def _align_bivariate(self, df1, df2, force_align=False):
if isinstance(df1, pd.DataFrame) and isinstance(df2, pd.DataFrame):
len1 = len(df1.index)
len2 = len(df2.index)
if (self.ann_dts is not None) and (self.trade_dts is not None):
if len1 > len2:
df2 = align(df2, self.ann_dts, self.trade_dts)
elif len1 < len2:
df1 = align(df1, self.ann_dts, self.trade_dts)
elif force_align:
df1 = align(df1, self.ann_dts, self.trade_dts)
df2 = align(df2, self.ann_dts, self.trade_dts)
return (df1, df2)
def test_remote_data_service_industry():
from quantos.data.align import align
import pandas as pd
ds = RemoteDataService()
arr = ds.get_index_comp(index='000300.SH', start_date=20130101, end_date=20170505)
df = ds.get_industry_raw(symbol=','.join(arr), type_='ZZ')
df = df.astype(dtype={'in_date': int})
# df_ann = df.loc[:, ['in_date', 'symbol']]
# df_ann = df_ann.set_index(['symbol', 'in_date'])
# df_ann = df_ann.unstack(level='symbol')
from quantos.data.dataview import DataView
dic_sec = DataView._group_df_to_dict(df, by='symbol')
dic_sec = {sec: df.reset_index() for sec, df in dic_sec.viewitems()}
df_ann = pd.concat([df.loc[:, 'in_date'].rename(sec) for sec, df in dic_sec.viewitems()], axis=1)
df_value = pd.concat([df.loc[:, 'industry1_code'].rename(sec) for sec, df in dic_sec.viewitems()], axis=1)
dates_arr = ds.get_trade_date(20140101, 20170505)
res = align(df_value, df_ann, dates_arr)
print
# df_ann = df.pivot(index='in_date', columns='symbol', values='in_date')
# df_value = df.pivot(index=None, columns='symbol', values='industry1_code')
def align_single_df(df_one_sec):
df_value = df_one_sec.loc[:, ['industry1_code']]
df_ann = df_one_sec.loc[:, ['in_date']]
res = align(df_value, df_ann, dates_arr)
return res
# res_list = [align_single_df(df) for sec, df in dic_sec.viewitems()]
res_list = [align_single_df(df) for sec, df in dic_sec.items()[:10]]
res = pd.concat(res_list, axis=1)
print res
def _align_univariate(self, df1):
if isinstance(df1, pd.DataFrame):
if (self.ann_dts is not None) and (self.trade_dts is not None):
len1 = len(df1.index)
len2 = len(self.trade_dts)
if len1 != len2:
return align(df1, self.ann_dts, self.trade_dts)
return df1
def get_industry_daily(self, symbol, start_date, end_date, type_='SW'):
"""
Get index components on each day during start_date and end_date.
Parameters
----------
symbol : str
separated by ','
start_date : int
end_date : int
type_ : {'SW', 'ZZ'}
Returns
-------
res : pd.DataFrame
index dates, columns symbols
values are industry code
"""
df_raw = self.get_industry_raw(symbol, type_=type_)
dic_sec = self._group_df_to_dict(df_raw, by='symbol')
dic_sec = {
sec: df.sort_values(by='in_date', axis=0).reset_index()
for sec, df in dic_sec.viewitems()
}
df_ann = pd.concat([
df.loc[:, 'in_date'].rename(sec)
for sec, df in dic_sec.viewitems()
],
axis=1)
df_value = pd.concat([
df.loc[:, 'industry1_code'].rename(sec)
for sec, df in dic_sec.viewitems()
],
axis=1)
dates_arr = self.get_trade_date(start_date, end_date)
df_industry = align.align(df_value, df_ann, dates_arr)
# TODO before industry classification is available, we assume they belong to their first group.
df_industry = df_industry.fillna(method='bfill')
df_industry = df_industry.astype(str)
return df_industry
def align_single_df(df_one_sec):
df_value = df_one_sec.loc[:, ['industry1_code']]
df_ann = df_one_sec.loc[:, ['in_date']]
res = align(df_value, df_ann, dates_arr)
return res