An elegant way to make transformation of something like transpose in pandas faster

Question

I have a pandas.Dataframe called a and the structure is as follows:

while I want to get the DataFrame structure is like:

where the b is like the transpose of a . By convert a to b , I use the code :

id_uni = a['id'].unique()
b = pd.DataFrame(columns=['id']+[str(i) for i in range(1,4)])
b['id'] = id_uni

for i in id_uni:
    for j in range(7):
        ind = (a['id'] == i) & (a['w'] == j)
        med = a.loc[ind, 't'].values 
        if med:  
            b.loc[b['id'] == i, str(j)] = med[0]
        else:
            b.loc[b['id'] == i, str(j)] = 0

The method is very brutal that I just use two for-loops to get all elements from a to b . And it is very slow. Do you have an efficient way to improve it?

Answer 1

You can use pivot :

print (df.pivot(index='id', columns='w', values='t'))
w    1    2   3
id             
0   54  147  12
1    1    0   1

---

df1 = df.pivot(index='id', columns='w', values='t').reset_index()
df1.columns.name=None
print (df1)
   id   1    2   3
0   0  54  147  12
1   1   1    0   1