Output index differs from input index (stoch, td_seq)

Hello @twopirllc,

upon releasing vectorbt v0.20 I noticed some pandas-ta tests failing. As you know, I previously created a simple parser to extract the meta of each pandas-ta indicator (it builds a dummy DataFrame with 50 random floating elements and passes it to pandas-ta to see what outputs to expect). The parser started to fail for some of the indicators:

>>> import vectorbt as vbt

>>> vbt.IndicatorFactory.get_pandas_ta_indicators(silence_warnings=False)
Function stoch: Couldn't parse the output: mismatching index
Function td_seq: Couldn't parse the output: mismatching index
Function trix: Couldn't parse the indicator: inputs are all NaN
Function ichimoku: 'NoneType' object has no attribute 'index'
UserWarning: Function kvo: 'NoneType' object has no attribute 'index'

I can’t remember which ones I wasn’t able to parse previously (probably Ichimoku), but the new ones are definitely stoch and td_seq. Upon looking closer, I noticed that stoch removes head entries that are NaN, while td_seq returns a simple range index instead of a datetime-like index.

To address this, I can quickly patch vectorbt to 1) fill the missing index if the output index is of the same type and 2) copy the input index if the output index has another type but the same length.

Would love to hear your thoughts on this matter.

Edit: I’m running 0.3.14b0

[polakowo]

@twopirllc,

perfect, I’ll look into the other two and maybe tweak the parser a bit. Thank you!

[twopirllc]

@bennycode,

Why does it need 9 inputs of high, low, close before the Stochastic Oscillator returns a result?

It has the required inputs to calculate ta.stoch as described in help(ta.stoch), which is closer to 7. Some (offset, fillna, fill_method) you won’t use unless you actually need them.

Is ta.stoch not working as expected?

Stochastic (STOCH)

    Args:
        high (pd.Series): Series of 'high's
        low (pd.Series): Series of 'low's
        close (pd.Series): Series of 'close's
        k (int): The Fast %K period. Default: 14
        d (int): The Slow %K period. Default: 3
        smooth_k (int): The Slow %D period. Default: 3
        mamode (str): See ```help(ta.ma)```. Default: 'sma'
        offset (int): How many periods to offset the result. Default: 0

    Kwargs:
        fillna (value, optional): pd.DataFrame.fillna(value)
        fill_method (value, optional): Type of fill method

However, I should probably add another mamode like TA-Lib does:

/*
 * TA_STOCH - Stochastic
 * 
 * Input  = High, Low, Close
 * Output = double, double
 * 
 * Optional Parameters
 * -------------------
 * optInFastK_Period:(From 1 to 100000)
 *    Time period for building the Fast-K line
 * 
 * optInSlowK_Period:(From 1 to 100000)
 *    Smoothing for making the Slow-K line. Usually set to 3
 * 
 * optInSlowK_MAType:
 *    Type of Moving Average for Slow-K
 * 
 * optInSlowD_Period:(From 1 to 100000)
 *    Smoothing for making the Slow-D line
 * 
 * optInSlowD_MAType:
 *    Type of Moving Average for Slow-D
 * 
 */

Kind Regards, KJ