Why we need it?
A lot of us migrate to Ducascopy from Metatrader. We studied mql well, but have around zero exp in java
mql converter works, but it's ugly and noway to get converted code
Custom indicator subforum closed, but there no way to get help in JStore. Its store and no more.
VisualJforex has no ability to develope indicators
WiKi pages is helpless
BUT we NEED to build new custom indicators.
I propose to ask here question about custom indicators build only.

Some questions
In API some indicators use

    private double[][] inputs = new double[1][];
    new InputParameterInfo("Price", InputParameterInfo.Type.DOUBLE)

another one

    private double[][][] inputs = new double[1][][];
    new InputParameterInfo("Price", InputParameterInfo.Type.PRICE)

What is better and why?

also the same situation with main cycle

        int i, k;
        for (i = 1, k = dtenkanMaxResult.getNumberOfElements(); i < k; i++) {

and

       int i, j;
        for (i = startIndex, j = 0; i <= endIndex; i++, j++) {                            

Which way is better?

1. JForex Wiki:
Currently input can be either DOUBLE, which means array of prices of one kind, or PRICE in which case system will give several arrays of prices - double[5][n] and BAR for array of IBar objects. Prices for PRICE type will be in the following order: open, close, high, low, volume.
When indicator needs for calculations several kinds of prices (open, close, high, low) or volume then one should use PRICE type of input.
For BAR type array of IBar objects also contains information about all kinds of prices, volume and time of the bar as well. This approach is more object oriented.
PRICE type is used when indicator needs only one kind of prices.
2. When indicator uses result of another indicator calculation one should use cycle with IndicatorResult.getNumberOfElements() number of iterations.
For iterating over elements of input arrays one should use startIndex and endIndex parameters as array index bounds.