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

# # Tips & Trics for anypytools

# ## Stopping and restarting simulations

# AnyPyProcess caches the simulation results. This allows us to stop the simulations, and then later restart them again.

# In[1]:


from anypytools import AnyPyProcess 
app = AnyPyProcess(num_processes=1)
macro = [
    'load "Knee.any"', 
    'operation Main.MyStudy.InverseDynamics',
    'run',
]

macrolist = [macro]*20


# In[2]:


app.start_macro(macrolist); 


# Here we stopped the simulation using the Notebook interrupt button. Calling the `start_macro()` function again continues the processing and re-run any task that wasn't completed in the first run and any task that exited with errors. 

# In[3]:


app.start_macro(macrolist);


# Note: That changing the input arguments `start_macro` or re-instanciating the `app` object will erase the cache and re-run all processes. 
# 

# ## Including meta-information in the output

# The `start_macro()` also returns extra meta info, but the information is not printed by the default `__repr__()` function.

# In[4]:


from anypytools import AnyPyProcess
from anypytools.macro_commands import Load, OperationRun, Dump

app = AnyPyProcess()
macro = [
    Load('Knee.any', defs={'N_STEP':10}), 
    OperationRun('Main.MyStudy.InverseDynamics'),
    Dump('Main.MyStudy.Output.MaxMuscleActivity'),
]
result = app.start_macro(macro)[0]
result


# But the information is there

# In[5]:


result["task_macro"]


# We can also see all task information by evaluating the result object as standard Python dictionary: 

# In[6]:


dict(result)


# ## Saving output to re-process at a later time

# The extra task meta info gives other posibilities. The results from running batch processing (i.e. output f `start_macro()` can be used as input to restart the same processing even if the AnyPyProcess have no cached results. 

# In[7]:


from anypytools import AnyPyProcess 

app = AnyPyProcess()
macro = [
    Load('Knee.any', defs={'N_STEP':10}), 
    OperationRun('Main.MyStudy.InverseDynamics'),
    Dump('Main.MyStudy.Output.MaxMuscleActivity'),
]
output = app.start_macro(macro) 


# In[8]:


app = AnyPyProcess()
app.start_macro(output)


# The effect is that the result of an analysis can be saved to files and later restarted.  The next example illustrates this.

# ## Example: Saving data to disk while running

# In[9]:


import os
from scipy.stats import distributions
from anypytools import AnyPyProcess, AnyMacro
from anypytools.macro_commands import Load, SetValue_random, OperationRun, Dump


# In[10]:


tibia_knee_srel = distributions.norm([0, 0.18, 0], [0.005, 0.005, 0.005] ) 
femur_knee_srel = distributions.norm([0, -0.3, 0], [0.005, 0.005, 0.005] ) 

app = AnyPyProcess(silent=True)
mg = AnyMacro(number_of_macros = 500)
mg.extend([
    Load('knee.any', defs = {'N_STEP':20}),
    SetValue_random('Main.MyModel.Tibia.Knee.sRel', tibia_knee_srel),
    SetValue_random('Main.MyModel.Femur.Knee.sRel', femur_knee_srel),
    OperationRun('Main.MyStudy.InverseDynamics'),
    Dump('Main.MyStudy.Output.MaxMuscleActivity'),
])

try:
    os.remove('data.db')
except OSError:
    pass

for macros in mg.create_macros_MonteCarlo(batch_size=50):
    app.start_macro(macros)
    app.save_results('data.db', append=True)
    print('Data saved')
print('Done')


# All this stored data can be be reloaded 

# In[11]:


reloaded_results = app.load_results('data.db')
print('Entries in file: {}'.format(len(reloaded_results)))


# In[12]:


reloaded_results[456:457]


# In[13]:


get_ipython().run_line_magic('matplotlib', 'inline')
import matplotlib.pyplot as plt

plt.plot(reloaded_results['MaxMuscleAct'].T, 'b', lw=0.2, alpha = 0.3);


# In[ ]: