Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

What is the structure of self.data ? #526

Answered by kernc
wolfcuring asked this question in Q&A
Discussion options

I defined a func to prepare for the indicator which will use both open, high, low, close columns as input. As my experience before, the package will first turn all pandas DataFrames into np.arrays. So I add a line to turn it back into pandas, so that I can use pd funcs to prepare the data. But it is throwing an error which indicates that DataFrame constructor failed.

`def slump(df):

df = pd.DataFrame(df)
big = ((df.High - df.Low) > 10).astype(int) 
up = ((df.Open - df.Close) < 0).astype(int)
return big*up

...

class Big10(Strategy):

n = 52 
def init(self):
 self.big = self.I(slump, self.data, plot=True) `

and the error is:

Indicator "slump" errored with exception: DataFrame constructor not properly called!

I guess the problem lies in my calling self.data , as the input. As I see in the tutorials, self.data.Close works fine. So, I guess self.data has all the 5 columns ?
You must be logged in to vote

The docs explicitly say Strategy.data is not a dataframe, so you can't just use it as if.
If you need data to become a DataFrame, you can use its .df accessor (self.data.df), or .s accessor for individual series (self.data.Close.s).

Finally, when in doubt, just look in the source:

class _Data:
"""
A data array accessor. Provides access to OHLCV "columns"
as a standard `pd.DataFrame` would, except it's not a DataFrame
and the returned "series" are _not_ `pd.Series` but `np.ndarray`
for performance reasons.
"""
def __init__(self, df: pd.DataFrame):
self.__df = df
self....

Replies: 1 comment 1 reply

Comment options

The docs explicitly say Strategy.data is not a dataframe, so you can't just use it as if.
If you need data to become a DataFrame, you can use its .df accessor (self.data.df), or .s accessor for individual series (self.data.Close.s).

Finally, when in doubt, just look in the source:

class _Data:
"""
A data array accessor. Provides access to OHLCV "columns"
as a standard `pd.DataFrame` would, except it's not a DataFrame
and the returned "series" are _not_ `pd.Series` but `np.ndarray`
for performance reasons.
"""
def __init__(self, df: pd.DataFrame):
self.__df = df
self.__i = len(df)
self.__pip: Optional[float] = None
self.__cache: Dict[str, _Array] = {}
self.__arrays: Dict[str, _Array] = {}
self._update()
def __getitem__(self, item):
return self.__get_array(item)
def __getattr__(self, item):
try:
return self.__get_array(item)
except KeyError:
raise AttributeError(f"Column '{item}' not in data") from None
def _set_length(self, i):
self.__i = i
self.__cache.clear()
def _update(self):
index = self.__df.index.copy()
self.__arrays = {col: _Array(arr, index=index)
for col, arr in self.__df.items()}
# Leave index as Series because pd.Timestamp nicer API to work with
self.__arrays['__index'] = index
def __repr__(self):
i = min(self.__i, len(self.__df) - 1)
index = self.__arrays['__index'][i]
items = ', '.join(f'{k}={v}' for k, v in self.__df.iloc[i].items())
return f'<Data i={i} ({index}) {items}>'
def __len__(self):
return self.__i
@property
def df(self) -> pd.DataFrame:
return (self.__df.iloc[:self.__i]
if self.__i < len(self.__df)
else self.__df)
@property
def pip(self) -> float:
if self.__pip is None:
self.__pip = 10**-np.median([len(s.partition('.')[-1])
for s in self.__arrays['Close'].astype(str)])
return self.__pip
def __get_array(self, key) -> _Array:
arr = self.__cache.get(key)
if arr is None:
arr = self.__cache[key] = self.__arrays[key][:self.__i]
return arr
@property
def Open(self) -> _Array:
return self.__get_array('Open')
@property
def High(self) -> _Array:
return self.__get_array('High')
@property
def Low(self) -> _Array:
return self.__get_array('Low')
@property
def Close(self) -> _Array:
return self.__get_array('Close')
@property
def Volume(self) -> _Array:
return self.__get_array('Volume')
@property
def index(self) -> pd.DatetimeIndex:
return self.__get_array('__index')
# Make pickling in Backtest.optimize() work with our catch-all __getattr__
def __getstate__(self):
return self.__dict__
def __setstate__(self, state):
self.__dict__ = state

😁
You must be logged in to vote
1 reply
Comment options

Cool ! It works. So I guess, it works, if I prepare all the conditions into one column and pass it onto OHLC data. Then, access it in a way, for example, when self.data.Signal == 1 buy, and when self.data.Signal == -1 close position ?
Answer selected by wolfcuring
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
None yet
2 participants

AltStyle によって変換されたページ (->オリジナル) /