#!/usr/bin/env python
# coding: utf-8

# This notebook is designed for ray integrations for signals
# 
# lets load askbid candlesticks file.

# In[20]:


import market_trade.constants
import market_trade.dataloader

candlesticks_filepaths = [filepath for filepath in market_trade.constants.CANDLESTICK_DATASETS_PATH.iterdir()]
candlesticks_filepath = candlesticks_filepaths[0]
duka_interface = market_trade.dataloader.DukaMTInterface(candlesticks_filepath)
duka_interface.bid_candlesticks


# Let's test signal.

# In[21]:


import market_trade.signals.Signal1

bid_candlesticks_df = duka_interface.bid_candlesticks[:10000]

ind_params = {'MeanType': 'SMA', 'window': 5, 'valueType': 'low', 'kDev': 2}
indEl1 = {
    'df': bid_candlesticks_df,
    'params': ind_params,
    'needFig': False,
    'showOnlyIndex': False,
    'drawFig': True
}
signal_result = market_trade.signals.Signal1.SignalBollingerBands1({'BB': indEl1})
signal_result.analiz


# Now let's design ray trainable.

# In[22]:


def trainable(config):
    ind_params = {'MeanType': 'SMA',
                  'window': config['window'],
                  'valueType': config['value_type'],
                  'kDev': config['k_dev']}
    indEl1 = {
        'df': bid_candlesticks_df,
        'params': ind_params,
        'needFig': False,
        'showOnlyIndex': False,
        'drawFig': True
    }
    signal_result = market_trade.signals.Signal1.SignalBollingerBands1({'BB': indEl1})
    tune.report(accuracy=signal_result.analiz["toch"])


# Let's create config space.

# In[23]:


from ray import tune

config = {
    'window': tune.qrandint(5,100,5),
    'value_type': tune.choice(['open', 'low', 'high', 'close']),
    'k_dev': tune.uniform(1,4)
}


# Let's run ray tune

# In[24]:


analysis = tune.run(trainable, config=config, num_samples=100, mode='max', metric='accuracy')


# Let's see who was the best

# In[31]:


analysis.best_config