CODEWITHSUNDEEP
pythonpandas
savchart
savchart

Reputation: 169

Iterate dataframe with window

I have two Dataframes:

value_df = pd.DataFrame({'value' : [10,20,30,40,50,60]})
iter_df = pd.DataFrame({'index_of_value' : [1,1,1,2,2,2], 'sec_value' : [1,2,3,4,5,6]})

I need to transform dataframe by view. I have solution for task where i group by 'index_of_value':

def write(group):
    seq_list = group.sec_value.tolist()
    index = np.unique(group.index_of_value).astype(int)
    seq_list.append(value_df.value.iloc[index[0]])
    seq_list = ','.join(str(e) for e in seq_list)
    return seq_list

result = pd.DataFrame(columns=['seq'])
result['seq'] = iter_df.groupby(['index_of_value']).apply(write)
result = result.reset_index()

But i need instead of groupby split dataframe on N-rows. For example we start from index = 0 with window = 10 and next iteration must start from index = 1 not from 10.

Output for window size = 3

    index_of_value  seq
0          1    1,2,3,20
1          2    2,3,4,20
2          1    3,4,5,20
3          2    4,5,6,30

Upvotes: 1

Views: 641

Answers (1)

alec_djinn
alec_djinn

Reputation: 10779

I use this little function to iterate over a dataframe in windows.

nrows is the window size. You can set the overlap if you want overlapping windows.

def split_df_with_overlap(df, nrows, overlap=0):
    for i in range(0, len(df) - overlap, nrows - overlap):
        yield df.iloc[i : i + nrows]

Upvotes: 2

Related Questions