Rather then computing iLast (as I did above,) I could "back up" to the last non-zero TimeFrame bar for the stair step effect. Note the dropping the minus one from iBar calculation and changing the for to a while loop, simplify things.
The lookback limit works very well for me in MT5 to limit the number of bars drawn by the custom indicator and save resources. However, when a new chart is downloading new data, I can sometimes witness garbled data appearing past bar 1000, which was never filled in by the custom indicator. I know that the indicator didn't fill in those values, as it never looked past 1000 bars, and it cancels if it cannot access any price data yet (downloading a new chart). Where does the garbled data come from? Of course, I found an easy way to fix this, which was to fill all 50,000 bars on the chart with EMPTY_VALUE. That solves the problem. The fix looks like this:
if (prev_calculated==0) for (j=0; j<rates_total; j++) indicatorBuffer[j]=EMPTY_VALUE;
This works, and is fast because it only has to loop once on the first tick, or only a few times when new data is being downloaded to a new chart. However, it looks a bit messy and has to loop 50,000 bars, or even 500,000 if that many are on the chart. Is there a nicer, cleaner solution for this?
Victor Epand: I know that the indicator didn't fill in those values, as it never looked past 1000 bars,
>Yes it did. It filled them when there were less than 1K bars on chart. Then newer history came in. Now there are 2K bars with values in the older part.
Always set buffer values, do not assume that they are empty values.
I never liked those arbitrary look back limits. I use the attached file to process as many bars as possible in 1.5 seconds and then continue periodically(either by time or by tick,) until all bars are done.
const int LB_MA = LengthMMA-1;
int iMA = rates_total - 1 - MathMax(LB_MA, prev_calculated);
int iLast = MathMax(LAST, iMA - 2000); // Do in groups
for(; iMA >= iLast; --iMA){
:
} // iMA
//----
#define REDRAW_BAR_ZERO true // Redraw Bar zero.
return rates_total - MathMax(REDRAW_BAR_ZERO, iLast);
extern ENUM_TIMEFRAMES TimeFrame = PERIOD_CURRENT;
//+------------------------------------------------------------------+
//| Custom indicator initialization function |
//+------------------------------------------------------------------+
int OnInit(){
if(TimeFrame == PERIOD_CURRENT) TimeFrame = (ENUM_TIMEFRAMES) _Period;
if(TimeFrame < _Period) return INIT_PARAMETERS_INCORRECT;
int iBar = Rates_total - 1 - MathMax(lookback, prev_calculated);
#define LAST 0
int iLast = MathMax(LAST, iBar - 2000); // Do in groups
for(; iBar >= iLast; --iBar){
int iTF = iBarShift(NULL, TimeFrame, Time[iBar]);
double hiCur = iHigh(NULL, TimeFrame, iTF), //High[iBar],
hiPre = iHigh(NULL, TimeFrame, iTF+1); //High[iBar+1];
:
(See images Multi TimeFrame Moving Average or Moving Average Ex from the codebase)
iLast = iBarShift(NULL, 0, Time[iLast] - Time[iLast] % (TimeFrame * 60));
return rates_total - REDRAW_BAR_ZERO - iLast;
The lookback limit works very well for me in MT5 to limit the number of bars drawn by the custom indicator and save resources. However, when a new chart is downloading new data, I can sometimes witness garbled data appearing past bar 1000, which was never filled in by the custom indicator. I know that the indicator didn't fill in those values, as it never looked past 1000 bars, and it cancels if it cannot access any price data yet (downloading a new chart). Where does the garbled data come from? Of course, I found an easy way to fix this, which was to fill all 50,000 bars on the chart with EMPTY_VALUE. That solves the problem. The fix looks like this:
if (prev_calculated==0) for (j=0; j<rates_total; j++) indicatorBuffer[j]=EMPTY_VALUE;
This works, and is fast because it only has to loop once on the first tick, or only a few times when new data is being downloaded to a new chart. However, it looks a bit messy and has to loop 50,000 bars, or even 500,000 if that many are on the chart. Is there a nicer, cleaner solution for this?
Always set buffer values, do not assume that they are empty values.