把柱头变成观察熊猫

Question

I'm not sure if my issue has a specific name (I remember listening to a lecture where the teacher said that a portion of knowledge is knowing the name of things).

Anyway, I'm working with some legacy systems and my data is output as the following:

df = pd.DataFrame({'Shop' : [1,2,3,4,5,6,7,8,9,10],'Week 1' : [15,25,11,22,0,-1,15,11,76,62],'Week 2' : [5,44,55,21,12,51,-10,25,81,46]})
print(df)


 Shop   Week 1  Week 2
0   1   15      5
1   2   25      44
2   3   11      55
3   4   22      21
4   5   0       12
5   6   -1      51
6   7   15     -10
7   8   11      25
8   9   76      81
9   10  62      46

In this instance, the week number should be an observation and the number is an value that should be assigned to it.

what I'm trying to do is the following.

transpose the DF but keep the index as the Shop. turn each instance into an observation so taking only the first 2 shops as an example:

    Shop    Week Hour
0   1       1    15
1   1       2    5
2   2       1    25
3   2       2    44

What would be the most pythonic way to achieve this? on a relatively medium sized df (500 rows 52 weeks)

Answer 1

Using wide_to_long

pd.wide_to_long(df,'Week ',i='Shop',j='week')
Out[770]: 
           Week 
Shop week       
1    1        15
2    1        25
3    1        11
4    1        22
5    1         0
6    1        -1
7    1        15
8    1        11
9    1        76
10   1        62
1    2         5
2    2        44
3    2        55
4    2        21
5    2        12
6    2        51
7    2       -10
8    2        25
9    2        81
10   2        46

#pd.wide_to_long(df,'Week ',i='Shop',j='week').sort_index(level=0).reset_index().rename(columns={'Week ':'Hour'})

Answer 2

You can rename columns, pd.melt and then sort_values :

df.columns = [i if not i.startswith('Week') else int(i[-1]) for i in df]

res = pd.melt(df, id_vars='Shop', var_name='Week', value_name='Hour')\
        .sort_values('Shop').reset_index(drop=True)

print(res)

    Shop Week  Hour
0      1    1    15
1      1    2     5
2      2    1    25
3      2    2    44
...
16     9    2    81
17     9    1    76
18    10    1    62
19    10    2    46

Answer 3

I would use something like this, though it's a bit messy with all the re-naming:

# Rename columns with dict comprehension so it can extend to more than week 1 and week 2
df2 = (df.rename(columns={i: int(i.split()[-1]) for i in df.columns[1:]})
       .set_index('Shop')
       .stack()
       .reset_index()
       .rename(columns={'level_1':'Week', 0:'Hour'}))

>>> df2

    Shop  Week  Hour
0      1     1    15
1      1     2     5
2      2     1    25
3      2     2    44
4      3     1    11
5      3     2    55
6      4     1    22
7      4     2    21
8      5     1     0
9      5     2    12
10     6     1    -1
11     6     2    51
12     7     1    15
13     7     2   -10
14     8     1    11
15     8     2    25
16     9     1    76
17     9     2    81
18    10     1    62
19    10     2    46