Sort at various levels in Python

Question

I am trying to sort a list of tuples like these:

[('Pineapple', 1), ('Orange', 3), ('Banana', 1), ('Apple', 1), ('Cherry', 2)]

The sorted list should be:

[('Orange', 3), ('Cherry', 2), ('Apple', 1), ('Banana', 1), ('Pineapple', 1)]

So, here 1st the list should be sorted based on tuple[1] in descending order, then if the tuple values (tuple[1]) match like forApple, Banana & Pineapple - list should be further sorted based on tuple[0] in ascending order.

I have tried the possible ways-

top_n.sort(key = operator.itemgetter(1, 0), reverse = True)
# Output: [(Orange, 3), (Cherry, 2), (Pineapple, 1), (Banana, 1), (Apple, 1)]

as "reverse = True", Pineapple, then Banana,...

I finally had to come up with a solution:

top_n.sort(key = operator.itemgetter(0), reverse = False)
top_n.sort(key = operator.itemgetter(1), reverse = True)

Is there any better way to get to the solution like my 1st approach. I am trying to explore more about Python, thus seeking such kind of solution.

Answer 1

Have your key return a tuple of the numeric value negated, and then the string. By negating, your numbers will be sorted in descending order, while the strings are sorted in ascending order:

top_n.sort(key=lambda t: (-t[1], t[0]))

Yes, this is a bit of a hack, but works anywhere you need to sort by two criteria in opposite directions, and one of those criteria is numeric.

Demo:

>>> top_n = [('Pineapple', 1), ('Orange', 3), ('Banana', 1), ('Apple', 1), ('Cherry', 2)]
>>> sorted(top_n, key=lambda t: (-t[1], t[0]))
[('Orange', 3), ('Cherry', 2), ('Apple', 1), ('Banana', 1), ('Pineapple', 1)]

Answer 2

In your case, Martijn Pieters solution is probably the best, but I was considering what you would do if you needed to do this for any number of parameters, and needed to do some ascending and some descending.

This approach creates a function to generate a sort index on the fly. Calling getsortfunction with the list of tuples to sort and a list containing the indices and if they should be in reverse order (for example (2,True) means second index in reverse order), returns a function which creates the sort index for an object. It is pretty ugly, but versatile.

def getsortfunction(values,indices):
    sorts = [sorted(list(set(x[indices[i][0]] for x in values)),reverse=indices[i][1]) for i in range(len(indices))]
    def sortfunction(y):
        return tuple(sorts[i].index(y[indices[i][0]]) for i in range(len(indices)))
    return sortfunction

Examples

a = [('Pineapple',1),('Orange',3),('Banana',1),('Apple',1),('Cherry',2)]
# sort a by index 1 first (in reverse order) and then by index 0 in non-reverse order
b = sorted(a,key=getsortfunction(a,[(1,True),(0,False)])) # gives desired list

With an additional criteria

c = [('Pineapple',1,'Hawaii'),('Orange',3,'Florida'),('Banana',1,'Hawaii'),('Apple',1,'Washington'),('Cherry',2,'Washington')]
# sort first by number (in reverse order) then by state, and finally by fruit
d = sorted(c,key=getsortfunction(c,[(1,True),(2,False),(0,False)]))

# sort c first by number (in reverse order), then by fruit, ignoring state
e = sorted(c,key=getsortfunction(c,[(1,True),(0,False)]))

The getsortfunction first builds a nested list of unique values in order and returns a function which maps each value to be sorted to a numeric tuple giving its indices in the sorted value list.

The biggest advantage of this is that you can determine the sort criteria at runtime (for instance from user requests).