stable/freqtrade/data/history.py

238 lines
8.7 KiB
Python
Raw Normal View History

2018-12-16 08:58:46 +00:00
"""
Handle historic data (ohlcv).
includes:
* load data for a pair (or a list of pairs) from disk
* download data from exchange and store to disk
"""
import logging
2018-12-15 12:54:35 +00:00
from pathlib import Path
from typing import Optional, List, Dict, Tuple, Any
import arrow
2018-12-15 13:28:37 +00:00
from pandas import DataFrame
from freqtrade import misc, OperationalException
from freqtrade.arguments import TimeRange
2018-12-15 13:28:37 +00:00
from freqtrade.data.converter import parse_ticker_dataframe
from freqtrade.exchange import Exchange
from freqtrade.misc import timeframe_to_minutes
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':
start_index = len(tickerlist) + timerange.stopts
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]
def load_tickerdata_file(
2018-12-15 13:10:45 +00:00
datadir: Optional[Path], pair: str,
ticker_interval: str,
2018-12-16 13:14:17 +00:00
timerange: Optional[TimeRange] = None) -> Optional[list]:
"""
Load a pair from file, either .json.gz or .json
2018-12-16 13:14:17 +00:00
:return tickerlist or None if unsuccesful
"""
path = make_testdata_path(datadir)
pair_s = pair.replace('/', '_')
2018-12-15 12:54:35 +00:00
file = path.joinpath(f'{pair_s}-{ticker_interval}.json')
pairdata = misc.file_load_json(file)
if not pairdata:
return None
if timerange:
pairdata = trim_tickerlist(pairdata, timerange)
return pairdata
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,
fill_up_missing: bool = True
2018-12-16 09:17:11 +00:00
) -> DataFrame:
"""
Loads cached ticker history for the given pair.
2018-12-16 13:14:17 +00:00
:return: DataFrame with ohlcv data
"""
2018-12-16 09:17:11 +00:00
# If the user force the refresh of pairs
if refresh_pairs:
if not exchange:
raise OperationalException("Exchange needs to be initialized when "
"calling load_data with refresh_pairs=True")
2019-01-01 12:42:30 +00:00
logger.info('Download data for pair and store them in %s', datadir)
download_pair_history(datadir=datadir,
exchange=exchange,
pair=pair,
ticker_interval=ticker_interval,
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)
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'))
return parse_ticker_dataframe(pairdata, ticker_interval, fill_up_missing)
else:
logger.warning('No data for pair: "%s", Interval: %s. '
'Use --refresh-pairs-cached to download the data',
pair, ticker_interval)
return None
2018-12-15 12:55:16 +00:00
def load_data(datadir: Optional[Path],
ticker_interval: str,
pairs: List[str],
2018-12-16 09:17:11 +00:00
refresh_pairs: bool = False,
exchange: Optional[Exchange] = None,
timerange: TimeRange = TimeRange(None, None, 0, 0),
fill_up_missing: bool = True) -> Dict[str, DataFrame]:
"""
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>)
"""
result = {}
for pair in pairs:
hist = load_pair_history(pair=pair, ticker_interval=ticker_interval,
2018-12-16 09:17:11 +00:00
datadir=datadir, timerange=timerange,
refresh_pairs=refresh_pairs,
exchange=exchange,
fill_up_missing=fill_up_missing)
if hist is not None:
result[pair] = hist
return result
2018-12-15 12:54:35 +00:00
def make_testdata_path(datadir: Optional[Path]) -> Path:
"""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()
def load_cached_data_for_updating(filename: Path, ticker_interval: str,
timerange: Optional[TimeRange]) -> Tuple[List[Any],
Optional[int]]:
"""
Load cached data and choose what part of the data should be updated
"""
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':
num_minutes = timerange.stopts * timeframe_to_minutes(ticker_interval)
since_ms = arrow.utcnow().shift(minutes=num_minutes).timestamp * 1000
# read the cached file
2018-12-15 12:54:35 +00:00
if filename.is_file():
with open(filename, "rt") as file:
2018-12-28 09:04:28 +00:00
data = misc.json_load(file)
2018-12-16 13:14:17 +00:00
# remove the last item, could be incomplete candle
if data:
data.pop()
else:
data = []
if data:
if since_ms and since_ms < data[0][0]:
# Earlier data than existing data requested, redownload all
data = []
else:
# a part of the data was already downloaded, so download unexist data only
since_ms = data[-1][0] + 1
return (data, since_ms)
def download_pair_history(datadir: Optional[Path],
exchange: Exchange,
pair: str,
ticker_interval: str = '5m',
timerange: Optional[TimeRange] = None) -> bool:
"""
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
:param pair: pair to download
:param ticker_interval: ticker interval
:param timerange: range of time to download
2018-12-16 13:14:17 +00:00
:return: bool with success state
"""
2018-12-16 09:29:53 +00:00
try:
path = make_testdata_path(datadir)
filepair = pair.replace("/", "_")
filename = path.joinpath(f'{filepair}-{ticker_interval}.json')
logger.info('Download the pair: "%s", Interval: %s', pair, ticker_interval)
data, since_ms = load_cached_data_for_updating(filename, ticker_interval, timerange)
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')
2018-12-16 09:29:53 +00:00
# Default since_ms to 30 days if nothing is given
new_data = exchange.get_history(pair=pair, ticker_interval=ticker_interval,
2018-12-16 09:29:53 +00:00
since_ms=since_ms if since_ms
else
int(arrow.utcnow().shift(days=-30).float_timestamp) * 1000)
data.extend(new_data)
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]))
2018-12-16 09:29:53 +00:00
misc.file_dump_json(filename, data)
return True
except BaseException:
2019-01-31 05:51:03 +00:00
logger.info('Failed to download the pair: "%s", Interval: %s',
pair, ticker_interval)
2019-01-31 05:51:03 +00:00
return False