def _agged(self, times, data=None):
"""
Get the aggregated data for the given times
data
intraday dataframe. If None, self.datas[0] is picked
times
list of 2-tuples with start and end times
Note
-----
1) For each time tuple given in list, data is aggregated
in OHLC form and added to a dataframe
2) Columns are numbered based on the order of the times
list so that the first tuple is named 0, the next 1 and so on
3) Expect a date and timestamp column in the list
"""
from fastbt.utils import recursive_merge
agged = {'open': 'first', 'high': 'max', 'low': 'min', 'close': 'last'}
dfs = []
if not (data):
data = self.datas[0]
for i, (s, e) in enumerate(times):
temp = data.set_index('timestamp').between_time(
s, e).groupby('date').agg(agged)
columns = {
x: '{x}{i}'.format(x=x, i=i)
for x in ['open', 'high', 'low', 'close']
}
temp.rename(columns=columns, inplace=True)
dfs.append(temp)
return recursive_merge(dfs, on=['date']).reset_index()
def corr(self, names=[], column='pnl'):
"""
Create a correlation matrix
names
names are the names of data sources to be merged
by default, all data sources are used
column
column name to merge
"""
keys = self._sources.keys()
if not(names):
names = keys
# Rename columns for better reporting
collect = []
for name in names:
src = self._sources.get(name)
if src is not None:
cols = ['date', column]
tmp = src[cols].rename(columns={column:name})
collect.append(tmp)
if len(collect) > 0:
frame = recursive_merge(collect, on=['date'], how='outer').fillna(0)
return frame.corr()
else:
return []
def _agged(self, times, data=None):
"""
Get the aggregated data for the given times
data
intraday dataframe. If None, self.datas[0] is picked
times
list of 2-tuples with start and end times
Note
-----
1) For each time tuple given in list, data is aggregated
in OHLC form and added to a dataframe
2) Columns are numbered based on the order of the times
list so that the first tuple is named 0, the next 1 and so on
3) Expect a date and timestamp column in the list
"""
from fastbt.utils import recursive_merge
agged = {"open": "first", "high": "max", "low": "min", "close": "last"}
dfs = []
if not (data):
data = self.datas[0]
for i, (s, e) in enumerate(times):
temp = (
data.set_index("timestamp")
.between_time(s, e)
.groupby("date")
.agg(agged)
)
columns = {
x: "{x}{i}".format(x=x, i=i) for x in ["open", "high", "low", "close"]
}
temp.rename(columns=columns, inplace=True)
dfs.append(temp)
return recursive_merge(dfs, on=["date"]).reset_index()
def get_column(self, column='pnl', on='date', how='outer'):
"""
Get a single column from all the dataframes and merge
them into a single dataframe
"""
names = self._sources.keys()
# Rename columns for better reporting
collect = []
for name in names:
src = self._sources.get(name)
if src is not None:
cols = ['date', column]
tmp = src[cols].rename(columns={column:name})
collect.append(tmp)
if len(collect) > 0:
frame = recursive_merge(collect, on=['date'], how='outer').fillna(0)
return frame