Is EA faster than MT4 internal series array Open[]?

It's much more likely that the EA is slow rather than the terminal being slow. The terminal is running a simple task which is well established. The EA is not fully tested. I give the EA a much higher chance of being the problem!

In the other thread you asked how I would handle the "gruesome" ArrayResize thing. I never actually answered although I did start formulating a response. The thing is that any time an ArrayResize happens the delay is quite badly defined. Thinking of this from a data structure point of view, I don't believe that MQL4 double arrays can be linked lists. New data has to be contiguous in order to make memory addressing simple. That forces a memory copy at some point when a new contiguous memory block is allocated. How long will this take? Well how big are the arrays? If you are logging tick data for days the arrays could start to get quite big.

To get an idea you could write a little script to measure it. Store the start time. Allocate a 100,000 element double array and fill with the index for each position ( bigArray[3]= 3 etc ) Increment the array size by 1000 elements in a loop say 100 times ( no need to fill the new elements). Store the end time. Print end time - start time. If the start and end times differ by more than a few seconds that may be your problem.

Well here is the script test code ...

double bigArray[1000000];

int init(){
   Comment( "start");
   for( int n=0; n<1000000; n++ )
      bigArray[n] = n;
}

int start(){
   
   Comment("");
   
   datetime start = TimeCurrent();
   for( int n=0; n<1000; n++ )
      ArrayResize(bigArray, 1000000 + n*1000);
   
   datetime end = TimeCurrent();
   Comment( "Duration was " + (end-start) + " seconds" ); 
   Print( "Duration was " + (end-start) + " seconds" ); 
   
   return(0);
}

It takes 17 seconds on my machine. I withdraw my objections to the ArrayResize. I guess they are using a low level hardware DMA transfer so it is blindly fast (17ms)

Hey dabbler,

Thank you for your input, and also for taking the time to read my long post :]

I ran your code several times, and it took about 4 seconds every time to complete on my machine (Core i5 @ 4GHz, 4GB RAM @ 1.6GHz).

So with that said, do you still think that the EA missed the ticks that Open[0] caught and not the other way around? To be honest I still tend to believe that my theory is correct, since it does seem very unlikely that there was a sharp change in price which my EA *didn't* catch, but then the next price in that same second which it *did* catch was exactly 1 pt. above the price it caught before for the previous second. It's harder to believe that there was such a spike in the price, then a few milliseconds later returned to the original price within a margin of 1 pt.

Of course this is still just speculation. To test this I altered my debug code to also output the Time/Bid of several ticks ahead in the Array, so if and when this happens again I'll be able to know if the EA catches these sharp price jumps or not.

I also think it's more likely, because Open[0] is not triggered by incoming ticks. It's trigger is the time of formation of a new bar. So it could be that at the exact millisecond at which Open[0] was triggered, a new tick was already incoming (i.e. "tick in progress"), which the EA caught, and then immediately after there was another tick with the price jump, and that's the one that Open[0] caught (and the EA as well).

What do you think?

Anyway I'll update on the result as soon as I have something! :)

nanquan

Okay, I have some updates with regards to this situation! Apparently it's even stranger than I thought!

First of all this happens only when there are disruptions in the data provided by the trading server. i.e. you will have a gap of several minutes with no incoming data, and then the next bar's Open price will suddenly jump drastically, because it's much farther ahead time wise. I contacted my broker about this, and this is caused due to overloads and disruptions on the DEMO server, since all demo accounts are setup on just one server. With LIVE accounts this will never happen since they span across multiple servers etc.

Regardless, this still poses a big question from a technical standpoint:

As you can see, there is a gap in data of approximately 17 minutes, from 15:23 to 15:40. The Open[] value of the bar right after the gap is 1.34146.

Here is the debug information from the TickData array:

2012.02.28 17:45:03     NEXUS-1 EURUSD,M15: Alert:
Index before: 1330442999-1.33978
Bar Open Time/Value: 1330443000-1.34146 <-- Time of 1330443000 is open time of new bar after the price gap, with Open[] value of 1.34146
Index from: 1330443000-1.33980 
1330443001-1.33983
1330443002-1.33987
1330443002-1.33994
1330443003-1.34003
1330443003-1.34000
1330443004-1.34001
1330443005-1.33999 <-- How can it be that 5 seconds later into that bar, the price recorded by EA internal 'TickData' array is 1.33999, when 'Low' value for that WHOLE bar is 1.34132 ?!?!?!?!?!?!?

As you can see, at the time of 1330443000 (new bar open time), Open[] price value is 1.34146.

At that same time, the value captured by internal 'TickData' array for time 1330443000 is 1.33980.

Now, even if EA had missed the "jump" in price that was recorded by Open[] at time of 1330443000, it should have still recorded the correct price right after the one it missed.

In other words, if we look at the attached chart image, we can see that after new bar opened at time of 1330443000, the price never went lower than 1.34132!

And yet, if we look at the data from internal EA 'TickData' array, the values recorded after the time of 1330443000, are still values which are nearer the price of the previous bar before the GAP.

So this makes no sense at all! How can it be that at time of 1330443005 (i.e. 5 whole seconds after the NEW bar Open time), the price recorded by 'TickData' array is 1.33999, when from looking at the chart data, after the time of 1330443000 the price never dipped below 1.34132!!!

Any ideas for a possible explanation? O_o

Thanks,

nanquan

Thanks for giving it a shot anyway, dabbler! If you think of anything please feel free to reply here:

https://forum.mql4.com/46568

As I've opened a new topic about this question now that I know what's happening!