如何使用第一行的值拆分一列？

Question

My dataset df looks like this:

time            Open
2017-01-01      2.2475
2017-01-02      3.2180
2017-01-03      5.2128
2017-01-04      1.2128
2017-01-05      2.2128
...., ....
2017-12-31      6.7388

I want to sort the Open column but by comparing the first ROW value in increasing order.

We will have 1st row value always on the top( 1st row ) and then sort starting the second row by comparing to 1st row to the closest value in increasing order. All the low value is kept on the bottom. Eg: 1.2128

[OP seeks a method where values greater than the first row in a select column should appear sequentially and ascending from row 2 to row n, and values less than the first row should then come after n (all of the preceding values).]

For example, the new df would be:

time            Open
2017-01-01      2.2475
2017-01-05      2.2128
2017-01-02      3.2180
2017-01-03      5.2128
...., ....
2017-12-31      6.7388
2017-01-04      1.2128

What did I do?

I can sort by column doing this:

df.sort_values(by='Open', ascending=False)

but that is by column . Now how do I sort by first ROW value, which is 2.2475

Answer 1

IIUC, given a df :

         time    Open
0  2017-01-01  2.2475
1  2017-01-02  3.2180
2  2017-01-03  5.2128
3  2017-01-04  1.2128
4  2017-01-05  2.2128
5  2017-12-31  6.7388

OP wants to sort as row_0 , (rows greater than row_0) , ( rows smaller than row_0) : This can be achieved using difference between each row and row_0:

s = df['Open'].sub(df['Open'][0]).to_dict()
df.iloc[sorted(s, key = lambda x: s.get(x) < 0)]

Output:

         time    Open
0  2017-01-01  2.2475
1  2017-01-02  3.2180
2  2017-01-03  5.2128
5  2017-12-31  6.7388
3  2017-01-04  1.2128
4  2017-01-05  2.2128

Answer 2

OP is after a method where the first row of a DataFrame column is used as a baseline for a split method of column sorting: values greater than this first row should appear sequentially and ascending from row 2 to row n, and values less than the first row should then come after n (all of the preceding values).

This can be achieved by the following function:

df = pd.DataFrame({'time': ['2017-01-01', '2017-01-02', '2017-01-03', '2017-01-04', '2017-01-05', '2017-01-06'], 
              'Open': [2.24, 1.21, 1.51, 3.21, 5.21, 6.21]})

def pin_row_and_sort(f):
    values_above = f.loc[f['Open'] >= f['Open'].iloc[0]].sort_values(by='Open')
    values_below = f.loc[f['Open'] < f['Open'].iloc[0]].sort_values(by='Open')
    return pd.concat([values_above, values_below])

new_frame = pin_row_and_sort(df)

I'd be keen to see any improvements/suggestions on this method. Or just down-vote without explaining why :)