands of god. higher-level languages are like sitting on a comfy chair next to god describing things you would like to happen and letting him do it his way and using his rules.
Congratulations ScalaFX for coming out of the closet!
0 traders viewing now