Merge pull request #2372 from xmatthias/kraken_ohlcv_emulate

download tick-based data to emulate candles
This commit is contained in:
hroff-1902
2019-10-19 19:32:37 +03:00
committed by GitHub
19 changed files with 799 additions and 15 deletions

View File

@@ -35,7 +35,8 @@ ARGS_LIST_TIMEFRAMES = ["exchange", "print_one_column"]
ARGS_CREATE_USERDIR = ["user_data_dir"]
ARGS_DOWNLOAD_DATA = ["pairs", "pairs_file", "days", "exchange", "timeframes", "erase"]
ARGS_DOWNLOAD_DATA = ["pairs", "pairs_file", "days", "download_trades", "exchange",
"timeframes", "erase"]
ARGS_PLOT_DATAFRAME = ["pairs", "indicators1", "indicators2", "plot_limit", "db_url",
"trade_source", "export", "exportfilename", "timerange", "ticker_interval"]

View File

@@ -273,6 +273,12 @@ AVAILABLE_CLI_OPTIONS = {
type=check_int_positive,
metavar='INT',
),
"download_trades": Arg(
'--dl-trades',
help='Download trades instead of OHLCV data. The bot will resample trades to the '
'desired timeframe as specified as --timeframes/-t.',
action='store_true',
),
"exchange": Arg(
'--exchange',
help=f'Exchange name (default: `{constants.DEFAULT_EXCHANGE}`). '

View File

@@ -312,6 +312,8 @@ class Configuration:
self._args_to_config(config, argname='days',
logstring='Detected --days: {}')
self._args_to_config(config, argname='download_trades',
logstring='Detected --dl-trades: {}')
def _process_runmode(self, config: Dict[str, Any]) -> None:

View File

@@ -114,3 +114,25 @@ def order_book_to_dataframe(bids: list, asks: list) -> DataFrame:
keys=['b_sum', 'b_size', 'bids', 'asks', 'a_size', 'a_sum'])
# logger.info('order book %s', frame )
return frame
def trades_to_ohlcv(trades: list, timeframe: str) -> list:
"""
Converts trades list to ohlcv list
:param trades: List of trades, as returned by ccxt.fetch_trades.
:param timeframe: Ticker timeframe to resample data to
:return: ohlcv timeframe as list (as returned by ccxt.fetch_ohlcv)
"""
from freqtrade.exchange import timeframe_to_minutes
ticker_minutes = timeframe_to_minutes(timeframe)
df = pd.DataFrame(trades)
df['datetime'] = pd.to_datetime(df['datetime'])
df = df.set_index('datetime')
df_new = df['price'].resample(f'{ticker_minutes}min').ohlc()
df_new['volume'] = df['amount'].resample(f'{ticker_minutes}min').sum()
df_new['date'] = df_new.index.astype("int64") // 10 ** 6
# Drop 0 volume rows
df_new = df_new.dropna()
columns = ["date", "open", "high", "low", "close", "volume"]
return list(zip(*[df_new[x].values.tolist() for x in columns]))

View File

@@ -17,7 +17,7 @@ from pandas import DataFrame
from freqtrade import OperationalException, misc
from freqtrade.configuration import TimeRange
from freqtrade.data.converter import parse_ticker_dataframe
from freqtrade.data.converter import parse_ticker_dataframe, trades_to_ohlcv
from freqtrade.exchange import Exchange, timeframe_to_minutes
logger = logging.getLogger(__name__)
@@ -82,6 +82,29 @@ def store_tickerdata_file(datadir: Path, pair: str,
misc.file_dump_json(filename, data, is_zip=is_zip)
def load_trades_file(datadir: Path, pair: str,
timerange: Optional[TimeRange] = None) -> List[Dict]:
"""
Load a pair from file, either .json.gz or .json
:return: tradelist or empty list if unsuccesful
"""
filename = pair_trades_filename(datadir, pair)
tradesdata = misc.file_load_json(filename)
if not tradesdata:
return []
return tradesdata
def store_trades_file(datadir: Path, pair: str,
data: list, is_zip: bool = True):
"""
Stores tickerdata to file
"""
filename = pair_trades_filename(datadir, pair)
misc.file_dump_json(filename, data, is_zip=is_zip)
def _validate_pairdata(pair, pairdata, timerange: TimeRange):
if timerange.starttype == 'date' and pairdata[0][0] > timerange.startts * 1000:
logger.warning('Missing data at start for pair %s, data starts at %s',
@@ -173,6 +196,12 @@ def pair_data_filename(datadir: Path, pair: str, ticker_interval: str) -> Path:
return filename
def pair_trades_filename(datadir: Path, pair: str) -> Path:
pair_s = pair.replace("/", "_")
filename = datadir.joinpath(f'{pair_s}-trades.json.gz')
return filename
def _load_cached_data_for_updating(datadir: Path, pair: str, ticker_interval: str,
timerange: Optional[TimeRange]) -> Tuple[List[Any],
Optional[int]]:
@@ -299,6 +328,92 @@ def refresh_backtest_ohlcv_data(exchange: Exchange, pairs: List[str], timeframes
return pairs_not_available
def download_trades_history(datadir: Path,
exchange: Exchange,
pair: str,
timerange: Optional[TimeRange] = None) -> bool:
"""
Download trade history from the exchange.
Appends to previously downloaded trades data.
"""
try:
since = timerange.startts * 1000 if timerange and timerange.starttype == 'date' else None
trades = load_trades_file(datadir, pair)
from_id = trades[-1]['id'] if trades else None
logger.debug("Current Start: %s", trades[0]['datetime'] if trades else 'None')
logger.debug("Current End: %s", trades[-1]['datetime'] if trades else 'None')
new_trades = exchange.get_historic_trades(pair=pair,
since=since if since else
int(arrow.utcnow().shift(
days=-30).float_timestamp) * 1000,
# until=xxx,
from_id=from_id,
)
trades.extend(new_trades[1])
store_trades_file(datadir, pair, trades)
logger.debug("New Start: %s", trades[0]['datetime'])
logger.debug("New End: %s", trades[-1]['datetime'])
logger.info(f"New Amount of trades: {len(trades)}")
return True
except Exception as e:
logger.error(
f'Failed to download historic trades for pair: "{pair}". '
f'Error: {e}'
)
return False
def refresh_backtest_trades_data(exchange: Exchange, pairs: List[str], datadir: Path,
timerange: TimeRange, erase=False) -> List[str]:
"""
Refresh stored trades data.
Used by freqtrade download-data
:return: Pairs not available
"""
pairs_not_available = []
for pair in pairs:
if pair not in exchange.markets:
pairs_not_available.append(pair)
logger.info(f"Skipping pair {pair}...")
continue
dl_file = pair_trades_filename(datadir, pair)
if erase and dl_file.exists():
logger.info(
f'Deleting existing data for pair {pair}.')
dl_file.unlink()
logger.info(f'Downloading trades for pair {pair}.')
download_trades_history(datadir=datadir, exchange=exchange,
pair=pair,
timerange=timerange)
return pairs_not_available
def convert_trades_to_ohlcv(pairs: List[str], timeframes: List[str],
datadir: Path, timerange: TimeRange, erase=False) -> None:
"""
Convert stored trades data to ohlcv data
"""
for pair in pairs:
trades = load_trades_file(datadir, pair)
for timeframe in timeframes:
ohlcv_file = pair_data_filename(datadir, pair, timeframe)
if erase and ohlcv_file.exists():
logger.info(f'Deleting existing data for pair {pair}, interval {timeframe}.')
ohlcv_file.unlink()
ohlcv = trades_to_ohlcv(trades, timeframe)
# Store ohlcv
store_tickerdata_file(datadir, pair, timeframe, data=ohlcv)
def get_timeframe(data: Dict[str, DataFrame]) -> Tuple[arrow.Arrow, arrow.Arrow]:
"""
Get the maximum timeframe for the given backtest data

View File

@@ -16,6 +16,8 @@ class Binance(Exchange):
_ft_has: Dict = {
"stoploss_on_exchange": True,
"order_time_in_force": ['gtc', 'fok', 'ioc'],
"trades_pagination": "id",
"trades_pagination_arg": "fromId",
}
def get_order_book(self, pair: str, limit: int = 100) -> dict:

View File

@@ -147,6 +147,8 @@ def retrier(f):
class Exchange:
_config: Dict = {}
# Parameters to add directly to buy/sell calls (like agreeing to trading agreement)
_params: Dict = {}
# Dict to specify which options each exchange implements
@@ -157,6 +159,9 @@ class Exchange:
"order_time_in_force": ["gtc"],
"ohlcv_candle_limit": 500,
"ohlcv_partial_candle": True,
"trades_pagination": "time", # Possible are "time" or "id"
"trades_pagination_arg": "since",
}
_ft_has: Dict = {}
@@ -200,6 +205,9 @@ class Exchange:
self._ohlcv_candle_limit = self._ft_has['ohlcv_candle_limit']
self._ohlcv_partial_candle = self._ft_has['ohlcv_partial_candle']
self._trades_pagination = self._ft_has['trades_pagination']
self._trades_pagination_arg = self._ft_has['trades_pagination_arg']
# Initialize ccxt objects
self._api = self._init_ccxt(
exchange_config, ccxt_kwargs=exchange_config.get('ccxt_config'))
@@ -742,6 +750,154 @@ class Exchange:
except ccxt.BaseError as e:
raise OperationalException(f'Could not fetch ticker data. Msg: {e}') from e
@retrier_async
async def _async_fetch_trades(self, pair: str,
since: Optional[int] = None,
params: Optional[dict] = None) -> List[Dict]:
"""
Asyncronously gets trade history using fetch_trades.
Handles exchange errors, does one call to the exchange.
:param pair: Pair to fetch trade data for
:param since: Since as integer timestamp in milliseconds
returns: List of dicts containing trades
"""
try:
# fetch trades asynchronously
if params:
logger.debug("Fetching trades for pair %s, params: %s ", pair, params)
trades = await self._api_async.fetch_trades(pair, params=params, limit=1000)
else:
logger.debug(
"Fetching trades for pair %s, since %s %s...",
pair, since,
'(' + arrow.get(since // 1000).isoformat() + ') ' if since is not None else ''
)
trades = await self._api_async.fetch_trades(pair, since=since, limit=1000)
return trades
except ccxt.NotSupported as e:
raise OperationalException(
f'Exchange {self._api.name} does not support fetching historical trade data.'
f'Message: {e}') from e
except (ccxt.NetworkError, ccxt.ExchangeError) as e:
raise TemporaryError(f'Could not load trade history due to {e.__class__.__name__}. '
f'Message: {e}') from e
except ccxt.BaseError as e:
raise OperationalException(f'Could not fetch trade data. Msg: {e}') from e
async def _async_get_trade_history_id(self, pair: str,
until: int,
since: Optional[int] = None,
from_id: Optional[str] = None) -> Tuple[str, List[Dict]]:
"""
Asyncronously gets trade history using fetch_trades
use this when exchange uses id-based iteration (check `self._trades_pagination`)
:param pair: Pair to fetch trade data for
:param since: Since as integer timestamp in milliseconds
:param until: Until as integer timestamp in milliseconds
:param from_id: Download data starting with ID (if id is known). Ignores "since" if set.
returns tuple: (pair, trades-list)
"""
trades: List[Dict] = []
if not from_id:
# Fetch first elements using timebased method to get an ID to paginate on
# Depending on the Exchange, this can introduce a drift at the start of the interval
# of up to an hour.
# e.g. Binance returns the "last 1000" candles within a 1h time interval
# - so we will miss the first trades.
t = await self._async_fetch_trades(pair, since=since)
from_id = t[-1]['id']
trades.extend(t[:-1])
while True:
t = await self._async_fetch_trades(pair,
params={self._trades_pagination_arg: from_id})
if len(t):
# Skip last id since its the key for the next call
trades.extend(t[:-1])
if from_id == t[-1]['id'] or t[-1]['timestamp'] > until:
logger.debug(f"Stopping because from_id did not change. "
f"Reached {t[-1]['timestamp']} > {until}")
# Reached the end of the defined-download period - add last trade as well.
trades.extend(t[-1:])
break
from_id = t[-1]['id']
else:
break
return (pair, trades)
async def _async_get_trade_history_time(self, pair: str, until: int,
since: Optional[int] = None) -> Tuple[str, List]:
"""
Asyncronously gets trade history using fetch_trades,
when the exchange uses time-based iteration (check `self._trades_pagination`)
:param pair: Pair to fetch trade data for
:param since: Since as integer timestamp in milliseconds
:param until: Until as integer timestamp in milliseconds
returns tuple: (pair, trades-list)
"""
trades: List[Dict] = []
while True:
t = await self._async_fetch_trades(pair, since=since)
if len(t):
since = t[-1]['timestamp']
trades.extend(t)
# Reached the end of the defined-download period
if until and t[-1]['timestamp'] > until:
logger.debug(
f"Stopping because until was reached. {t[-1]['timestamp']} > {until}")
break
else:
break
return (pair, trades)
async def _async_get_trade_history(self, pair: str,
since: Optional[int] = None,
until: Optional[int] = None,
from_id: Optional[str] = None) -> Tuple[str, List[Dict]]:
"""
Async wrapper handling downloading trades using either time or id based methods.
"""
if self._trades_pagination == 'time':
return await self._async_get_trade_history_time(
pair=pair, since=since,
until=until or ccxt.Exchange.milliseconds())
elif self._trades_pagination == 'id':
return await self._async_get_trade_history_id(
pair=pair, since=since,
until=until or ccxt.Exchange.milliseconds(), from_id=from_id
)
else:
raise OperationalException(f"Exchange {self.name} does use neither time, "
f"nor id based pagination")
def get_historic_trades(self, pair: str,
since: Optional[int] = None,
until: Optional[int] = None,
from_id: Optional[str] = None) -> Tuple[str, List]:
"""
Gets candle history using asyncio and returns the list of candles.
Handles all async doing.
Async over one pair, assuming we get `_ohlcv_candle_limit` candles per call.
:param pair: Pair to download
:param ticker_interval: Interval to get
:param since: Timestamp in milliseconds to get history from
:param until: Timestamp in milliseconds. Defaults to current timestamp if not defined.
:param from_id: Download data starting with ID (if id is known)
:returns List of tickers
"""
if not self.exchange_has("fetchTrades"):
raise OperationalException("This exchange does not suport downloading Trades.")
return asyncio.get_event_loop().run_until_complete(
self._async_get_trade_history(pair=pair, since=since,
until=until, from_id=from_id))
@retrier
def cancel_order(self, order_id: str, pair: str) -> None:
if self._config['dry_run']:

View File

@@ -14,6 +14,10 @@ logger = logging.getLogger(__name__)
class Kraken(Exchange):
_params: Dict = {"trading_agreement": "agree"}
_ft_has: Dict = {
"trades_pagination": "id",
"trades_pagination_arg": "since",
}
@retrier
def get_balances(self) -> dict:

View File

@@ -72,8 +72,10 @@ def json_load(datafile: IO):
def file_load_json(file):
gzipfile = file.with_suffix(file.suffix + '.gz')
if file.suffix != ".gz":
gzipfile = file.with_suffix(file.suffix + '.gz')
else:
gzipfile = file
# Try gzip file first, otherwise regular json file.
if gzipfile.is_file():
logger.debug('Loading ticker data from file %s', gzipfile)

View File

@@ -8,7 +8,9 @@ import arrow
from freqtrade import OperationalException
from freqtrade.configuration import Configuration, TimeRange
from freqtrade.configuration.directory_operations import create_userdata_dir
from freqtrade.data.history import refresh_backtest_ohlcv_data
from freqtrade.data.history import (convert_trades_to_ohlcv,
refresh_backtest_ohlcv_data,
refresh_backtest_trades_data)
from freqtrade.exchange import available_exchanges, ccxt_exchanges
from freqtrade.resolvers import ExchangeResolver
from freqtrade.state import RunMode
@@ -88,9 +90,19 @@ def start_download_data(args: Dict[str, Any]) -> None:
# Init exchange
exchange = ExchangeResolver(config['exchange']['name'], config).exchange
pairs_not_available = refresh_backtest_ohlcv_data(
exchange, pairs=config["pairs"], timeframes=config["timeframes"],
dl_path=Path(config['datadir']), timerange=timerange, erase=config.get("erase"))
if config.get('download_trades'):
pairs_not_available = refresh_backtest_trades_data(
exchange, pairs=config["pairs"], datadir=Path(config['datadir']),
timerange=timerange, erase=config.get("erase"))
# Convert downloaded trade data to different timeframes
convert_trades_to_ohlcv(
pairs=config["pairs"], timeframes=config["timeframes"],
datadir=Path(config['datadir']), timerange=timerange, erase=config.get("erase"))
else:
pairs_not_available = refresh_backtest_ohlcv_data(
exchange, pairs=config["pairs"], timeframes=config["timeframes"],
dl_path=Path(config['datadir']), timerange=timerange, erase=config.get("erase"))
except KeyboardInterrupt:
sys.exit("SIGINT received, aborting ...")