Python pandas nlargest() not working properly with keep = 'all'

user_16495591 0
Viewed 207

When I try to use the function below

top3 = df1.nlargest(3, 'perChange', keep='all')

Even if keep = 'all', the output is

92     3.828120
255   -0.673854
256   -0.673854
Name: perChange, dtype: float64

However, the sorted array is

92      3.828120
255    -0.673854
256    -0.673854
304    -1.906793
340    -2.643661
355    -3.421462
359    -3.549768

What I want is

92      3.828120
255    -0.673854
256    -0.673854
304    -1.906793

How can I solve the problem?

Edit: For the people who interest to solve it, here is the example code which show the problem.

# initialize data of lists.
data = {'value': [3.828120, -0.673854, -0.673854, -1.906793, -2.643661]}
  
# Create DataFrame
test_df = pd.DataFrame(data)
test_df = test_df.nlargest(3, 'value', keep='all')
# Print the output.
print(test_df['value'])
pandas dataframe python
1 Answers

It seems to be a bug for nlargest().

I write a function that does the job I want. I sorted the df and use the function below to find the index.

def findnlargestI(df,n,col):
    lastValue = 0
    lastRank = 0
    for i in range(0,len(df.index)):
            currentValue = df[col].iloc[i]
            if currentValue != lastValue:
                rank = lastRank + 1
                lastRank = rank
                lastValue = currentValue
            
            if rank > n:
                return i 
top3= df1.iloc[0:findnlargestI(df1,3,'perChange'),:]
user_164955910
Related