In [1]:

    

import pandas as pd
import numpy as np
import karps as ks
import karps.functions as f
from karps.display import show_phase


    

In [2]:

    

def harmonic_mean(col):
    count = f.as_double(f.count(col))
    inv_sum = 1.0/f.sum(1.0/col)
    return inv_sum * count


    

In [3]:

    

df0 = pd.DataFrame([1.0, 2.0])
df0


    

    
Out[3]:







  

    

      

      
0
    
  
  

    

      
0
      
1.0
    
    

      
1
      
2.0
    
  

In [4]:

    

harmonic_mean(df0)


    

    
Out[4]:





1.3333333333333333

In [5]:

    

# Create a HUGE dataframe
df = ks.dataframe([1.0, 2.0], name="my_input")
df


    

    
Out[5]:





/my_input@org.spark.DistributedLiteral:double

In [6]:

    

# And apply our function:
cached_df = f.autocache(df)
hmean = harmonic_mean(cached_df)
hmean


    

    
Out[6]:





/multiply6!org.spark.LocalStructuredTransform:double

In [7]:

    

s = ks.session("demo1e")


    

In [10]:

    

s.eval(hmean)


    

    
Out[10]:





1.3333333333333333

In [ ]:

    

s = ks.session("demo1b")


    

In [11]:

    

s.compute(hmean)


    

    
Out[11]:





<karps.computation.Computation at 0x109815d30>

In [12]:

    

_11.values()


    

    
Out[12]:





(double, double_value: 1.3333333333333333
)

In [ ]:

    

df = pd.concat({'col1':c1, 'col2':c2})
df


    

In [ ]:

    

df = pd.DataFrame([1.0, 2.0])


    

In [ ]:

    

f.count(df)


    

In [ ]:

    

df2 = pd.DataFrame(list(zip(range(5), [x%2 for x in range(1,6)])))
df2


    

In [ ]:

    

z = pd.Series.groupby(df2[0], by=df2[1])
z.max()


    

In [ ]:

    

pd.Series({'x':x})


    

In [ ]:

    

type(df[df.columns[0]])


    

In [ ]:

    

pd.Series.__add__(df[0], df[0])


    

In [ ]:

    

s=(x.sum())
type(s), s


    

In [ ]:

    

np.cast(s, np.float64)


    

In [ ]:

    

x.count()


    

In [ ]: