mirror of
https://github.com/freqtrade/freqtrade.git
synced 2024-11-10 10:21:59 +00:00
182 lines
6.5 KiB
Python
182 lines
6.5 KiB
Python
import json
|
|
import logging
|
|
from datetime import datetime
|
|
from typing import Any, ClassVar, List, Optional, Sequence
|
|
|
|
from sqlalchemy import DateTime, ForeignKey, Integer, String, Text, UniqueConstraint, select
|
|
from sqlalchemy.orm import Mapped, mapped_column, relationship
|
|
|
|
from freqtrade.constants import DATETIME_PRINT_FORMAT
|
|
from freqtrade.persistence.base import ModelBase, SessionType
|
|
from freqtrade.util import dt_now
|
|
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class _CustomData(ModelBase):
|
|
"""
|
|
CustomData database model
|
|
Keeps records of metadata as key/value store
|
|
for trades or global persistent values
|
|
One to many relationship with Trades:
|
|
- One trade can have many metadata entries
|
|
- One metadata entry can only be associated with one Trade
|
|
"""
|
|
|
|
__tablename__ = "trade_custom_data"
|
|
__allow_unmapped__ = True
|
|
session: ClassVar[SessionType]
|
|
|
|
# Uniqueness should be ensured over pair, order_id
|
|
# its likely that order_id is unique per Pair on some exchanges.
|
|
__table_args__ = (UniqueConstraint("ft_trade_id", "cd_key", name="_trade_id_cd_key"),)
|
|
|
|
id = mapped_column(Integer, primary_key=True)
|
|
ft_trade_id = mapped_column(Integer, ForeignKey("trades.id"), index=True)
|
|
|
|
trade = relationship("Trade", back_populates="custom_data")
|
|
|
|
cd_key: Mapped[str] = mapped_column(String(255), nullable=False)
|
|
cd_type: Mapped[str] = mapped_column(String(25), nullable=False)
|
|
cd_value: Mapped[str] = mapped_column(Text, nullable=False)
|
|
created_at: Mapped[datetime] = mapped_column(DateTime, nullable=False, default=dt_now)
|
|
updated_at: Mapped[Optional[datetime]] = mapped_column(DateTime, nullable=True)
|
|
|
|
# Empty container value - not persisted, but filled with cd_value on query
|
|
value: Any = None
|
|
|
|
def __repr__(self):
|
|
create_time = (
|
|
self.created_at.strftime(DATETIME_PRINT_FORMAT) if self.created_at is not None else None
|
|
)
|
|
update_time = (
|
|
self.updated_at.strftime(DATETIME_PRINT_FORMAT) if self.updated_at is not None else None
|
|
)
|
|
return (
|
|
f"CustomData(id={self.id}, key={self.cd_key}, type={self.cd_type}, "
|
|
+ f"value={self.cd_value}, trade_id={self.ft_trade_id}, created={create_time}, "
|
|
+ f"updated={update_time})"
|
|
)
|
|
|
|
@classmethod
|
|
def query_cd(
|
|
cls, key: Optional[str] = None, trade_id: Optional[int] = None
|
|
) -> Sequence["_CustomData"]:
|
|
"""
|
|
Get all CustomData, if trade_id is not specified
|
|
return will be for generic values not tied to a trade
|
|
:param trade_id: id of the Trade
|
|
"""
|
|
filters = []
|
|
if trade_id is not None:
|
|
filters.append(_CustomData.ft_trade_id == trade_id)
|
|
if key is not None:
|
|
filters.append(_CustomData.cd_key.ilike(key))
|
|
|
|
return _CustomData.session.scalars(select(_CustomData).filter(*filters)).all()
|
|
|
|
|
|
class CustomDataWrapper:
|
|
"""
|
|
CustomData middleware class
|
|
Abstracts the database layer away so it becomes optional - which will be necessary to support
|
|
backtesting and hyperopt in the future.
|
|
"""
|
|
|
|
use_db = True
|
|
custom_data: List[_CustomData] = []
|
|
unserialized_types = ["bool", "float", "int", "str"]
|
|
|
|
@staticmethod
|
|
def _convert_custom_data(data: _CustomData) -> _CustomData:
|
|
if data.cd_type in CustomDataWrapper.unserialized_types:
|
|
data.value = data.cd_value
|
|
if data.cd_type == "bool":
|
|
data.value = data.cd_value.lower() == "true"
|
|
elif data.cd_type == "int":
|
|
data.value = int(data.cd_value)
|
|
elif data.cd_type == "float":
|
|
data.value = float(data.cd_value)
|
|
else:
|
|
data.value = json.loads(data.cd_value)
|
|
return data
|
|
|
|
@staticmethod
|
|
def reset_custom_data() -> None:
|
|
"""
|
|
Resets all key-value pairs. Only active for backtesting mode.
|
|
"""
|
|
if not CustomDataWrapper.use_db:
|
|
CustomDataWrapper.custom_data = []
|
|
|
|
@staticmethod
|
|
def delete_custom_data(trade_id: int) -> None:
|
|
_CustomData.session.query(_CustomData).filter(_CustomData.ft_trade_id == trade_id).delete()
|
|
_CustomData.session.commit()
|
|
|
|
@staticmethod
|
|
def get_custom_data(*, trade_id: int, key: Optional[str] = None) -> List[_CustomData]:
|
|
if CustomDataWrapper.use_db:
|
|
filters = [
|
|
_CustomData.ft_trade_id == trade_id,
|
|
]
|
|
if key is not None:
|
|
filters.append(_CustomData.cd_key.ilike(key))
|
|
filtered_custom_data = _CustomData.session.scalars(
|
|
select(_CustomData).filter(*filters)
|
|
).all()
|
|
|
|
else:
|
|
filtered_custom_data = [
|
|
data_entry
|
|
for data_entry in CustomDataWrapper.custom_data
|
|
if (data_entry.ft_trade_id == trade_id)
|
|
]
|
|
if key is not None:
|
|
filtered_custom_data = [
|
|
data_entry
|
|
for data_entry in filtered_custom_data
|
|
if (data_entry.cd_key.casefold() == key.casefold())
|
|
]
|
|
return [CustomDataWrapper._convert_custom_data(d) for d in filtered_custom_data]
|
|
|
|
@staticmethod
|
|
def set_custom_data(trade_id: int, key: str, value: Any) -> None:
|
|
value_type = type(value).__name__
|
|
|
|
if value_type not in CustomDataWrapper.unserialized_types:
|
|
try:
|
|
value_db = json.dumps(value)
|
|
except TypeError as e:
|
|
logger.warning(f"could not serialize {key} value due to {e}")
|
|
return
|
|
else:
|
|
value_db = str(value)
|
|
|
|
if trade_id is None:
|
|
trade_id = 0
|
|
|
|
custom_data = CustomDataWrapper.get_custom_data(trade_id=trade_id, key=key)
|
|
if custom_data:
|
|
data_entry = custom_data[0]
|
|
data_entry.cd_value = value_db
|
|
data_entry.updated_at = dt_now()
|
|
else:
|
|
data_entry = _CustomData(
|
|
ft_trade_id=trade_id,
|
|
cd_key=key,
|
|
cd_type=value_type,
|
|
cd_value=value_db,
|
|
created_at=dt_now(),
|
|
)
|
|
data_entry.value = value
|
|
|
|
if CustomDataWrapper.use_db and value_db is not None:
|
|
_CustomData.session.add(data_entry)
|
|
_CustomData.session.commit()
|
|
else:
|
|
if not custom_data:
|
|
CustomDataWrapper.custom_data.append(data_entry)
|
|
# Existing data will have updated interactively.
|