Source code for pandas_ta.trend.increasing

# -*- coding: utf-8 -*-
from pandas_ta.utils import get_drift, get_offset, is_percent, verify_series

[docs]def increasing(close, length=None, strict=None, asint=None, percent=None, drift=None, offset=None, **kwargs): """Indicator: Increasing""" # Validate Arguments length = int(length) if length and length > 0 else 1 strict = strict if isinstance(strict, bool) else False asint = asint if isinstance(asint, bool) else True close = verify_series(close, length) drift = get_drift(drift) offset = get_offset(offset) percent = float(percent) if is_percent(percent) else False if close is None: return # Calculate Result close_ = (1 + 0.01 * percent) * close if percent else close if strict: # Returns value as float64? Have to cast to bool increasing = close > close_.shift(drift) for x in range(3, length + 1): increasing = increasing & (close.shift(x - (drift + 1)) > close_.shift(x - drift)) increasing.fillna(0, inplace=True) increasing = increasing.astype(bool) else: increasing = close_.diff(length) > 0 if asint: increasing = increasing.astype(int) # Offset if offset != 0: increasing = increasing.shift(offset) # Handle fills if "fillna" in kwargs: increasing.fillna(kwargs["fillna"], inplace=True) if "fill_method" in kwargs: increasing.fillna(method=kwargs["fill_method"], inplace=True) # Name and Categorize it _percent = f"_{0.01 * percent}" if percent else '' _props = f"{'S' if strict else ''}INC{'p' if percent else ''}" increasing.name = f"{_props}_{length}{_percent}" increasing.category = "trend" return increasing
increasing.__doc__ = \ """Increasing Returns True if the series is increasing over a period, False otherwise. If the kwarg 'strict' is True, it returns True if it is continuously increasing over the period. When using the kwarg 'asint', then it returns 1 for True or 0 for False. Calculation: if strict: increasing = all(i < j for i, j in zip(close[-length:], close[1:])) else: increasing = close.diff(length) > 0 if asint: increasing = increasing.astype(int) Args: close (pd.Series): Series of 'close's length (int): It's period. Default: 1 strict (bool): If True, checks if the series is continuously increasing over the period. Default: False percent (float): Percent as an integer. Default: None asint (bool): Returns as binary. Default: True drift (int): The difference period. Default: 1 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 Returns: pd.Series: New feature generated. """