new = pd.DataFrame({'table': \['a','b', 'c', 'd'\], 'desc': \['','','',''\], 'total':\[22,22,22,22\]})
old = pd.DataFrame({'table': \['a','b', 'e'\], 'desc': \['foo','foo','foo'\], 'total':\[11,11,11\]})

all = pd.merge(new, old, how='outer', on=\['table', 'total'\])

Output:

table desc_x  total desc_y
0     a            22    NaN
1     b            22    NaN
2     c            22    NaN
3     d            22    NaN
4     a    NaN     11    foo

Desired Output:

table desc  total
0     a   foo     22
1     b   foo     22
2     c           22
3     d           22
4     a   foo     11

I attempted to outer join, but it removed the descriptions of a and b.`

CodePudding user response：

• it makes no sense given what you are trying to achieve is to outer join on table and total. Changed to outer join on table
• table can then be modified to use preference that is implied in your desired output and cleanup columns

new = pd.DataFrame({'table': ['a','b', 'c', 'd'], 'desc': ['','','',''], 'total':[22,22,22,22]})
old = pd.DataFrame({'table': ['a','b', 'e'], 'desc': ['foo','foo','foo'], 'total':[11,11,11]})

all = pd.merge(new, old, how='outer', on=['table'])

# select prefered columns
all["desc"] = all["desc_x"].replace('', np.nan).fillna(all["desc_y"]).fillna("")
all["total"] = all["total_x"].fillna(all["total_y"])

# clean up columns
all = all.drop(columns=[c for c in all.columns if c[-2:] in ["_x", "_y"]])

all