import pandas as pd

import datetime as dt

from oanda_api import OandaAPI

import json

PAIRS = ['GBP_USD', 'GBP_CAD', 'GBP_JPY', 'GBP_NZD',

'GBP_CHF', 'EUR_GBP', 'EUR_USD', 'EUR_CAD', 'EUR_JPY',

'EUR_NZD', 'EUR_CHF', 'USD_CAD', 'USD_JPY', 'USD_CHF',

'CAD_JPY', 'CAD_CHF', 'NZD_USD', 'NZD_CAD', 'NZD_JPY',

'NZD_CHF', 'CHF_JPY']

DF_COLS = ['PAIR', 'time', 'volume', 'mid_o', 'mid_h', 'mid_l', 'mid_c', 'MACD_CROSS', 'BB_SIGNAL',

'BODY_PERC', 'DIRECTION', 'HAMMER', 'DOJI', 'MARUBOZU',

'SPINNING_TOP', 'ENGULFING', 'TREND']

SMALL_BODY = 0.20

CLOSE_DISTANCE = 0.15

DOJI_BODY = 0.05

FULL_BODY = 0.95

CENTER_DISTANCE_HIGH = 0.55

CENTER_DISTANCE_LOW = 0.45

BIG_BODY = 0.6

ENGULFING_FACTOR = 1.1

def apply_top_end_distance(row):

if row.DIRECTION == 1:

return row.mid_h - row.mid_c

else:

return row.mid_h - row.mid_o

def apply_bottom_end_distance(row):

if row.DIRECTION == 1:

return row.mid_o - row.mid_l

else:

return row.mid_c - row.mid_l

def apply_hammer(row):

if row.BODY_PERC < SMALL_BODY:

if row.DIST_TOP_PERC < CLOSE_DISTANCE or row.DIST_BOTTOM_PERC < CLOSE_DISTANCE:

return True

return False

def apply_spinning_top(row):

if row.BODY_PERC < SMALL_BODY:

if row.DIST_TOP_PERC < CENTER_DISTANCE_HIGH and row.DIST_TOP_PERC > CENTER_DISTANCE_LOW:

return True

return False

def engulfing(row):

if row.PREV_DIRECTION != row.DIRECTION:

if row.PREV_BODY_PERC > BIG_BODY and row.BODY_PERC > BIG_BODY:

if row.BODY_RANGE > row.PREV_BODY_RANGE * ENGULFING_FACTOR:

return True

return False

def apply_stats(df):

df['RANGE'] = df.mid_h - df.mid_l

df['BODY_RANGE'] = abs(df.mid_c - df.mid_o)

df['CENTER'] = (df.mid_h - df.mid_l) / 2 + df.mid_l

df['BODY_PERC'] = df.BODY_RANGE / df.RANGE

df['DIRECTION'] = df.mid_c - df.mid_o

df['DIRECTION'] = df['DIRECTION'].apply(lambda x: 1 if x >= 0 else -1)

df['DIST_TOP'] = df.apply(apply_top_end_distance, axis=1)

df['DIST_BOTTOM'] = df.apply(apply_bottom_end_distance, axis=1)

df['DIST_TOP_PERC'] = df.DIST_TOP / df.RANGE

df['DIST_BOTTOM_PERC'] = df.DIST_BOTTOM / df.RANGE

df['PREV_BODY_RANGE'] = df.BODY_RANGE.shift(1)

df['PREV_BODY_PERC'] = df.BODY_PERC.shift(1)

df['PREV_DIRECTION'] = df.DIRECTION.shift(1)

df.dropna(inplace=True)

return df

def apply_patterns(df):

df['HAMMER'] = df.apply(apply_hammer, axis=1)

df['DOJI'] = df['BODY_PERC'].apply(lambda x: True if x < DOJI_BODY else False)

df['MARUBOZU'] = df['BODY_PERC'].apply(lambda x: True if x > FULL_BODY else False)

df['SPINNING_TOP'] = df.apply(apply_spinning_top, axis=1)

df['ENGULFING'] = df.apply(engulfing, axis=1)

return df

def is_cross(row, val_now, val_prev):

if row[val_now] >= 0 and row[val_prev] < 0:

return 1

if row[val_now] <= 0 and row[val_prev] > 0:

return -1

return 0

def is_trend(row):

if row.EMA_SM > row.EMA_MD and row.EMA_MD > row.EMA_LN:

return 1

if row.EMA_SM < row.EMA_MD and row.EMA_MD < row.EMA_LN:

return -1

return 0

def Get_BB_Signal(row):

if row.mid_c > row.BB_UPPER and row.mid_c > row.mid_o:

return 1

if row.mid_c < row.BB_LOWER and row.mid_c < row.mid_o:

return 1

return 0

def BBands(df):

df['TP'] = (df.mid_h + df.mid_l + df.mid_c) / 3

df['MA_TP'] = df.TP.rolling(window=20).mean()

df['MA_STD'] = df.TP.rolling(window=20).std()

df['BB_UPPER'] = df.MA_TP + 2 * df.MA_STD

df['BB_LOWER'] = df.MA_TP - 2 * df.MA_STD

df['BB_SIGNAL'] = df.apply(Get_BB_Signal, axis=1)

return df

def Macd(df):

df['EMA_LONG'] = df.mid_c.ewm(span=26, min_periods=26).mean()

df['EMA_SHORT'] = df.mid_c.ewm(span=12, min_periods=12).mean()

df['MACD'] = df.EMA_SHORT - df.EMA_LONG

df['SIGNAL'] = df.MACD.ewm(span=9).mean()

df['HIST'] = df.MACD - df.SIGNAL

df['PREV_HIST'] = df.HIST.shift(1)

df['MACD_CROSS'] = df.apply(is_cross, val_now='HIST', val_prev='PREV_HIST', axis=1)

return df

def apply_trend(df):

df[f'EMA_SM'] = df.mid_c.ewm(span=8, min_periods=8).mean()

df[f'EMA_MD'] = df.mid_c.ewm(span=20, min_periods=20).mean()

df[f'EMA_LN'] = df.mid_c.ewm(span=50, min_periods=50).mean()

df['TREND'] = df.apply(is_trend, axis=1)

return df

def get_pair_data(pair, api):

code, df = api.fetch_candles(pair, granularity='M1')

df['PAIR'] = pair

if code == 200:

df = Macd(df)

df = BBands(df)

df = apply_stats(df)

df = apply_patterns(df)

df = apply_trend(df)

df.dropna(inplace=True)

df = df[DF_COLS]

return df.iloc[-1:]

return None

def prepare_data():

api = OandaAPI()

data = []

for p in PAIRS:

row = get_pair_data(p, api)

if row is not None:

data.append(row)

final_df = pd.concat(data)

final_df['time'] = [dt.datetime.strftime(x, "%Y-%m-%d %H:%M:%S") for x in final_df.time]

data_api = {

'price_data' : final_df.to_dict(orient='records'),

'updated' : dt.datetime.strftime(dt.datetime.utcnow(), "%Y-%m-%d %H:%M:%S")

}

return json.dumps(data_api, indent=2)

def save_file():

api_data = prepare_data()

with open('data.json', 'w') as f:

f.write(api_data)

if __name__ == "__main__":

save_file()