How to properly use operator.itemgetter to do double sort

Question

I am trying to use itemgetter to do a double sort in python, but I can't seem to grasp it. What I want to do is first sort by 'var2' and then by 'var4', but keeping it sorted by 'var2'. I have the following piece of code that should help (I adapted it from another SO question):

df = df[['var1', 'var2', 'var3', 'var4']]
df = sorted(df, key=operator.itemgetter(1,2))

but I am not sure what the arguments mean for itemgetter. I tried running it the way it is, but all I get are the variables' names.

I also tried doing

df = sorted(df, key=operator.itemgetter(2,4))

but I get the following error: 'IndexError: string index out of range'.

Please help.

Edit: example

I have four variables: date, time, price and a number. I want to sort the dataframe by date, but within each date I want to sort it by the number. I hope this makes sense.

date            time        price   number

09/02/2008  00:20:38    46.0    9987

09/03/2009  07:00:49    46.65   8551

07/05/2008  07:00:51    46.75   13681

08/02/2008  07:00:57    46.75   14022

09/02/2008  07:01:00    46.75   10270

09/08/2008  07:01:11    46.75   14850

09/02/2008  07:01:22    46.75   20568

08/02/2008  07:01:24    46.75   15683

09/02/2008  07:02:16    46.65   11698

Answer 1

operator.itemgetter(a, b, c) is equivalent to lambda x: x[a][b][c] , not to lambda x: (x[a], x[b], x[c]) .

What you really want is:

sorted(df, key=lambda x: (x[1], x[3]))

Also note that I changed the indices 2 and 4 to 1 and 3; Python indexes start with 0.

---

Since you appear to be using pandas DataFrames, not lists (next time, mention that in your question), here's how you sort a DataFrame by value:

df.sort_values(['time', 'number'])

Call df.sort_values with a column or a list of columns to sort by. Don't assign this to anything; df.sort_values works in-place.