2018-12-16 08:58:46 +00:00
|
|
|
"""
|
|
|
|
Handle historic data (ohlcv).
|
2019-04-09 09:27:35 +00:00
|
|
|
|
|
|
|
Includes:
|
2018-12-16 08:58:46 +00:00
|
|
|
* load data for a pair (or a list of pairs) from disk
|
|
|
|
* download data from exchange and store to disk
|
|
|
|
"""
|
2019-05-25 14:51:52 +00:00
|
|
|
|
2018-12-13 05:12:10 +00:00
|
|
|
import logging
|
2019-05-25 14:51:52 +00:00
|
|
|
import operator
|
|
|
|
from datetime import datetime
|
2018-12-15 12:54:35 +00:00
|
|
|
from pathlib import Path
|
2019-05-25 14:51:52 +00:00
|
|
|
from typing import Any, Dict, List, Optional, Tuple
|
2018-12-13 05:12:10 +00:00
|
|
|
|
|
|
|
import arrow
|
2018-12-15 13:28:37 +00:00
|
|
|
from pandas import DataFrame
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2019-05-25 14:51:52 +00:00
|
|
|
from freqtrade import OperationalException, misc
|
2019-07-11 18:23:23 +00:00
|
|
|
from freqtrade.configuration import TimeRange
|
2018-12-15 13:28:37 +00:00
|
|
|
from freqtrade.data.converter import parse_ticker_dataframe
|
2019-04-09 09:27:35 +00:00
|
|
|
from freqtrade.exchange import Exchange, timeframe_to_minutes
|
2019-04-04 17:56:40 +00:00
|
|
|
|
2018-12-13 05:12:10 +00:00
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
|
|
|
|
|
|
def trim_tickerlist(tickerlist: List[Dict], timerange: TimeRange) -> List[Dict]:
|
|
|
|
"""
|
|
|
|
Trim tickerlist based on given timerange
|
|
|
|
"""
|
|
|
|
if not tickerlist:
|
|
|
|
return tickerlist
|
|
|
|
|
|
|
|
start_index = 0
|
|
|
|
stop_index = len(tickerlist)
|
|
|
|
|
|
|
|
if timerange.starttype == 'line':
|
|
|
|
stop_index = timerange.startts
|
|
|
|
if timerange.starttype == 'index':
|
|
|
|
start_index = timerange.startts
|
|
|
|
elif timerange.starttype == 'date':
|
|
|
|
while (start_index < len(tickerlist) and
|
|
|
|
tickerlist[start_index][0] < timerange.startts * 1000):
|
|
|
|
start_index += 1
|
|
|
|
|
|
|
|
if timerange.stoptype == 'line':
|
2019-08-15 04:35:50 +00:00
|
|
|
start_index = max(len(tickerlist) + timerange.stopts, 0)
|
2018-12-13 05:12:10 +00:00
|
|
|
if timerange.stoptype == 'index':
|
|
|
|
stop_index = timerange.stopts
|
|
|
|
elif timerange.stoptype == 'date':
|
|
|
|
while (stop_index > 0 and
|
|
|
|
tickerlist[stop_index-1][0] > timerange.stopts * 1000):
|
|
|
|
stop_index -= 1
|
|
|
|
|
|
|
|
if start_index > stop_index:
|
|
|
|
raise ValueError(f'The timerange [{timerange.startts},{timerange.stopts}] is incorrect')
|
|
|
|
|
|
|
|
return tickerlist[start_index:stop_index]
|
|
|
|
|
|
|
|
|
2019-08-14 18:45:24 +00:00
|
|
|
def load_tickerdata_file(datadir: Optional[Path], pair: str, ticker_interval: str,
|
|
|
|
timerange: Optional[TimeRange] = None) -> Optional[list]:
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2018-12-15 18:52:52 +00:00
|
|
|
Load a pair from file, either .json.gz or .json
|
2019-06-23 20:10:37 +00:00
|
|
|
:return: tickerlist or None if unsuccesful
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2019-05-21 17:49:02 +00:00
|
|
|
filename = pair_data_filename(datadir, pair, ticker_interval)
|
|
|
|
pairdata = misc.file_load_json(filename)
|
2018-12-28 09:25:12 +00:00
|
|
|
if not pairdata:
|
2019-08-14 18:45:24 +00:00
|
|
|
return []
|
2018-12-13 05:12:10 +00:00
|
|
|
|
|
|
|
if timerange:
|
|
|
|
pairdata = trim_tickerlist(pairdata, timerange)
|
|
|
|
return pairdata
|
|
|
|
|
|
|
|
|
2019-08-14 18:48:42 +00:00
|
|
|
def store_tickerdata_file(datadir: Optional[Path], pair: str,
|
|
|
|
ticker_interval: str, data: list, is_zip: bool = False):
|
|
|
|
"""
|
|
|
|
Stores tickerdata to file
|
|
|
|
"""
|
|
|
|
filename = pair_data_filename(datadir, pair, ticker_interval)
|
|
|
|
misc.file_dump_json(filename, data, is_zip=is_zip)
|
|
|
|
|
|
|
|
|
2018-12-15 19:31:25 +00:00
|
|
|
def load_pair_history(pair: str,
|
|
|
|
ticker_interval: str,
|
|
|
|
datadir: Optional[Path],
|
2018-12-16 09:17:11 +00:00
|
|
|
timerange: TimeRange = TimeRange(None, None, 0, 0),
|
|
|
|
refresh_pairs: bool = False,
|
|
|
|
exchange: Optional[Exchange] = None,
|
2019-06-09 12:40:45 +00:00
|
|
|
fill_up_missing: bool = True,
|
|
|
|
drop_incomplete: bool = True
|
2018-12-16 09:17:11 +00:00
|
|
|
) -> DataFrame:
|
2018-12-15 19:31:25 +00:00
|
|
|
"""
|
|
|
|
Loads cached ticker history for the given pair.
|
2019-06-09 12:40:45 +00:00
|
|
|
:param pair: Pair to load data for
|
|
|
|
:param ticker_interval: Ticker-interval (e.g. "5m")
|
|
|
|
:param datadir: Path to the data storage location.
|
|
|
|
:param timerange: Limit data to be loaded to this timerange
|
|
|
|
:param refresh_pairs: Refresh pairs from exchange.
|
|
|
|
(Note: Requires exchange to be passed as well.)
|
|
|
|
:param exchange: Exchange object (needed when using "refresh_pairs")
|
|
|
|
:param fill_up_missing: Fill missing values with "No action"-candles
|
|
|
|
:param drop_incomplete: Drop last candle assuming it may be incomplete.
|
2018-12-16 13:14:17 +00:00
|
|
|
:return: DataFrame with ohlcv data
|
2018-12-15 19:31:25 +00:00
|
|
|
"""
|
|
|
|
|
2019-05-17 16:05:36 +00:00
|
|
|
# The user forced the refresh of pairs
|
2018-12-16 09:17:11 +00:00
|
|
|
if refresh_pairs:
|
2018-12-16 09:33:08 +00:00
|
|
|
download_pair_history(datadir=datadir,
|
|
|
|
exchange=exchange,
|
|
|
|
pair=pair,
|
2019-04-07 13:14:40 +00:00
|
|
|
ticker_interval=ticker_interval,
|
2018-12-16 09:33:08 +00:00
|
|
|
timerange=timerange)
|
2018-12-16 09:17:11 +00:00
|
|
|
|
2019-01-01 12:42:30 +00:00
|
|
|
pairdata = load_tickerdata_file(datadir, pair, ticker_interval, timerange=timerange)
|
|
|
|
|
2018-12-15 19:31:25 +00:00
|
|
|
if pairdata:
|
|
|
|
if timerange.starttype == 'date' and pairdata[0][0] > timerange.startts * 1000:
|
|
|
|
logger.warning('Missing data at start for pair %s, data starts at %s',
|
|
|
|
pair, arrow.get(pairdata[0][0] // 1000).strftime('%Y-%m-%d %H:%M:%S'))
|
|
|
|
if timerange.stoptype == 'date' and pairdata[-1][0] < timerange.stopts * 1000:
|
|
|
|
logger.warning('Missing data at end for pair %s, data ends at %s',
|
|
|
|
pair,
|
|
|
|
arrow.get(pairdata[-1][0] // 1000).strftime('%Y-%m-%d %H:%M:%S'))
|
2019-06-15 11:46:19 +00:00
|
|
|
return parse_ticker_dataframe(pairdata, ticker_interval, pair=pair,
|
2019-06-09 12:40:45 +00:00
|
|
|
fill_missing=fill_up_missing,
|
|
|
|
drop_incomplete=drop_incomplete)
|
2018-12-15 19:31:25 +00:00
|
|
|
else:
|
2019-05-17 16:05:36 +00:00
|
|
|
logger.warning(
|
|
|
|
f'No history data for pair: "{pair}", interval: {ticker_interval}. '
|
|
|
|
'Use --refresh-pairs-cached option or download_backtest_data.py '
|
|
|
|
'script to download the data'
|
|
|
|
)
|
2018-12-15 19:31:25 +00:00
|
|
|
return None
|
|
|
|
|
|
|
|
|
2018-12-15 12:55:16 +00:00
|
|
|
def load_data(datadir: Optional[Path],
|
2018-12-13 05:12:10 +00:00
|
|
|
ticker_interval: str,
|
|
|
|
pairs: List[str],
|
2018-12-16 09:17:11 +00:00
|
|
|
refresh_pairs: bool = False,
|
2018-12-13 05:12:10 +00:00
|
|
|
exchange: Optional[Exchange] = None,
|
2018-12-31 18:13:34 +00:00
|
|
|
timerange: TimeRange = TimeRange(None, None, 0, 0),
|
2019-05-29 18:10:48 +00:00
|
|
|
fill_up_missing: bool = True,
|
|
|
|
live: bool = False
|
|
|
|
) -> Dict[str, DataFrame]:
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2018-12-16 09:17:11 +00:00
|
|
|
Loads ticker history data for a list of pairs the given parameters
|
2018-12-16 13:14:17 +00:00
|
|
|
:return: dict(<pair>:<tickerlist>)
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2019-05-29 18:25:07 +00:00
|
|
|
result: Dict[str, DataFrame] = {}
|
2019-05-29 18:10:48 +00:00
|
|
|
if live:
|
2019-05-29 18:25:07 +00:00
|
|
|
if exchange:
|
|
|
|
logger.info('Live: Downloading data for all defined pairs ...')
|
|
|
|
exchange.refresh_latest_ohlcv([(pair, ticker_interval) for pair in pairs])
|
|
|
|
result = {key[0]: value for key, value in exchange._klines.items() if value is not None}
|
|
|
|
else:
|
|
|
|
raise OperationalException(
|
|
|
|
"Exchange needs to be initialized when using live data."
|
|
|
|
)
|
2019-05-29 18:10:48 +00:00
|
|
|
else:
|
|
|
|
logger.info('Using local backtesting data ...')
|
|
|
|
|
|
|
|
for pair in pairs:
|
|
|
|
hist = load_pair_history(pair=pair, ticker_interval=ticker_interval,
|
|
|
|
datadir=datadir, timerange=timerange,
|
|
|
|
refresh_pairs=refresh_pairs,
|
|
|
|
exchange=exchange,
|
|
|
|
fill_up_missing=fill_up_missing)
|
|
|
|
if hist is not None:
|
|
|
|
result[pair] = hist
|
2018-12-13 05:12:10 +00:00
|
|
|
return result
|
|
|
|
|
|
|
|
|
2019-05-22 11:04:58 +00:00
|
|
|
def make_testdata_path(datadir: Optional[Path]) -> Path:
|
2018-12-13 05:12:10 +00:00
|
|
|
"""Return the path where testdata files are stored"""
|
2018-12-15 12:54:35 +00:00
|
|
|
return datadir or (Path(__file__).parent.parent / "tests" / "testdata").resolve()
|
2018-12-13 05:12:10 +00:00
|
|
|
|
|
|
|
|
2019-05-21 17:49:02 +00:00
|
|
|
def pair_data_filename(datadir: Optional[Path], pair: str, ticker_interval: str) -> Path:
|
2019-05-22 11:04:58 +00:00
|
|
|
path = make_testdata_path(datadir)
|
2019-05-21 17:49:02 +00:00
|
|
|
pair_s = pair.replace("/", "_")
|
|
|
|
filename = path.joinpath(f'{pair_s}-{ticker_interval}.json')
|
|
|
|
return filename
|
|
|
|
|
|
|
|
|
2019-08-14 19:39:53 +00:00
|
|
|
def load_cached_data_for_updating(datadir: Optional[Path], pair: str, ticker_interval: str,
|
2018-12-15 18:52:52 +00:00
|
|
|
timerange: Optional[TimeRange]) -> Tuple[List[Any],
|
|
|
|
Optional[int]]:
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2019-08-15 18:13:19 +00:00
|
|
|
Load cached data to download more data.
|
|
|
|
If timerange is passed in, checks wether data from an before the stored data will be downloaded.
|
|
|
|
If that's the case than what's available should be completely overwritten.
|
2019-08-14 18:45:24 +00:00
|
|
|
Only used by download_pair_history().
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
|
|
|
|
|
|
|
since_ms = None
|
|
|
|
|
|
|
|
# user sets timerange, so find the start time
|
|
|
|
if timerange:
|
|
|
|
if timerange.starttype == 'date':
|
|
|
|
since_ms = timerange.startts * 1000
|
|
|
|
elif timerange.stoptype == 'line':
|
2019-04-07 13:14:40 +00:00
|
|
|
num_minutes = timerange.stopts * timeframe_to_minutes(ticker_interval)
|
2018-12-13 05:12:10 +00:00
|
|
|
since_ms = arrow.utcnow().shift(minutes=num_minutes).timestamp * 1000
|
|
|
|
|
|
|
|
# read the cached file
|
2019-08-15 18:09:00 +00:00
|
|
|
# Intentionally don't pass timerange in - since we need to load the full dataset.
|
|
|
|
data = load_tickerdata_file(datadir, pair, ticker_interval)
|
2019-08-14 18:48:42 +00:00
|
|
|
# remove the last item, could be incomplete candle
|
|
|
|
if data:
|
|
|
|
data.pop()
|
2018-12-13 05:12:10 +00:00
|
|
|
else:
|
|
|
|
data = []
|
|
|
|
|
|
|
|
if data:
|
|
|
|
if since_ms and since_ms < data[0][0]:
|
2018-12-15 18:52:52 +00:00
|
|
|
# Earlier data than existing data requested, redownload all
|
2018-12-13 05:12:10 +00:00
|
|
|
data = []
|
|
|
|
else:
|
2018-12-15 18:52:52 +00:00
|
|
|
# a part of the data was already downloaded, so download unexist data only
|
2018-12-13 05:12:10 +00:00
|
|
|
since_ms = data[-1][0] + 1
|
|
|
|
|
|
|
|
return (data, since_ms)
|
|
|
|
|
|
|
|
|
2018-12-16 09:33:08 +00:00
|
|
|
def download_pair_history(datadir: Optional[Path],
|
2019-05-17 16:05:36 +00:00
|
|
|
exchange: Optional[Exchange],
|
2018-12-16 09:33:08 +00:00
|
|
|
pair: str,
|
2019-04-07 13:14:40 +00:00
|
|
|
ticker_interval: str = '5m',
|
2018-12-16 09:33:08 +00:00
|
|
|
timerange: Optional[TimeRange] = None) -> bool:
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
|
|
|
Download the latest ticker intervals from the exchange for the pair passed in parameters
|
|
|
|
The data is downloaded starting from the last correct ticker interval data that
|
|
|
|
exists in a cache. If timerange starts earlier than the data in the cache,
|
|
|
|
the full data will be redownloaded
|
|
|
|
|
|
|
|
Based on @Rybolov work: https://github.com/rybolov/freqtrade-data
|
2019-05-17 16:05:36 +00:00
|
|
|
|
2018-12-13 05:12:10 +00:00
|
|
|
:param pair: pair to download
|
2019-04-07 13:14:40 +00:00
|
|
|
:param ticker_interval: ticker interval
|
2018-12-13 05:12:10 +00:00
|
|
|
:param timerange: range of time to download
|
2018-12-16 13:14:17 +00:00
|
|
|
:return: bool with success state
|
2018-12-13 05:12:10 +00:00
|
|
|
"""
|
2019-05-17 16:05:36 +00:00
|
|
|
if not exchange:
|
|
|
|
raise OperationalException(
|
|
|
|
"Exchange needs to be initialized when downloading pair history data"
|
|
|
|
)
|
|
|
|
|
2018-12-16 09:29:53 +00:00
|
|
|
try:
|
2019-05-17 16:05:36 +00:00
|
|
|
logger.info(
|
|
|
|
f'Download history data for pair: "{pair}", interval: {ticker_interval} '
|
|
|
|
f'and store in {datadir}.'
|
|
|
|
)
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2019-08-14 18:48:42 +00:00
|
|
|
data, since_ms = load_cached_data_for_updating(datadir, pair, ticker_interval, timerange)
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2018-12-16 09:29:53 +00:00
|
|
|
logger.debug("Current Start: %s", misc.format_ms_time(data[1][0]) if data else 'None')
|
|
|
|
logger.debug("Current End: %s", misc.format_ms_time(data[-1][0]) if data else 'None')
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2018-12-16 09:29:53 +00:00
|
|
|
# Default since_ms to 30 days if nothing is given
|
2019-08-14 08:14:54 +00:00
|
|
|
new_data = exchange.get_historic_ohlcv(pair=pair, ticker_interval=ticker_interval,
|
|
|
|
since_ms=since_ms if since_ms
|
|
|
|
else
|
|
|
|
int(arrow.utcnow().shift(
|
|
|
|
days=-30).float_timestamp) * 1000)
|
2018-12-16 09:29:53 +00:00
|
|
|
data.extend(new_data)
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2018-12-16 09:29:53 +00:00
|
|
|
logger.debug("New Start: %s", misc.format_ms_time(data[0][0]))
|
|
|
|
logger.debug("New End: %s", misc.format_ms_time(data[-1][0]))
|
2018-12-13 05:12:10 +00:00
|
|
|
|
2019-08-14 18:48:42 +00:00
|
|
|
store_tickerdata_file(datadir, pair, ticker_interval, data=data)
|
2018-12-16 09:29:53 +00:00
|
|
|
return True
|
2019-05-17 16:05:36 +00:00
|
|
|
|
2019-05-21 17:49:02 +00:00
|
|
|
except Exception as e:
|
2019-05-17 16:05:36 +00:00
|
|
|
logger.error(
|
2019-05-21 17:49:02 +00:00
|
|
|
f'Failed to download history data for pair: "{pair}", interval: {ticker_interval}. '
|
|
|
|
f'Error: {e}'
|
2019-05-17 16:05:36 +00:00
|
|
|
)
|
2019-01-31 05:51:03 +00:00
|
|
|
return False
|
2019-05-25 14:51:52 +00:00
|
|
|
|
|
|
|
|
|
|
|
def get_timeframe(data: Dict[str, DataFrame]) -> Tuple[arrow.Arrow, arrow.Arrow]:
|
|
|
|
"""
|
|
|
|
Get the maximum timeframe for the given backtest data
|
|
|
|
:param data: dictionary with preprocessed backtesting data
|
|
|
|
:return: tuple containing min_date, max_date
|
|
|
|
"""
|
|
|
|
timeframe = [
|
|
|
|
(arrow.get(frame['date'].min()), arrow.get(frame['date'].max()))
|
|
|
|
for frame in data.values()
|
|
|
|
]
|
|
|
|
return min(timeframe, key=operator.itemgetter(0))[0], \
|
|
|
|
max(timeframe, key=operator.itemgetter(1))[1]
|
|
|
|
|
|
|
|
|
2019-06-15 11:31:27 +00:00
|
|
|
def validate_backtest_data(data: DataFrame, pair: str, min_date: datetime,
|
2019-05-25 14:51:52 +00:00
|
|
|
max_date: datetime, ticker_interval_mins: int) -> bool:
|
|
|
|
"""
|
|
|
|
Validates preprocessed backtesting data for missing values and shows warnings about it that.
|
|
|
|
|
2019-06-15 11:31:27 +00:00
|
|
|
:param data: preprocessed backtesting data (as DataFrame)
|
|
|
|
:param pair: pair used for log output.
|
2019-05-25 14:51:52 +00:00
|
|
|
:param min_date: start-date of the data
|
|
|
|
:param max_date: end-date of the data
|
|
|
|
:param ticker_interval_mins: ticker interval in minutes
|
|
|
|
"""
|
|
|
|
# total difference in minutes / interval-minutes
|
|
|
|
expected_frames = int((max_date - min_date).total_seconds() // 60 // ticker_interval_mins)
|
|
|
|
found_missing = False
|
2019-06-15 11:31:27 +00:00
|
|
|
dflen = len(data)
|
|
|
|
if dflen < expected_frames:
|
|
|
|
found_missing = True
|
|
|
|
logger.warning("%s has missing frames: expected %s, got %s, that's %s missing values",
|
|
|
|
pair, expected_frames, dflen, expected_frames - dflen)
|
2019-05-25 14:51:52 +00:00
|
|
|
return found_missing
|