def get_training_and_testing_candles(start_date_str: str, finish_date_str: str) -> Tuple[
Dict[str, Dict[str, Union[Union[str, ndarray], Any]]], Dict[str, Dict[str, Union[Union[str, ndarray], Any]]]]:
start_date = jh.arrow_to_timestamp(arrow.get(start_date_str, 'YYYY-MM-DD'))
finish_date = jh.arrow_to_timestamp(arrow.get(finish_date_str, 'YYYY-MM-DD')) - 60000
# Load candles (first try cache, then database)
candles = load_candles(start_date_str, finish_date_str)
# divide into training(85%) and testing(15%) sets
training_candles = {}
testing_candles = {}
days_diff = jh.date_diff_in_days(jh.timestamp_to_arrow(start_date), jh.timestamp_to_arrow(finish_date))
divider_index = int(days_diff * 0.85) * 1440
for key in candles:
training_candles[key] = {
'exchange': candles[key]['exchange'],
'symbol': candles[key]['symbol'],
'candles': candles[key]['candles'][0:divider_index],
}
testing_candles[key] = {
'exchange': candles[key]['exchange'],
'symbol': candles[key]['symbol'],
'candles': candles[key]['candles'][divider_index:],
}
return training_candles, testing_candles
def get_training_and_testing_candles(start_date_str: str,
finish_date_str: str):
start_date = jh.arrow_to_timestamp(arrow.get(start_date_str, 'YYYY-MM-DD'))
finish_date = jh.arrow_to_timestamp(
arrow.get(finish_date_str, 'YYYY-MM-DD')) - 60000
# Load candles (first try cache, then database)
from jesse.modes.backtest_mode import load_candles
candles = load_candles(start_date_str, finish_date_str)
# divide into training(85%) and testing(15%) sets
training_candles = {}
testing_candles = {}
days_diff = jh.date_diff_in_days(jh.timestamp_to_arrow(start_date),
jh.timestamp_to_arrow(finish_date))
divider_index = int(days_diff * 0.85) * 1440
for key in candles:
training_candles[key] = {
'exchange': candles[key]['exchange'],
'symbol': candles[key]['symbol'],
'candles': candles[key]['candles'][0:divider_index],
}
testing_candles[key] = {
'exchange': candles[key]['exchange'],
'symbol': candles[key]['symbol'],
'candles': candles[key]['candles'][divider_index:],
}
return training_candles, testing_candles
def candles_info(candles_array: np.ndarray) -> dict:
period = jh.date_diff_in_days(jh.timestamp_to_arrow(
candles_array[0][0]), jh.timestamp_to_arrow(candles_array[-1][0])) + 1
if period > 365:
duration = f'{period} days ({round(period / 365, 2)} years)'
elif period > 30:
duration = f'{period} days ({round(period / 30, 2)} months)'
else:
duration = f'{period} days'
return {
'duration': duration,
'starting_time': candles_array[0][0],
'finishing_time': (candles_array[-1][0] + 60_000),
}
def candles(candles_array: np.ndarray) -> List[List[str]]:
period = jh.date_diff_in_days(jh.timestamp_to_arrow(
candles_array[0][0]), jh.timestamp_to_arrow(candles_array[-1][0])) + 1
if period > 365:
duration = f'{period} days ({round(period / 365, 2)} years)'
elif period > 30:
duration = f'{period} days ({round(period / 30, 2)} months)'
else:
duration = f'{period} days'
return [
['period', duration],
[
'starting-ending date',
f'{jh.timestamp_to_time(candles_array[0][0])[:10]} => {jh.timestamp_to_time(candles_array[-1][0] + 60_000)[:10]}'
],
]
def candles(candles_array):
period = jh.date_diff_in_days(jh.timestamp_to_arrow(
candles_array[0][0]), jh.timestamp_to_arrow(candles_array[-1][0])) + 1
if period > 365:
duration = '{} days ({} years)'.format(period, round(period / 365, 2))
elif period > 30:
duration = '{} days ({} months)'.format(period, round(period / 30, 2))
else:
duration = '{} days'.format(period)
return [
['period', duration],
[
'starting-ending date', '{} => {}'.format(
jh.timestamp_to_time(candles_array[0][0])[:10],
jh.timestamp_to_time(candles_array[-1][0] + 60_000)[:10])
],
]
def test_timestamp_to_arrow():
arrow_time = arrow.get('2015-08-01')
assert jh.timestamp_to_arrow(1438387200000) == arrow_time
def generate_initial_population(self) -> None:
"""
generates the initial population
"""
loop_length = int(self.population_size / self.cpu_cores)
with click.progressbar(length=loop_length, label='Generating initial population...') as progressbar:
for i in range(loop_length):
people = []
with Manager() as manager:
dna_bucket = manager.list([])
workers = []
def get_fitness(dna: str, dna_bucket: list) -> None:
try:
fitness_score, fitness_log_training, fitness_log_testing = self.fitness(dna)
dna_bucket.append((dna, fitness_score, fitness_log_training, fitness_log_testing))
except Exception as e:
proc = os.getpid()
logger.error(f'process failed - ID: {str(proc)}')
logger.error("".join(traceback.TracebackException.from_exception(e).format()))
raise e
try:
for _ in range(self.cpu_cores):
dna = ''.join(choices(self.charset, k=self.solution_len))
w = Process(target=get_fitness, args=(dna, dna_bucket))
w.start()
workers.append(w)
# join workers
for w in workers:
w.join()
if w.exitcode > 0:
logger.error(f'a process exited with exitcode: {str(w.exitcode)}')
except KeyboardInterrupt:
print(
jh.color('Terminating session...', 'red')
)
# terminate all workers
for w in workers:
w.terminate()
# shutdown the manager process manually since garbage collection cannot won't get to do it for us
manager.shutdown()
# now we can terminate the main session safely
jh.terminate_app()
except:
raise
for d in dna_bucket:
people.append({
'dna': d[0],
'fitness': d[1],
'training_log': d[2],
'testing_log': d[3]
})
# update dashboard
click.clear()
progressbar.update(1)
print('\n')
table_items = [
['Started at', jh.timestamp_to_arrow(self.start_time).humanize()],
['Index', f'{len(self.population)}/{self.population_size}'],
['errors/info', f'{len(store.logs.errors)}/{len(store.logs.info)}'],
['Trading Route', f'{router.routes[0].exchange}, {router.routes[0].symbol}, {router.routes[0].timeframe}, {router.routes[0].strategy_name}'],
# TODO: add generated DNAs?
# ['-'*10, '-'*10],
# ['DNA', people[0]['dna']],
# ['fitness', round(people[0]['fitness'], 6)],
# ['training|testing logs', people[0]['log']],
]
if jh.is_debugging():
table_items.insert(3, ['Population Size', self.population_size])
table_items.insert(3, ['Iterations', self.iterations])
table_items.insert(3, ['Solution Length', self.solution_len])
table_items.insert(3, ['-' * 10, '-' * 10])
table.key_value(table_items, 'Optimize Mode', alignments=('left', 'right'))
# errors
if jh.is_debugging() and len(report.errors()):
print('\n')
table.key_value(report.errors(), 'Error Logs')
for p in people:
self.population.append(p)
# sort the population
self.population = list(sorted(self.population, key=lambda x: x['fitness'], reverse=True))
def evolve(self) -> List[Any]:
"""
the main method, that runs the evolutionary algorithm
"""
# generate the population if starting
if self.started_index == 0:
self.generate_initial_population()
if len(self.population) < 0.5 * self.population_size:
raise ValueError('Too many errors: less then half of the planned population size could be generated.')
# if even our best individual is too weak, then we better not continue
if self.population[0]['fitness'] == 0.0001:
print(jh.color('Cannot continue because no individual with the minimum fitness-score was found. '
'Your strategy seems to be flawed or maybe it requires modifications. ', 'yellow'))
jh.terminate_app()
loop_length = int(self.iterations / self.cpu_cores)
i = self.started_index
with click.progressbar(length=loop_length, label='Evolving...') as progressbar:
while i < loop_length:
with Manager() as manager:
people = manager.list([])
workers = []
def get_baby(people: List) -> None:
try:
# let's make a baby together LOL
baby = self.make_love()
# let's mutate baby's genes, who knows, maybe we create a x-man or something
baby = self.mutate(baby)
people.append(baby)
except Exception as e:
proc = os.getpid()
logger.error(f'process failed - ID: {str(proc)}')
logger.error("".join(traceback.TracebackException.from_exception(e).format()))
raise e
try:
for _ in range(self.cpu_cores):
w = Process(target=get_baby, args=[people])
w.start()
workers.append(w)
for w in workers:
w.join()
if w.exitcode > 0:
logger.error(f'a process exited with exitcode: {str(w.exitcode)}')
except KeyboardInterrupt:
print(
jh.color('Terminating session...', 'red')
)
# terminate all workers
for w in workers:
w.terminate()
# shutdown the manager process manually since garbage collection cannot won't get to do it for us
manager.shutdown()
# now we can terminate the main session safely
jh.terminate_app()
except:
raise
# update dashboard
click.clear()
progressbar.update(1)
print('\n')
table_items = [
['Started At', jh.timestamp_to_arrow(self.start_time).humanize()],
['Index/Total', f'{(i + 1) * self.cpu_cores}/{self.iterations}'],
['errors/info', f'{len(store.logs.errors)}/{len(store.logs.info)}'],
['Route', f'{router.routes[0].exchange}, {router.routes[0].symbol}, {router.routes[0].timeframe}, {router.routes[0].strategy_name}']
]
if jh.is_debugging():
table_items.insert(
3,
['Population Size, Solution Length',
f'{self.population_size}, {self.solution_len}']
)
table.key_value(table_items, 'info', alignments=('left', 'right'))
# errors
if jh.is_debugging() and len(report.errors()):
print('\n')
table.key_value(report.errors(), 'Error Logs')
print('\n')
print('Best DNA candidates:')
print('\n')
# print fittest individuals
if jh.is_debugging():
fittest_list = [['Rank', 'DNA', 'Fitness', 'Training log || Testing log'], ]
else:
fittest_list = [['Rank', 'DNA', 'Training log || Testing log'], ]
if self.population_size > 50:
number_of_ind_to_show = 15
elif self.population_size > 20:
number_of_ind_to_show = 10
elif self.population_size > 9:
number_of_ind_to_show = 9
else:
raise ValueError('self.population_size cannot be less than 10')
for j in range(number_of_ind_to_show):
log = f"win-rate: {self.population[j]['training_log']['win-rate']}%, total: {self.population[j]['training_log']['total']}, PNL: {self.population[j]['training_log']['PNL']}% || win-rate: {self.population[j]['testing_log']['win-rate']}%, total: {self.population[j]['testing_log']['total']}, PNL: {self.population[j]['testing_log']['PNL']}%"
if self.population[j]['testing_log']['PNL'] is not None and self.population[j]['training_log'][
'PNL'] > 0 and self.population[j]['testing_log'][
'PNL'] > 0:
log = jh.style(log, 'bold')
if jh.is_debugging():
fittest_list.append(
[
j + 1,
self.population[j]['dna'],
self.population[j]['fitness'],
log
],
)
else:
fittest_list.append(
[
j + 1,
self.population[j]['dna'],
log
],
)
if jh.is_debugging():
table.multi_value(fittest_list, with_headers=True, alignments=('left', 'left', 'right', 'left'))
else:
table.multi_value(fittest_list, with_headers=True, alignments=('left', 'left', 'left'))
# one person has to die and be replaced with the newborn baby
for baby in people:
random_index = randint(1, len(self.population) - 1) # never kill our best perforemr
try:
self.population[random_index] = baby
except IndexError:
print('=============')
print(f'self.population_size: {self.population_size}')
print(f'self.population length: {len(self.population)}')
jh.terminate_app()
self.population = list(sorted(self.population, key=lambda x: x['fitness'], reverse=True))
# reaching the fitness goal could also end the process
if baby['fitness'] >= self.fitness_goal:
progressbar.update(self.iterations - i)
print('\n')
print(f'fitness goal reached after iteration {i}')
return baby
# save progress after every n iterations
if i != 0 and int(i * self.cpu_cores) % 50 == 0:
self.save_progress(i)
# store a take_snapshot of the fittest individuals of the population
if i != 0 and i % int(100 / self.cpu_cores) == 0:
self.take_snapshot(i * self.cpu_cores)
i += 1
print('\n\n')
print(f'Finished {self.iterations} iterations.')
return self.population
def livetrade() -> List[Union[List[Union[str, Any]], List[str], List[Union[
str, int]], List[Union[str, Dict[str, Union[str, int]], Dict[
str, str], Dict[str, bool], Dict[str, Union[Dict[str, Union[
int, str, List[Dict[str, Union[str, int]]]]], Dict[str, Union[
float, str, int, List[Dict[str, Union[str, int]]]]]]], Dict[
str, int]]]]]:
# sum up balance of all trading exchanges
starting_balance = 0
current_balance = 0
for e in store.exchanges.storage:
starting_balance += store.exchanges.storage[e].starting_assets[
jh.app_currency()]
current_balance += store.exchanges.storage[e].assets[jh.app_currency()]
starting_balance = round(starting_balance, 2)
current_balance = round(current_balance, 2)
arr = [[
'started at',
jh.timestamp_to_arrow(store.app.starting_time).humanize()
], ['current time',
jh.timestamp_to_time(jh.now_to_timestamp())[:19]],
['errors/info', f'{len(store.logs.errors)}/{len(store.logs.info)}'],
['active orders',
store.orders.count_all_active_orders()],
['open positions',
store.positions.count_open_positions()]]
# TODO: for now, we assume that we trade on one exchange only. Later, we need to support for more than one exchange at a time
first_exchange = selectors.get_exchange(router.routes[0].exchange)
if first_exchange.type == 'futures':
arr.append([
'started/current balance', f'{starting_balance}/{current_balance}'
])
else:
# loop all trading exchanges
for exchange in selectors.get_all_exchanges():
# loop all assets
for asset_name, asset_balance in exchange.assets.items():
if asset_name == jh.base_asset(router.routes[0].symbol):
current_price = selectors.get_current_price(
router.routes[0].exchange, router.routes[0].symbol)
arr.append([
f'{asset_name}',
f'{round(exchange.available_assets[asset_name], 5)}/{round(asset_balance, 5)} ({jh.format_currency(round(asset_balance * current_price, 2))} { jh.quote_asset(router.routes[0].symbol)})'
])
else:
arr.append([
f'{asset_name}',
f'{round(exchange.available_assets[asset_name], 5)}/{round(asset_balance, 5)}'
])
# short trades summary
if len(store.completed_trades.trades):
df = pd.DataFrame.from_records(
[t.to_dict() for t in store.completed_trades.trades])
total = len(df)
winning_trades = df.loc[df['PNL'] > 0]
losing_trades = df.loc[df['PNL'] < 0]
pnl = round(df['PNL'].sum(), 2)
pnl_percentage = round((pnl / starting_balance) * 100, 2)
arr.append([
'total/winning/losing trades',
f'{total}/{len(winning_trades)}/{len(losing_trades)}'
])
arr.append(['PNL (%)', f'${pnl} ({pnl_percentage}%)'])
if config['app']['debug_mode']:
arr.append(['debug mode', config['app']['debug_mode']])
if config['app']['is_test_driving']:
arr.append(['Test Drive', config['app']['is_test_driving']])
return arr
def evolve(self) -> list:
"""
the main method, that runs the evolutionary algorithm
"""
# clear the logs to start from a clean slate
jh.clear_file('storage/logs/optimize-mode.txt')
logger.log_optimize_mode('Optimization session started')
if self.started_index == 0:
logger.log_optimize_mode(
f"Generating {self.population_size} population size (random DNAs) using {self.cpu_cores} CPU cores"
)
self.generate_initial_population()
if len(self.population) < 0.5 * self.population_size:
msg = f'Too many errors! less than half of the expected population size could be generated. Only {len(self.population)} indviduals from planned {self.population_size} are usable.'
logger.log_optimize_mode(msg)
raise ValueError(msg)
# if even our best individual is too weak, then we better not continue
if self.population[0]['fitness'] == 0.0001:
msg = 'Cannot continue because no individual with the minimum fitness-score was found. Your strategy seems to be flawed or maybe it requires modifications. '
logger.log_optimize_mode(msg)
raise exceptions.InvalidStrategy(msg)
loop_length = int(self.iterations / self.cpu_cores)
i = self.started_index
progressbar = Progressbar(loop_length)
while i < loop_length:
with Manager() as manager:
people_bucket = manager.list([])
workers = []
try:
for _ in range(self.cpu_cores):
mommy = self.select_person()
daddy = self.select_person()
w = Process(
target=create_baby,
args=(
people_bucket, mommy, daddy, self.solution_len, self.charset,
jh.get_config('env.optimization'), router.formatted_routes,
router.formatted_extra_routes,
self.strategy_hp, self.training_candles, self.testing_candles,
self.optimal_total
)
)
w.start()
workers.append(w)
for w in workers:
w.join()
if w.exitcode > 0:
logger.error(f'a process exited with exitcode: {w.exitcode}')
except exceptions.Termination:
self._handle_termination(manager, workers)
# update dashboard
click.clear()
self.update_progressbar(progressbar)
# general_info streams
general_info = {
'started_at': jh.timestamp_to_arrow(self.start_time).humanize(),
'index': f'{(i + 1) * self.cpu_cores}/{self.iterations}',
'errors_info_count': f'{len(store.logs.errors)}/{len(store.logs.info)}',
'trading_route': f'{router.routes[0].exchange}, {router.routes[0].symbol}, {router.routes[0].timeframe}, {router.routes[0].strategy_name}',
'average_execution_seconds': self.average_execution_seconds
}
if jh.is_debugging():
general_info['population_size'] = self.population_size
general_info['iterations'] = self.iterations
general_info['solution_length'] = self.solution_len
sync_publish('general_info', general_info)
if self.population_size > 50:
number_of_ind_to_show = 40
elif self.population_size > 20:
number_of_ind_to_show = 15
elif self.population_size > 9:
number_of_ind_to_show = 9
else:
raise ValueError('self.population_size cannot be less than 10')
best_candidates = [{
'rank': j + 1,
'dna': self.population[j]['dna'],
'fitness': round(self.population[j]['fitness'], 4),
'training_win_rate': self.population[j]['training_log']['win-rate'],
'training_total_trades': self.population[j]['training_log']['total'],
'training_pnl': self.population[j]['training_log']['PNL'],
'testing_win_rate': self.population[j]['testing_log']['win-rate'],
'testing_total_trades': self.population[j]['testing_log']['total'],
'testing_pnl': self.population[j]['testing_log']['PNL'],
} for j in range(number_of_ind_to_show)]
sync_publish('best_candidates', best_candidates)
# one person has to die and be replaced with the newborn baby
for baby in people_bucket:
# never kill our best performer
random_index = randint(1, len(self.population) - 1)
self.population[random_index] = baby
self.population = list(sorted(self.population, key=lambda x: x['fitness'], reverse=True))
# reaching the fitness goal could also end the process
if baby['fitness'] >= self.fitness_goal:
self.update_progressbar(progressbar, finished=True)
sync_publish('alert', {
'message': f'Fitness goal reached after iteration {i*self.cpu_cores}',
'type': 'success'
})
return baby
# TODO: bring back progress resumption
# # save progress after every n iterations
# if i != 0 and int(i * self.cpu_cores) % 50 == 0:
# self.save_progress(i)
# TODO: bring back
# # store a take_snapshot of the fittest individuals of the population
# if i != 0 and i % int(100 / self.cpu_cores) == 0:
# self.take_snapshot(i * self.cpu_cores)
i += 1
sync_publish('alert', {
'message': f"Finished {self.iterations} iterations. Check your best DNA candidates, "
f"if you don't like any of them, try modifying your strategy.",
'type': 'success'
})
return self.population
def generate_initial_population(self) -> None:
"""
generates the initial population
"""
length = int(self.population_size / self.cpu_cores)
progressbar = Progressbar(length)
for i in range(length):
people = []
with Manager() as manager:
dna_bucket = manager.list([])
workers = []
try:
for _ in range(self.cpu_cores):
dna = ''.join(choices(self.charset, k=self.solution_len))
w = Process(
target=get_and_add_fitness_to_the_bucket,
args=(
dna_bucket, jh.get_config('env.optimization'), router.formatted_routes, router.formatted_extra_routes,
self.strategy_hp, dna, self.training_candles, self.testing_candles,
self.optimal_total
)
)
w.start()
workers.append(w)
# join workers
for w in workers:
w.join()
if w.exitcode > 0:
logger.error(f'a process exited with exitcode: {w.exitcode}')
except exceptions.Termination:
self._handle_termination(manager, workers)
for d in dna_bucket:
people.append({
'dna': d[0],
'fitness': d[1],
'training_log': d[2],
'testing_log': d[3]
})
# update dashboard
self.update_progressbar(progressbar)
# general_info streams
general_info = {
'started_at': jh.timestamp_to_arrow(self.start_time).humanize(),
'index': f'{(i + 1) * self.cpu_cores}/{self.population_size}',
'errors_info_count': f'{len(store.logs.errors)}/{len(store.logs.info)}',
'trading_route': f'{router.routes[0].exchange}, {router.routes[0].symbol}, {router.routes[0].timeframe}, {router.routes[0].strategy_name}',
'average_execution_seconds': self.average_execution_seconds
}
if jh.is_debugging():
general_info['population_size'] = self.population_size
general_info['iterations'] = self.iterations
general_info['solution_length'] = self.solution_len
sync_publish('general_info', general_info)
for p in people:
self.population.append(p)
sync_publish('progressbar', {
'current': 100,
'estimated_remaining_seconds': 0
})
# sort the population
self.population = list(sorted(self.population, key=lambda x: x['fitness'], reverse=True))
def _get_candles_from_backup_exchange(exchange: str,
backup_driver: CandleExchange,
symbol: str, start_timestamp: int,
end_timestamp: int):
total_candles = []
# try fetching from database first
backup_candles = Candle.select(
Candle.timestamp, Candle.open, Candle.close, Candle.high, Candle.low,
Candle.volume).where(
Candle.timestamp.between(start_timestamp, end_timestamp),
Candle.exchange == backup_driver.name,
Candle.symbol == symbol).order_by(Candle.timestamp.asc()).tuples()
already_exists = len(
backup_candles) == (end_timestamp - start_timestamp) / 60_000 + 1
if already_exists:
# loop through them and set new ID and exchange
for c in backup_candles:
total_candles.append({
'id': jh.generate_unique_id(),
'symbol': symbol,
'exchange': exchange,
'timestamp': c[0],
'open': c[1],
'close': c[2],
'high': c[3],
'low': c[4],
'volume': c[5]
})
return total_candles
# try fetching from market now
days_count = jh.date_diff_in_days(jh.timestamp_to_arrow(start_timestamp),
jh.timestamp_to_arrow(end_timestamp))
# make sure it's rounded up so that we import maybe more candles, but not less
if days_count < 1:
days_count = 1
if type(days_count) is float and not days_count.is_integer():
days_count = math.ceil(days_count)
candles_count = days_count * 1440
start_date = jh.timestamp_to_arrow(start_timestamp).floor('day')
for _ in range(candles_count):
temp_start_timestamp = start_date.int_timestamp * 1000
temp_end_timestamp = temp_start_timestamp + (backup_driver.count -
1) * 60000
# to make sure it won't try to import candles from the future! LOL
if temp_start_timestamp > jh.now_to_timestamp():
break
# prevent duplicates
count = Candle.select().where(
Candle.timestamp.between(temp_start_timestamp, temp_end_timestamp),
Candle.symbol == symbol,
Candle.exchange == backup_driver.name).count()
already_exists = count == backup_driver.count
if not already_exists:
# it's today's candles if temp_end_timestamp < now
if temp_end_timestamp > jh.now_to_timestamp():
temp_end_timestamp = arrow.utcnow().floor(
'minute').int_timestamp * 1000 - 60000
# fetch from market
candles = backup_driver.fetch(symbol, temp_start_timestamp)
if not len(candles):
raise CandleNotFoundInExchange(
'No candles exists in the market for this day: {} \n'
'Try another start_date'.format(
jh.timestamp_to_time(temp_start_timestamp)[:10], ))
# fill absent candles (if there's any)
candles = _fill_absent_candles(candles, temp_start_timestamp,
temp_end_timestamp)
# store in the database
_insert_to_database(candles)
# add as much as driver's count to the temp_start_time
start_date = start_date.shift(minutes=backup_driver.count)
# sleep so that the exchange won't get angry at us
if not already_exists:
time.sleep(backup_driver.sleep_time)
# now try fetching from database again. Why? because we might have fetched more
# than what's needed, but we only want as much was requested. Don't worry, the next
# request will probably fetch from database and there won't be any waste!
backup_candles = Candle.select(
Candle.timestamp, Candle.open, Candle.close, Candle.high, Candle.low,
Candle.volume).where(
Candle.timestamp.between(start_timestamp, end_timestamp),
Candle.exchange == backup_driver.name,
Candle.symbol == symbol).order_by(Candle.timestamp.asc()).tuples()
already_exists = len(
backup_candles) == (end_timestamp - start_timestamp) / 60_000 + 1
if already_exists:
# loop through them and set new ID and exchange
for c in backup_candles:
total_candles.append({
'id': jh.generate_unique_id(),
'symbol': symbol,
'exchange': exchange,
'timestamp': c[0],
'open': c[1],
'close': c[2],
'high': c[3],
'low': c[4],
'volume': c[5]
})
return total_candles
def load_required_candles(exchange: str, symbol: str, start_date_str: str,
finish_date_str: str):
"""
loads initial candles that required before executing strategies.
210 for the biggest timeframe and more for the rest
"""
start_date = jh.arrow_to_timestamp(arrow.get(start_date_str, 'YYYY-MM-DD'))
finish_date = jh.arrow_to_timestamp(
arrow.get(finish_date_str, 'YYYY-MM-DD')) - 60000
# validate
if start_date == finish_date:
raise ValueError('start_date and finish_date cannot be the same.')
if start_date > finish_date:
raise ValueError('start_date cannot be bigger than finish_date.')
if finish_date > arrow.utcnow().int_timestamp * 1000:
raise ValueError('Can\'t backtest the future!')
max_timeframe = jh.max_timeframe(config['app']['considering_timeframes'])
short_candles_count = jh.get_config(
'env.data.warmup_candles_num',
210) * jh.timeframe_to_one_minutes(max_timeframe)
pre_finish_date = start_date - 60_000
pre_start_date = pre_finish_date - short_candles_count * 60_000
# make sure starting from the beginning of the day instead
pre_start_date = jh.timestamp_to_arrow(pre_start_date).floor(
'day').int_timestamp * 1000
# update candles_count to count from the beginning of the day instead
short_candles_count = int((pre_finish_date - pre_start_date) / 60_000)
key = jh.key(exchange, symbol)
cache_key = '{}-{}-{}'.format(jh.timestamp_to_date(pre_start_date),
jh.timestamp_to_date(pre_finish_date), key)
cached_value = cache.get_value(cache_key)
# if cache exists
if cached_value:
candles_tuple = cached_value
# not cached, get and cache for later calls in the next 5 minutes
else:
# fetch from database
candles_tuple = tuple(
Candle.select(Candle.timestamp, Candle.open, Candle.close,
Candle.high, Candle.low, Candle.volume).where(
Candle.timestamp.between(pre_start_date,
pre_finish_date),
Candle.exchange == exchange,
Candle.symbol == symbol).order_by(
Candle.timestamp.asc()).tuples())
# cache it for near future calls
cache.set_value(cache_key,
candles_tuple,
expire_seconds=60 * 60 * 24 * 7)
candles = np.array(candles_tuple)
if len(candles) < short_candles_count + 1:
first_existing_candle = tuple(
Candle.select(Candle.timestamp).where(
Candle.exchange == exchange, Candle.symbol == symbol).order_by(
Candle.timestamp.asc()).limit(1).tuples())
if not len(first_existing_candle):
raise CandleNotFoundInDatabase(
'No candle for {} {} is present in the database. Try importing candles.'
.format(exchange, symbol))
first_existing_candle = first_existing_candle[0][0]
last_existing_candle = tuple(
Candle.select(Candle.timestamp).where(
Candle.exchange == exchange, Candle.symbol == symbol).order_by(
Candle.timestamp.desc()).limit(1).tuples())[0][0]
first_backtestable_timestamp = first_existing_candle + (
pre_finish_date - pre_start_date) + (60_000 * 1440)
# if first backtestable timestamp is in the future, that means we have some but not enough candles
if first_backtestable_timestamp > jh.today_to_timestamp():
raise CandleNotFoundInDatabase(
'Not enough candle for {} {} is present in the database. Jesse requires "210 * biggest_timeframe" warm-up candles. '
'Try importing more candles from an earlier date.'.format(
exchange, symbol))
raise CandleNotFoundInDatabase(
'Not enough candles for {} {} exists to run backtest from {} => {}. \n'
'First available date is {}\n'
'Last available date is {}'.format(
exchange,
symbol,
start_date_str,
finish_date_str,
jh.timestamp_to_date(first_backtestable_timestamp),
jh.timestamp_to_date(last_existing_candle),
))
return candles
