def fix_factor(self, stock_list: list, factor: f or list, start_date: str or datetime.date,
end_date: str or datetime.date):
# 格式化参数
stock_list = convert_11code(stock_list)
collection2factor_map = dis_collection2factor_map(factor, self.factor2collection_map)
start_date = convert2datetime(start_date)
end_date = convert2datetime(end_date)
end_date = end_date + datetime.timedelta(hours=23, minutes=59, seconds=59)
# 查询停牌时间
calendar = TradeCalendar().calendar(start_date.strftime("%Y%m%d"),
end_date.strftime("%Y%m%d"))
trading_calendar = calendar['date'][calendar['trade']]
trading_calendar_index = pandas.DataFrame(trading_calendar,
columns=['index']).set_index('index')
rett = trading_calendar_index.copy()
# 确定要查询的集合和字段值
collection, field = list(collection2factor_map.items())[0]
field = field[0].name
snap = ['SecuCode', 'PubDate', field]
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
# 查询
db_coll = connect_coll(collection, self._db)
ret = db_coll.find({'SecuCode': {'$in': stock_list},
"PubDate": {"$gte": start_date, "$lte": end_date}}, doc_snap)
# 生成查询结果
data = pandas.DataFrame(list(ret))
# 对查询结果进行规范化
if not data.empty:
data[snap[1]] = data[snap[1]].map(lambda x: x.strftime('%Y-%m-%d'))
data = data[snap]
data.columns = ['stock', 'time', field]
data = pandas.crosstab(data['time'], data['stock'], values=data[field], aggfunc='last')
# 更新 rett
rett = data.merge(rett, left_index=True, right_index=True, how='outer')
rett = rett.fillna(method='pad') # 先向后填充数据
rett = rett.ix[trading_calendar_index.index] # 再以日历限制一次日期
to_concat_ret = pandas.DataFrame(dict(zip(stock_list, [1] * len(stock_list))),index=['1'])
rett = pandas.concat([rett, to_concat_ret])
rett = rett.drop(['1'])
# 整理结果
rett = rett.astype(float)
rett = rett.to_records()
rett.dtype.names = ['date'] + list(rett.dtype.names)[1:]
# 暂时返回structured array，后面可以让用户选择返回pandas
# 固定了因子的表格，表的行索引是股票，列索引是时间
return numpy.array(rett)
def fix_time(self, stock_list: list or str, factors: list or f,
trade_date: datetime.date or str):
stock_list = convert_11code(stock_list)
collection2factor_map = dis_collection2factor_map(
factors, self.factor2collection_map)
start_date = datetime.datetime(2010, 8, 14)
end_date = start_date + datetime.timedelta(
hours=23, minutes=59, seconds=59)
dtypes = [("stock", "U11")]
for ft in factors:
dtypes.append((ft.name, "<f8"))
result = numpy.full((len(stock_list), ), numpy.NAN, dtype=dtypes)
result["stock"] = stock_list
indexes = {}
for collection in collection2factor_map:
factor_name_list = gen_factor_name_list(
collection2factor_map[collection])
snap = [
'SecuCode',
'PubDate',
]
snap.extend(factor_name_list)
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
db_coll = connect_coll(collection, self._db)
data = db_coll.find(
{
'SecuCode': {
'$in': stock_list
},
"PubDate": {
"$gte": start_date,
"$lte": end_date
}
}, doc_snap).sort("PubDate", pymongo.ASCENDING)
for d in data:
code = d['SecuCode']
idx = indexes.get(code)
if idx is None:
idx = find_date_in_array(code, result["stock"])
if idx == -1:
system_log.warning(
f"[Finance.fix_time] code index not found. record={d}"
)
continue
indexes[code] = idx
for ft in factor_name_list:
result[idx][ft] = d[ft]
return result
def fix_factor(self, stock_list: list, factor: f or list, time: str
or datetime.date, frequency: (1, 2, 3, 4)):
# 格式化参数
stock_list = convert_11code(stock_list)
collection2factor_map = dis_collection2factor_map(
factor, self.factor2collection_map)
collection, field = list(collection2factor_map.items())[0]
field = field[0].name
snap = ['SecuCode', 'PubDate', field]
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
# 测试时间点
start_date = datetime.datetime(2015, 1, 1)
end_date = datetime.datetime(2015, 3, 31)
# 查询
db_coll = connect_coll(collection, self._db)
ret = db_coll.find(
{
'SecuCode': {
'$in': stock_list
},
"PubDate": {
"$gte": start_date,
"$lte": end_date
}
}, doc_snap).sort('PubDate', pymongo.ASCENDING)
dtypes = [("date", "uint32")]
# 转换证券代码标识
for code in stock_list:
dtypes.append((code, "<f8"))
trade_days = TradeCalendar().calendar(start_date, end_date)
result = numpy.full((trade_days.shape[0], ), numpy.NaN, dtype=dtypes)
result["date"] = trade_days
indexes = {}
for r in ret:
t = yyyymmdd_date(r['PubDate'])
idx = indexes.get(t)
if not idx:
idx = find_date_in_array(t, result["date"])
if idx == -1: # 非交易日
system_log.warning(
f"[Finance.fix_factor] date index not found. record={r}"
)
continue
indexes[t] = idx
result[r.get('SecuCode')][idx] = r[field]
return result
def fix_symbol(self, stock: str, factors: list or f, start_date: datetime.date or str,
end_date: datetime.date or str):
# 格式化参数
stock = convert_11code(stock)[0]
collection2factor_map = dis_collection2factor_map(factors, self.factor2collection_map)
start_date = convert2datetime(start_date)
end_date = convert2datetime(end_date)
end_date = end_date + datetime.timedelta(hours=23, minutes=59, seconds=59)
# 查询停牌时间
calendar = TradeCalendar().calendar(start_date.strftime("%Y%m%d"),
end_date.strftime("%Y%m%d"))
trading_calendar = calendar['date'][calendar['trade']]
trading_calendar_index = pandas.DataFrame(trading_calendar,
columns=['index']).set_index('index')
rett = trading_calendar_index.copy()
# 确定要查询的集合和字段值
_year = int(start_date.strftime("%Y")) - 1
start_date = datetime.datetime(_year, 1, 1, 0, 0, 0)
for collection in collection2factor_map:
factor_name_list = gen_factor_name_list(collection2factor_map[collection])
snap = ['PubDate', ]
snap.extend(factor_name_list)
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
# 查询
db_coll = connect_coll(collection, self._db)
data = db_coll.find({'SecuCode': stock,
"PubDate": {"$gte": start_date, "$lte": end_date}}, doc_snap)
# 生成查询结果
data = pandas.DataFrame(list(data))
# 对查询结果进行规范化
if not data.empty:
data[snap[0]] = data[snap[0]].map(lambda x: x.strftime('%Y-%m-%d'))
data = data[snap]
data = data.set_index(snap[0])
# 循环更新 rett
rett = rett.merge(data, left_index=True, right_index=True, how='outer')
rett[factor_name_list] = rett[factor_name_list].fillna(method='pad') # 先向后填充数据
rett = rett.ix[trading_calendar_index.index] # 再以日历限制一次日期
# 整理结果
rett = rett.astype(float).to_records()
rett.dtype.names = ['date'] + list(rett.dtype.names)[1:]
return numpy.array(rett)
def fix_time(self, stock_list: list or str, factors: list or f,
trade_date: datetime.date or str):
# 格式化参数
stock_list = convert_11code(stock_list)
collection2factor_map = dis_collection2factor_map(factors, self.factor2collection_map)
start_date = convert2datetime(trade_date)
end_date = start_date + datetime.timedelta(hours=23, minutes=59, seconds=59)
# 补充缺失股票
rett = pandas.DataFrame(stock_list, columns=['stock']).set_index('stock')
# 确定要查询的集合和字段值
_year = int(start_date.strftime("%Y")) - 1
start_date = datetime.datetime(_year, 1, 1, 0, 0, 0)
for collection in collection2factor_map:
factor_name_list = gen_factor_name_list(collection2factor_map[collection])
snap = ['SecuCode', 'PubDate',]
snap.extend(factor_name_list)
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
# 查询
db_coll = connect_coll(collection, self._db)
data = db_coll.find({'SecuCode': {'$in': stock_list},
"PubDate": {"$gte": start_date, "$lte": end_date}}, doc_snap)
# 生成查询结果
data = pandas.DataFrame(list(data))
# 对查询结果进行规范化
if not data.empty:
data = data[snap].set_index('PubDate')
data.columns = ['stock'] + factor_name_list
if len(data):
data = data.groupby('stock')
data = pandas.DataFrame([i[1].iloc[-1] for i in data]).set_index('stock')
else:
data = data.set_index('stock')
# 循环更新 rett
rett = rett.merge(data, left_index=True, right_index=True, how='outer')
# 整理结果
rett = rett.astype(float).to_records()
rett.dtype.names = ['stock'] + list(rett.dtype.names)[1:]
return numpy.array(rett)
def fix_symbol(self, stock: str, factors: list or f, time: str
or datetime.date, frequency: (1, 2, 3, 4)):
stock = convert_11code(stock)[0]
print(stock)
collection2factor_map = dis_collection2factor_map(
factors, self.factor2collection_map)
# 测试时间点
start_date = datetime.datetime(2010, 1, 1)
end_date = datetime.datetime(2015, 3, 31)
calendar = TradeCalendar()
trade_days = calendar.calendar(start_date, end_date)
print("==>", trade_days)
dtypes = [("date", "uint32")]
for ft in factors:
dtypes.append((ft.name, "<f8"))
print("###", dtypes)
for collection in collection2factor_map:
factor_name_list = gen_factor_name_list(
collection2factor_map[collection])
print("-->", factor_name_list)
snap = [
'PubDate',
]
snap.extend(factor_name_list)
doc_snap = {k: 1 for k in snap}
doc_snap["_id"] = 0
db_coll = connect_coll(collection, self._db)
data = db_coll.find(
{
'SecuCode': stock,
"PubDate": {
"$gte": start_date,
"$lte": end_date
}
}, doc_snap).sort("PubDate", pymongo.ASCENDING)
# 初始化结果
result = numpy.full((trade_days.shape[0], ),
numpy.NAN,
dtype=dtypes)
result["date"] = trade_days
indexes = {}
for d in data:
t = yyyymmdd_date(d["PubDate"])
print(t)
idx = indexes.get(t)
print(idx, "---")
if not idx:
idx = find_date_in_array(t, result["date"])
print(idx, "===")
if idx == -1:
system_log.warning(
f"[Finance.fix_symbol] date index not found. record={d}"
)
continue
indexes[t] = idx
for ft in factor_name_list:
result[idx][ft] = d[ft]
return result