def ta_candles_as_culb(df: _pd.DataFrame, open="Open", high="High", low="Low", close="Close", volume="Volume", relative_close=False, drop_nan_volume=True):
o = _get_pandas_object(df, open)
c = _get_pandas_object(df, close)
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
oc = _pd.concat([o, c], axis=1)
c_1 = c.shift(1)
if not c.empty:
# calculate close, upper_shadow, lower_shadow, body
res = _pd.DataFrame({
"close": (c / c_1 - 1) if relative_close else c,
"upper": (h / oc.max(axis=1) - 1),
"lower": (oc.min(axis=1) / l - 1),
"body": (c / o - 1)
}, index=df.index)
else:
_log.warning("empty DataFrame!")
res = _pd.DataFrame({
"close": [],
"upper": [],
"lower": [],
"body": []
})
if volume is not None:
rel_vol = df[volume].pct_change()
if not drop_nan_volume or not rel_vol.isnull().all():
res[volume] = rel_vol
return res
def ta_cross(df: _pd.DataFrame, a=None, b=None, period=1):
# get pandas objects for crossing
if a is None and b is None:
assert len(df.columns) == 2, f"ambiguous crossing of {df.columns}"
a = df[df.columns[0]]
b = df[df.columns[1]]
elif b is None:
b = _get_pandas_object(df, a)
a = df
elif a is None:
b = _get_pandas_object(df, b)
a = df
else:
a = _get_pandas_object(df, a)
b = _get_pandas_object(df, b)
# get periods
a1 = a.shift(period)
b1 = b.shift(period)
# if a1 < b1 and a > b then a crosses over b
a_over_b = ((a1 < b1) & (a > b)).astype(int)
# if a1 > b1 and a < b then a cross under b
a_under_b = ((a1 > b1) & (a < b)).astype(int) * -1
return a_over_b + a_under_b
def ta_ultimate_osc(df: _pd.DataFrame,
period1=7,
period2=14,
period3=28,
close="Close",
high="High",
low="Low") -> _pd.Series:
# BP = Close - Minimum(Low or Prior Close).
# TR = Maximum(High or Prior Close) - Minimum(Low or Prior Close)
prev_close = _get_pandas_object(df, close).shift(1)
downs = (_get_pandas_object(
df,
low if low is not None else close).to_frame().join(prev_close)).min(
axis=1)
ups = (_get_pandas_object(
df,
high if high is not None else close).to_frame().join(prev_close)).max(
axis=1)
temp = _pd.DataFrame(
{
"bp": _get_pandas_object(df, close) - downs,
"tr": ups - downs
},
index=df.index)
periods = [period1, period2, period3]
avs = []
for period in periods:
# Average7 = (7 - period BP Sum) / (7 - period TR Sum)
av = temp.rolling(period).sum()
avs.append(av["bp"] / av["tr"])
# UO = [(4 x Average7) + (2 x Average14) + Average28] / (4 + 2 + 1)
return ((4 * avs[0] + 2 * avs[1] + avs[2]) /
7).rename(f"ultimate_osc_{period1},{period2},{period3}")
def ta_candles_as_culb(df: _pd.DataFrame,
open="Open",
high="High",
low="Low",
close="Close",
volume="Volume",
relative_close=False):
o = _get_pandas_object(df, open)
c = _get_pandas_object(df, close)
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
oc = _pd.concat([o, c], axis=1)
c_1 = c.shift(1)
# calculate close, upper_shadow, lower_shadow, body
res = _pd.DataFrame(
{
"close": (c / c_1 - 1) if relative_close else c,
"upper": (h / oc.max(axis=1) - 1),
"lower": (oc.min(axis=1) / l - 1),
"body": (c / o - 1)
},
index=df.index)
if volume is not None:
res[volume] = df[volume].pct_change()
return res
def ta_bop(df: _pd.DataFrame, open="Open", high="High", low="Low", close="Close") -> _PANDAS:
# (CLOSE – OPEN) / (HIGH – LOW)
o = _get_pandas_object(df, open)
c = _get_pandas_object(df, close)
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
return ((c - o) / (h - l)).rename("bop")
def ta_cci(df: _pd.DataFrame, period=14, high="High", low="Low", close="Close", alpha=0.015) -> _PANDAS:
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
c = _get_pandas_object(df, close)
tp = (h + l + c) / 3
tp_sma = _f.ta_sma(tp, period)
md = tp.rolling(period).apply(lambda x: _np.abs(x - x.mean()).sum() / period)
return ((1 / alpha) * (tp - tp_sma) / md / 100).rename(f"cci_{period}")
def ta_tr(df: _PANDAS, high="High", low="Low", close="Close", relative=True) -> _PANDAS:
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
c = _get_pandas_object(df, close).shift(1)
if relative:
ranges = (h / l - 1).rename("a").to_frame() \
.join((h / c - 1).rename("b").abs()) \
.join((l / c - 1).rename("c").abs())
else:
ranges = (h - l).rename("a").to_frame()\
.join((h - c).rename("b").abs())\
.join((l - c).rename("c").abs())
return ranges.max(axis=1).rename("true_range")
def ta_williams_R(df: _pd.DataFrame,
period=14,
close="Close",
high="High",
low="Low") -> _pd.Series:
temp = _get_pandas_object(df, close).to_frame()
temp = temp.join(
_get_pandas_object(df, high if high is not None else
close).rolling(period).max().rename("highest_high"))
temp = temp.join(
_get_pandas_object(df, low if low is not None else
close).rolling(period).min().rename("lowest_low"))
return ((temp["highest_high"] - temp[close]) /
(temp["highest_high"] -
temp["lowest_low"])).rename(f"williams_R_{period}")
def ta_adx(df: _PANDAS, period=14, high="High", low="Low", close="Close", relative=True) -> _PANDAS:
from ..utils import difference
h = _get_pandas_object(df, high)
l = _get_pandas_object(df, low)
temp = _pd.DataFrame({
"up": difference(h, h.shift(1), relative, 0),
"down": difference(l.shift(1), l, relative, 0)
}, index=df.index)
atr = ta_atr(df, period, high, low, close, relative=relative)
pdm = _f.ta_wilders(temp.apply(lambda r: r[0] if r["up"] > r["down"] and r["up"] > 0 else 0, raw=False, axis=1), period)
ndm = _f.ta_wilders(temp.apply(lambda r: r[1] if r["down"] > r["up"] and r["down"] > 0 else 0, raw=False, axis=1), period)
pdi = pdm / atr
ndi = ndm / atr
adx = _f.ta_wilders((pdi - ndi).abs() / (pdi + ndi).abs(), period)
return _pd.DataFrame({"+DM": pdm, "-DM": ndm, "+DI": pdi, "-DI": ndi, "ADX": adx}, index=df.index)
