EFS Questions from z11

[z11]

aahh, so effectively for my objective the answer to my first question is No as well...thanks Jason.

By the way, is there any other way in eSignal that I can achieve these two tasks?

[gspiker]

Z11,

For number 1 you could, in many (but not all) cases, use Universal Globals or data files. For example if all you wanted to do is create an index of 50 symbols, you could run 50/7 = 8 (its new math!) EFS's each which calculate the average of 6-7 symbols and report it up the chain to a master EFS, which takes all 8 globals (holding the averages) and takes another average of these. I have never tried this, but I suspect it will work (unless eSignal added a heck of lot of logic to their engine to block such things...which i doubt).

G

[z11]

Hi Garth ... that was what I was afraid I might have to do to accomplish this except the whole thing would involve more convulated logic and studies which I have to build internally as only native studied can be called is the one on the symbol in the chart ...I will try to hack something this weekend and report if that would work.

on another note, again today (this morning before and after job data) the main efs (the one that "gets" global vairables) did not go through its iteration even though the chart in which it resides was updating just fine...so it could not have been data got slow down...the reason I believe it did not go through iteration was because I had a simple (not convoluted) test if the system is in a position and is in the profit for certain amount and then mkt retraces a certain amount it suppose to exit ...so the only data involved in this test is couple of local variables and the raw price (close())... no globals involved... well in EC it got long at 12848 and rode it to 12905 and did not get out until 12875!!! so that suggests somehow the EFS didnot get to go through its iteration until at least much later...

Any idea why that EFS would get stale? and not going through its iteration?

[gspiker]

Z11,

I'm not sure I completely understand what you are saying. It sounds like you think the EFS wasn't run at all for the ticks that happened during this period? I have never experienced this for code which was properly written, so if this is what you are saying I really wouldn't have a clue.

If you are saying that the globals were not being updated, I haven't seen this either, but there is some latency associated with the setting and getting Universal Globals, so I could imagine that if you were using a small enough interval, or if the market was moving fast enough, that data could be missed due to the latency. I believe Steve has mentioned has mentioned that he has seen globals being missed as well (but also had a corrective action for this case).

In the case where you think globals are being missed, you might try either time stamping data, to verify, or alternately (and a better solution), try to do a synchronous handshake (ack/nack) and buffer data until you get a confirmation that the data was read. This would add complexity, but should work. It will not fix any latency issues associated with this however (ie: data could lag, but at least you will get all the data).

as an aside: I use globals on 5 minutes charts, mostly ES, IBM, and such and haven't ever noticed globals being lost....

Garth

[z11]

Is there a way to force an EFS to reload itself on a time threshold from within iteself...say on every new minute do a force reload (effectively a recalculation) of that efs?

The phenomenon I am observing is that the signal that efs gives is different from what it will show if it get reloaded...so I am thinking on the last tick of the bar that signal is not being updated but when reloaded the logic will take into consideration of that now historical bar and calculates the signal based on intended logic.

I don't have computeonclose set on that EFS.

Any thoughts are appreciated...thanks.

[gspiker]

Z11,

Such behavior points to a mistake in the EFS's logic. I would closely review what code you have in checks for newbar and where those blocks of code exist relative to displaying of the data and alert generation.


I have fixed a ton of other peoples code that showed this type of behavior. It was all internal logic...so unless you discoverd a new bug, look to your logic.

G

[z11]

Hi Garth,

In this particular EFS, I don't check for newbar state ...it just updates and does calculation on every tick (iteration) and then displays a final result (a colored bar) in that efs pane.

It does not have an overall condition such as if (barIndex<0) for the main body of code....it "should" wash straight through on each new tick.

[eSignal_JasonK]

Hello z11,

My experience agrees with Garth's comments about the problem you are describing. I personally haven't come across this situation that wasn't caused by an error in the code logic. The only way to tell for sure is to see your code. Please post your code or a stripped down version that reproduces the problem so we can give you a more specific solution.

[Guest]

Z11,

I have also had similar experiences along the lines of what you are describing. At this point, I can only guess what the most likely issue is, but I am in agreement with what Garth and Jason discussed.

There is code that has been posted that addresses similar issues. I recommend you look at some of the efs's that either Garth, Jason, Alex, or myself have posted on the board for a solution that may apply.

[gspiker]

Z11,


And what exactly do you think this condition will be true (this is a test, you will be timed, and a wrong answer is punishable ;-)...

(barIndex<0)


I suspect this is similar to the problems you were seing in your other example earlier...

G

[z11]

Hi Garth:

my understanding is that statement is true for any bar other than the current bar (in real time) if you do not have setcomputeonclose on. I believe this prevents logic to be executed while the EFS is being loaded so that only real time bar is being excuted on.

[z11]

hi Steve,

I did a search under your user id but did not see any relevant code addressing the issue in this context. Would you give me a hint where I could find those EFS postings you mentioned?

Thanks.

[z11]

On a separate note, I thought:

var vHigh=getValue("High",0,-10,"ES Z4,15");

would put last 10 highs of eMini's 15 minute bars? This EFS resides in a Adv Chart with a 5 minute time frame.

What am I doing wrong?

[gspiker]

Z11,

It does not have an overall condition

Sorry, I misread this to say "it does", not "it does not". So the point I was going to make doesn't apply to this EFS...

G

[z11]

ok, on that getValue("High"....) question...
I am getting the correct raw values back but I have to synchronize my calculation ... they are out of whack right now.