brighter-trading/tests/test_DataCache_v2.py

522 lines
24 KiB
Python

import pytz
from DataCache_v2 import DataCache
from ExchangeInterface import ExchangeInterface
import unittest
import pandas as pd
import datetime as dt
import os
from Database import SQLite, Database
class DataGenerator:
def __init__(self, timeframe_str):
"""
Initialize the DataGenerator with a timeframe string like '2h', '5m', '1d', '1w', '1M', or '1y'.
"""
# Initialize attributes with placeholder values
self.timeframe_amount = None
self.timeframe_unit = None
# Set the actual timeframe
self.set_timeframe(timeframe_str)
def set_timeframe(self, timeframe_str):
"""
Set the timeframe unit and amount based on a string like '2h', '5m', '1d', '1w', '1M', or '1y'.
"""
self.timeframe_amount = int(timeframe_str[:-1])
unit = timeframe_str[-1]
if unit == 's':
self.timeframe_unit = 'seconds'
elif unit == 'm':
self.timeframe_unit = 'minutes'
elif unit == 'h':
self.timeframe_unit = 'hours'
elif unit == 'd':
self.timeframe_unit = 'days'
elif unit == 'w':
self.timeframe_unit = 'weeks'
elif unit == 'M':
self.timeframe_unit = 'months'
elif unit == 'Y':
self.timeframe_unit = 'years'
else:
raise ValueError(
"Unsupported timeframe unit. Use 's,m,h,d,w,M,Y'.")
def create_table(self, num_rec=None, start=None, end=None):
"""
Create a table with simulated data. If both start and end are provided, num_rec is derived from the interval.
If neither are provided the table will have num_rec and end at the current time.
Parameters:
num_rec (int, optional): The number of records to generate.
start (datetime, optional): The start time for the first record.
end (datetime, optional): The end time for the last record.
Returns:
pd.DataFrame: A DataFrame with the simulated data.
"""
# Ensure provided datetime parameters are timezone aware
if start and start.tzinfo is None:
raise ValueError('start datetime must be timezone aware.')
if end and end.tzinfo is None:
raise ValueError('end datetime must be timezone aware.')
# If neither start nor end are provided.
if start is None and end is None:
end = dt.datetime.now(dt.timezone.utc)
if num_rec is None:
raise ValueError("num_rec must be provided if both start and end are not specified.")
# If start and end are provided.
if start is not None and end is not None:
total_duration = (end - start).total_seconds()
interval_seconds = self.timeframe_amount * self._get_seconds_per_unit(self.timeframe_unit)
num_rec = int(total_duration // interval_seconds) + 1
# If only end is provided.
if end is not None and start is None:
if num_rec is None:
raise ValueError("num_rec must be provided if both start and end are not specified.")
interval_seconds = self.timeframe_amount * self._get_seconds_per_unit(self.timeframe_unit)
start = end - dt.timedelta(seconds=(num_rec - 1) * interval_seconds)
start = start.replace(tzinfo=pytz.utc)
# Ensure start is aligned to the timeframe interval
start = self.round_down_datetime(start, self.timeframe_unit[0], self.timeframe_amount)
# Generate times
times = [self.unix_time_millis(start + self._delta(i)) for i in range(num_rec)]
df = pd.DataFrame({
'market_id': 1,
'open_time': times,
'open': [100 + i for i in range(num_rec)],
'high': [110 + i for i in range(num_rec)],
'low': [90 + i for i in range(num_rec)],
'close': [105 + i for i in range(num_rec)],
'volume': [1000 + i for i in range(num_rec)]
})
return df
@staticmethod
def _get_seconds_per_unit(unit):
"""Helper method to convert timeframe units to seconds."""
units_in_seconds = {
'seconds': 1,
'minutes': 60,
'hours': 3600,
'days': 86400,
'weeks': 604800,
'months': 2592000, # Assuming 30 days per month
'years': 31536000 # Assuming 365 days per year
}
if unit not in units_in_seconds:
raise ValueError(f"Unsupported timeframe unit: {unit}")
return units_in_seconds[unit]
def generate_incomplete_data(self, query_offset, num_rec=5):
"""
Generate data that is incomplete, i.e., starts before the query but doesn't fully satisfy it.
"""
query_start_time = self.x_time_ago(query_offset)
start_time_for_data = self.get_start_time(query_start_time)
return self.create_table(num_rec, start=start_time_for_data)
@staticmethod
def generate_missing_section(df, drop_start=5, drop_end=8):
"""
Generate data with a missing section.
"""
df = df.drop(df.index[drop_start:drop_end]).reset_index(drop=True)
return df
def get_start_time(self, query_start_time):
margin = 2
delta_args = {self.timeframe_unit: margin * self.timeframe_amount}
return query_start_time - dt.timedelta(**delta_args)
def x_time_ago(self, offset):
"""
Returns a datetime object representing the current time minus the offset in the specified units.
"""
delta_args = {self.timeframe_unit: offset}
return dt.datetime.utcnow().replace(tzinfo=pytz.utc) - dt.timedelta(**delta_args)
def _delta(self, i):
"""
Returns a timedelta object for the ith increment based on the timeframe unit and amount.
"""
delta_args = {self.timeframe_unit: i * self.timeframe_amount}
return dt.timedelta(**delta_args)
@staticmethod
def unix_time_millis(dt_obj: dt.datetime):
"""
Convert a datetime object to Unix time in milliseconds.
"""
if dt_obj.tzinfo is None:
raise ValueError('dt_obj needs to be timezone aware.')
epoch = dt.datetime(1970, 1, 1).replace(tzinfo=pytz.UTC)
return int((dt_obj - epoch).total_seconds() * 1000)
@staticmethod
def round_down_datetime(dt_obj: dt.datetime, unit: str, interval: int) -> dt.datetime:
if dt_obj.tzinfo is None:
raise ValueError('dt_obj needs to be timezone aware.')
if unit == 's': # Round down to the nearest interval of seconds
seconds = (dt_obj.second // interval) * interval
dt_obj = dt_obj.replace(second=seconds, microsecond=0)
elif unit == 'm': # Round down to the nearest interval of minutes
minutes = (dt_obj.minute // interval) * interval
dt_obj = dt_obj.replace(minute=minutes, second=0, microsecond=0)
elif unit == 'h': # Round down to the nearest interval of hours
hours = (dt_obj.hour // interval) * interval
dt_obj = dt_obj.replace(hour=hours, minute=0, second=0, microsecond=0)
elif unit == 'd': # Round down to the nearest interval of days
days = (dt_obj.day // interval) * interval
dt_obj = dt_obj.replace(day=days, hour=0, minute=0, second=0, microsecond=0)
elif unit == 'w': # Round down to the nearest interval of weeks
dt_obj -= dt.timedelta(days=dt_obj.weekday() % (interval * 7))
dt_obj = dt_obj.replace(hour=0, minute=0, second=0, microsecond=0)
elif unit == 'M': # Round down to the nearest interval of months
months = ((dt_obj.month - 1) // interval) * interval + 1
dt_obj = dt_obj.replace(month=months, day=1, hour=0, minute=0, second=0, microsecond=0)
elif unit == 'y': # Round down to the nearest interval of years
years = (dt_obj.year // interval) * interval
dt_obj = dt_obj.replace(year=years, month=1, day=1, hour=0, minute=0, second=0, microsecond=0)
return dt_obj
class TestDataCacheV2(unittest.TestCase):
def setUp(self):
# Set up database and exchanges
self.exchanges = ExchangeInterface()
self.exchanges.connect_exchange(exchange_name='binance', user_name='test_guy', api_keys=None)
self.db_file = 'test_db.sqlite'
self.database = Database(db_file=self.db_file)
# Create necessary tables
sql_create_table_1 = f"""
CREATE TABLE IF NOT EXISTS test_table (
id INTEGER PRIMARY KEY,
market_id INTEGER,
open_time INTEGER UNIQUE ON CONFLICT IGNORE,
open REAL NOT NULL,
high REAL NOT NULL,
low REAL NOT NULL,
close REAL NOT NULL,
volume REAL NOT NULL,
FOREIGN KEY (market_id) REFERENCES market (id)
)"""
sql_create_table_2 = """
CREATE TABLE IF NOT EXISTS exchange (
id INTEGER PRIMARY KEY AUTOINCREMENT,
name TEXT UNIQUE
)"""
sql_create_table_3 = """
CREATE TABLE IF NOT EXISTS markets (
id INTEGER PRIMARY KEY AUTOINCREMENT,
symbol TEXT,
exchange_id INTEGER,
FOREIGN KEY (exchange_id) REFERENCES exchange(id)
)"""
with SQLite(db_file=self.db_file) as con:
con.execute(sql_create_table_1)
con.execute(sql_create_table_2)
con.execute(sql_create_table_3)
self.data = DataCache(self.exchanges)
self.data.db = self.database
self.ex_details = ['BTC/USD', '2h', 'binance', 'test_guy']
self.key = f'{self.ex_details[0]}_{self.ex_details[1]}_{self.ex_details[2]}'
def tearDown(self):
if os.path.exists(self.db_file):
os.remove(self.db_file)
def test_set_cache(self):
print('\nTesting set_cache() method without no-overwrite flag:')
self.data.set_cache(data='data', key=self.key)
attr = self.data.__getattribute__('cached_data')
self.assertEqual(attr[self.key], 'data')
print(' - Set cache without no-overwrite flag passed.')
print('Testing set_cache() once again with new data without no-overwrite flag:')
self.data.set_cache(data='more_data', key=self.key)
attr = self.data.__getattribute__('cached_data')
self.assertEqual(attr[self.key], 'more_data')
print(' - Set cache with new data without no-overwrite flag passed.')
print('Testing set_cache() method once again with more data with no-overwrite flag set:')
self.data.set_cache(data='even_more_data', key=self.key, do_not_overwrite=True)
attr = self.data.__getattribute__('cached_data')
self.assertEqual(attr[self.key], 'more_data')
print(' - Set cache with no-overwrite flag passed.')
def test_cache_exists(self):
print('Testing cache_exists() method:')
self.assertFalse(self.data.cache_exists(key=self.key))
print(' - Check for non-existent cache passed.')
self.data.set_cache(data='data', key=self.key)
self.assertTrue(self.data.cache_exists(key=self.key))
print(' - Check for existent cache passed.')
def test_update_candle_cache(self):
print('Testing update_candle_cache() method:')
# Initialize the DataGenerator with the 5-minute timeframe
data_gen = DataGenerator('5m')
# Create initial DataFrame and insert into cache
df_initial = data_gen.create_table(num_rec=3, start=dt.datetime(2024, 8, 9, 0, 0, 0))
print(f'Inserting this table into cache:\n{df_initial}\n')
self.data.set_cache(data=df_initial, key=self.key)
# Create new DataFrame to be added to cache
df_new = data_gen.create_table(num_rec=3, start=dt.datetime(2024, 8, 9, 0, 15, 0))
print(f'Updating cache with this table:\n{df_new}\n')
self.data.update_candle_cache(more_records=df_new, key=self.key)
# Retrieve the resulting DataFrame from cache
result = self.data.get_cache(key=self.key)
print(f'The resulting table in cache is:\n{result}\n')
# Create the expected DataFrame
expected = data_gen.create_table(num_rec=6, start=dt.datetime(2024, 8, 9, 0, 0, 0))
print(f'The expected open_time values are:\n{expected["open_time"].tolist()}\n')
# Assert that the open_time values in the result match those in the expected DataFrame, in order
assert result['open_time'].tolist() == expected['open_time'].tolist(), \
f"open_time values in result are {result['open_time'].tolist()}" \
f" but expected {expected['open_time'].tolist()}"
print(f'The results open_time values match:\n{result["open_time"].tolist()}\n')
print(' - Update cache with new records passed.')
def test_update_cached_dict(self):
print('Testing update_cached_dict() method:')
self.data.set_cache(data={}, key=self.key)
self.data.update_cached_dict(cache_key=self.key, dict_key='sub_key', data='value')
cache = self.data.get_cache(key=self.key)
self.assertEqual(cache['sub_key'], 'value')
print(' - Update dictionary in cache passed.')
def test_get_cache(self):
print('Testing get_cache() method:')
self.data.set_cache(data='data', key=self.key)
result = self.data.get_cache(key=self.key)
self.assertEqual(result, 'data')
print(' - Retrieve cache passed.')
def _test_get_records_since(self, set_cache=True, set_db=True, query_offset=None, num_rec=None, ex_details=None,
simulate_scenarios=None):
"""
Test the get_records_since() method by generating a table of simulated data,
inserting it into cache and/or database, and then querying the records.
Parameters:
set_cache (bool): If True, the generated table is inserted into the cache.
set_db (bool): If True, the generated table is inserted into the database.
query_offset (int, optional): The offset in the timeframe units for the query.
num_rec (int, optional): The number of records to generate in the simulated table.
ex_details (list, optional): Exchange details to generate the cache key.
simulate_scenarios (str, optional): The type of scenario to simulate. Options are:
- 'not_enough_data': The table data doesn't go far enough back.
- 'incomplete_data': The table doesn't have enough records to satisfy the query.
- 'missing_section': The table has missing records in the middle.
"""
print('Testing get_records_since() method:')
# Use provided ex_details or fallback to the class attribute.
ex_details = ex_details or self.ex_details
# Generate a cache/database key using exchange details.
key = f'{ex_details[0]}_{ex_details[1]}_{ex_details[2]}'
# Set default number of records if not provided.
num_rec = num_rec or 12
table_timeframe = ex_details[1] # Extract timeframe from exchange details.
# Initialize DataGenerator with the given timeframe.
data_gen = DataGenerator(table_timeframe)
if simulate_scenarios == 'not_enough_data':
# Set query_offset to a time earlier than the start of the table data.
query_offset = (num_rec + 5) * data_gen.timeframe_amount
else:
# Default to querying for 1 record length less than the table duration.
query_offset = query_offset or (num_rec - 1) * data_gen.timeframe_amount
if simulate_scenarios == 'incomplete_data':
# Set start time to generate fewer records than required.
start_time_for_data = data_gen.x_time_ago(num_rec * data_gen.timeframe_amount)
num_rec = 5 # Set a smaller number of records to simulate incomplete data.
else:
# No specific start time for data generation.
start_time_for_data = None
# Create the initial data table.
df_initial = data_gen.create_table(num_rec, start=start_time_for_data)
if simulate_scenarios == 'missing_section':
# Simulate missing section in the data by dropping records.
df_initial = data_gen.generate_missing_section(df_initial, drop_start=2, drop_end=5)
# Convert 'open_time' to datetime for better readability.
temp_df = df_initial.copy()
temp_df['open_time'] = pd.to_datetime(temp_df['open_time'], unit='ms')
print(f'Table Created:\n{temp_df}')
if set_cache:
# Insert the generated table into cache.
print('Inserting table into cache.')
self.data.set_cache(data=df_initial, key=key)
if set_db:
# Insert the generated table into the database.
print('Inserting table into database.')
with SQLite(self.db_file) as con:
df_initial.to_sql(key, con, if_exists='replace', index=False)
# Calculate the start time for querying the records.
start_datetime = data_gen.x_time_ago(query_offset)
# Ensure start_datetime is timezone-aware (UTC).
if start_datetime.tzinfo is None:
start_datetime = start_datetime.replace(tzinfo=dt.timezone.utc)
# Defaults to current time if not provided to get_records_since()
query_end_time = dt.datetime.utcnow().replace(tzinfo=dt.timezone.utc)
print(f'Requesting records from {start_datetime} to {query_end_time}')
# Query the records since the calculated start time.
result = self.data.get_records_since(start_datetime=start_datetime, ex_details=ex_details)
# Filter the initial data table to match the query time.
expected = df_initial[df_initial['open_time'] >= data_gen.unix_time_millis(start_datetime)].reset_index(
drop=True)
temp_df = expected.copy()
temp_df['open_time'] = pd.to_datetime(temp_df['open_time'], unit='ms')
print(f'Expected table:\n{temp_df}')
# Print the result from the query for comparison.
temp_df = result.copy()
temp_df['open_time'] = pd.to_datetime(temp_df['open_time'], unit='ms')
print(f'Resulting table:\n{temp_df}')
if simulate_scenarios in ['not_enough_data', 'incomplete_data', 'missing_section']:
# Check that the result has more rows than the expected incomplete data.
assert result.shape[0] > expected.shape[
0], "Result has fewer or equal rows compared to the incomplete data."
print("\nThe returned DataFrames has filled in the missing data!")
else:
# Ensure the result and expected dataframes match in shape and content.
assert result.shape == expected.shape, f"Shape mismatch: {result.shape} vs {expected.shape}"
pd.testing.assert_series_equal(result['open_time'], expected['open_time'], check_dtype=False)
print("\nThe DataFrames have the same shape and the 'open_time' columns match.")
# Verify that the oldest timestamp in the result is within the allowed time difference.
oldest_timestamp = pd.to_datetime(result['open_time'].min(), unit='ms').tz_localize('UTC')
time_diff = oldest_timestamp - start_datetime
max_allowed_time_diff = dt.timedelta(**{data_gen.timeframe_unit: data_gen.timeframe_amount})
assert dt.timedelta(0) <= time_diff <= max_allowed_time_diff, \
f"Oldest timestamp {oldest_timestamp} is not within " \
f"{data_gen.timeframe_amount} {data_gen.timeframe_unit} of {start_datetime}"
print(f'The first timestamp is {time_diff} from {start_datetime}')
# Verify that the newest timestamp in the result is within the allowed time difference.
newest_timestamp = pd.to_datetime(result['open_time'].max(), unit='ms').tz_localize('UTC')
time_diff_end = abs(query_end_time - newest_timestamp)
assert dt.timedelta(0) <= time_diff_end <= max_allowed_time_diff, \
f"Newest timestamp {newest_timestamp} is not within {data_gen.timeframe_amount} " \
f"{data_gen.timeframe_unit} of {query_end_time}"
print(f'The last timestamp is {time_diff_end} from {query_end_time}')
print(' - Fetch records within the specified time range passed.')
def test_get_records_since(self):
print('\nTest get_records_since with records set in cache')
self._test_get_records_since()
print('\nTest get_records_since with records not in cache')
self._test_get_records_since(set_cache=False)
print('\nTest get_records_since with records not in database')
self._test_get_records_since(set_cache=False, set_db=False)
print('\nTest get_records_since with a different timeframe')
self._test_get_records_since(query_offset=None, num_rec=None,
ex_details=['BTC/USD', '15m', 'binance', 'test_guy'])
print('\nTest get_records_since where data does not go far enough back')
self._test_get_records_since(simulate_scenarios='not_enough_data')
print('\nTest get_records_since with incomplete data')
self._test_get_records_since(simulate_scenarios='incomplete_data')
print('\nTest get_records_since with missing section in data')
self._test_get_records_since(simulate_scenarios='missing_section')
def test_other_timeframes(self):
# print('\nTest get_records_since with a different timeframe')
# ex_details = ['BTC/USD', '15m', 'binance', 'test_guy']
# start_datetime = dt.datetime.now(dt.timezone.utc) - dt.timedelta(hours=2)
# # Query the records since the calculated start time.
# result = self.data.get_records_since(start_datetime=start_datetime, ex_details=ex_details)
# last_record_time = pd.to_datetime(result['open_time'].max(), unit='ms').tz_localize('UTC')
# assert last_record_time > dt.datetime.now(dt.timezone.utc) - dt.timedelta(minutes=15.1)
#
# print('\nTest get_records_since with a different timeframe')
# ex_details = ['BTC/USD', '5m', 'binance', 'test_guy']
# start_datetime = dt.datetime.now(dt.timezone.utc) - dt.timedelta(hours=1)
# # Query the records since the calculated start time.
# result = self.data.get_records_since(start_datetime=start_datetime, ex_details=ex_details)
# last_record_time = pd.to_datetime(result['open_time'].max(), unit='ms').tz_localize('UTC')
# assert last_record_time > dt.datetime.now(dt.timezone.utc) - dt.timedelta(minutes=5.1)
print('\nTest get_records_since with a different timeframe')
ex_details = ['BTC/USD', '4h', 'binance', 'test_guy']
start_datetime = dt.datetime.now(dt.timezone.utc) - dt.timedelta(hours=12)
# Query the records since the calculated start time.
result = self.data.get_records_since(start_datetime=start_datetime, ex_details=ex_details)
last_record_time = pd.to_datetime(result['open_time'].max(), unit='ms').tz_localize('UTC')
assert last_record_time > dt.datetime.now(dt.timezone.utc) - dt.timedelta(hours=4.1)
def test_populate_db(self):
print('Testing _populate_db() method:')
# Create a table of candle records.
data_gen = DataGenerator(self.ex_details[1])
data = data_gen.create_table(num_rec=5)
self.data._populate_db(ex_details=self.ex_details, data=data)
with SQLite(self.db_file) as con:
result = pd.read_sql(f'SELECT * FROM "{self.key}"', con)
self.assertFalse(result.empty)
print(' - Populate database with data passed.')
def test_fetch_candles_from_exchange(self):
print('Testing _fetch_candles_from_exchange() method:')
start_time = dt.datetime.utcnow() - dt.timedelta(days=1)
end_time = dt.datetime.utcnow()
result = self.data._fetch_candles_from_exchange(symbol='BTC/USD', interval='2h', exchange_name='binance',
user_name='test_guy', start_datetime=start_time,
end_datetime=end_time)
self.assertIsInstance(result, pd.DataFrame)
self.assertFalse(result.empty)
print(' - Fetch candle data from exchange passed.')
if __name__ == '__main__':
unittest.main()