在备用列上对齐日期Pandas Dataframe

Question

I have a Pandas Dataframe, the columns 1-3-5-7...contain dates, the columns 2-4-6-8-.. contain data values. The dates in the columns do not correspond. I want a single column containing all dates and the remaining columns containing just values Example:

input

      date val1       date   val2        date val3 
2007-12-01 35.6  2007-12-05 101.1  2007-12-05 89.1
2007-12-02 36.7. 2007-12-06 102.3  2007-12-07 89.3
2007-12-05 36.7  2007-12-07 108.3. 2007-12-08 89.5
2007-12-06 36.9  2007-12-08 110.0  2007-12-09 89.3
2007-12-07 36.9. 2007-12-09 102.3  2007-12-12 89.9

output

      date   val1   val2   val3 
2007-12-01   35.6     na     na 
2007-12-02   36.7     na     na 
2007-12-05   36.7  101.1   89.1 
2007-12-06   36.9  102.3     na 
2007-12-07   36.9  108.3   89.3 
2007-12-08     na  110.0   89.5
2007-12-09     na  102.3   89.3
2007-12-12     na     na   89.9

Answer 1

You can iteratively join all the couple of columns into a new empty dataframe.

dft = pd.DataFrame({"date": []})
N = len(df.columns)
for n in range(N // 2):
    dft = dft.merge(df.iloc[:, 2*n:2*(n+1)], on='date', how='outer')

Notice that we define an empty column date to merge on it the first iteration. The 'outer' key says that all the values coming both from the left (initial) and right (merged) dataframe are to be kept, and nans added where needed. Hope this helps.

Answer 2

You can try so(It can happen that columns with same name are renamed):

df:
         date   val1      date.1    val2      date.2  val3
0  2007-12-01   35.6  2007-12-05   101.1  2007-12-05  89.1
1  2007-12-02  36.7.  2007-12-06   102.3  2007-12-07  89.3
2  2007-12-05   36.7  2007-12-07  108.3.  2007-12-08  89.5
3  2007-12-06   36.9  2007-12-08   110.0  2007-12-09  89.3
4  2007-12-07  36.9.  2007-12-09   102.3  2007-12-12  89.9

for index, i in enumerate(xrange(0,len(df.columns),2)):
    col = df.columns[i]
    name = 'df' + str(index)
    name = df.iloc[:,i:i+2]
    if index == 0:
        dft = name
    name.columns = ['date', ('value' + str(i/2+1))]
    if index !=0:
        dft = dft.merge(name, on='date', how='outer')
print dft

Output:

         date value1  value2  value3
0  2007-12-01   35.6     NaN     NaN
1  2007-12-02  36.7.     NaN     NaN
2  2007-12-05   36.7   101.1    89.1
3  2007-12-06   36.9   102.3     NaN
4  2007-12-07  36.9.  108.3.    89.3
5  2007-12-08    NaN   110.0    89.5
6  2007-12-09    NaN   102.3    89.3
7  2007-12-12    NaN     NaN    89.9