How do I sort a list based on the presence of a string?

Question

I currently have a list of lists, something of this kind:

[['NJ', '10', '2000', '500', '20', '02-03-19', '15:20'], 
 ['NJ', '15', '1500', '600', '20', '02-03-19', '15:30'], 
 ['NYC', '25', '1500', '500', '10', '02-03-19', '15:30'], 
 ['NYC', '15', '1200', '700', '1', '02-03-19', '15:35']]

And I need to sort them with several elements in mind, for example, let's say, in terms of index numbers, 0 is area, 1 is weight, 2 is distance, 3 is height, 4 is autonomy, 5 is date and 6 is a timestamp. I am already sorting all of the elements using this:

list.sort(key=itemgetter(0, time_sorter, 4, 3))

Where time_sorter() is a function I had previously built that sorts lists based on element time stamps. My problem is that I need to sort this list with "area" in mind. For example, say I need to sort the list with all those elements in mind, as I am, but I would also like to sort it, simultaneously, in way that the elements which have "NYC" on their 0 index position are placed first, how do I go about this?

Bonus question would be: If I have multiple parameters in itemgetter() or the sort "key" parameter itself, and I want the sorting to be reverse but for only one of those arguments, how do I go about that?

Is my best option really to separate all these sorts into several functions and call those in the sort key?

Thanks in advance!

Answer 1

Make a function that returns the sort key for each item. The function will be called multiple times, each time with a single item as parameter, and you can call other functions inside it if you want.

def _key(row):
    return row[0], time_sorter(row), row[4], row[3]

Whatever you return will be used to sort the items:

mylist.sort(key=_key)

itemgetter is just a simple function factory that creates a function to get items. If you need anything more complex you should make your own function

Answer 2

With numbers, you can reverse the sorting on some of the keys but not others by selectively using a minus sign, eg vectors.sort(key=lambda v: (vx, -vy)) sorts a list of vectors first by their x component in ascending order, then by y component in descending order as a tie-breaker. Unfortunately, there is no equivalent way to do this with strings.

A good option is to write a class to represent your data as objects instead of lists, and define a custom ordering on your class. The functools.total_ordering decorator makes this easier, so you only have to define __eq__ and __lt__ instead of all six comparison methods:

from functools import total_ordering

@total_ordering
class PlaceAndTime:
    def __init__(self, area, weight, distance, height, autonomy, date, time):
        self.area = area
        self.weight = weight
        self.distance = distance
        self.height = height
        self.autonomy = autonomy
        self.date = date
        self.time = time

    def __eq__(self, other):
        return self.__dict__ == other.__dict__

    def __lt__(self, other):
        # test whether self < other
        if not isinstance(other, PlaceAndTime):
            raise ValueError('Not comparable')
        elif self.area != other.area:
            # sort by area in reverse order using > instead of <
            return self.area > other.area
        elif self.date != other.date:
            return self.date < other.date
        elif self.time != other.time:
            return self.time < other.time
        elif self.autonomy != other.autonomy:
            # sort by autonomy in reverse order using > instead of <
            return self.autonomy > other.autonomy
        else:
            return self.height < other.height

This way you can sort a list of objects simply using objects.sort() without providing a key. If you specifically want to sort one area like 'NYC' to the start of the list, you can still use a custom key:

objects.sort(key=lambda obj: (obj.area != 'NYC', obj))

This works because when obj.area == 'NYC' the comparison will be False , otherwise it will be True , so the comparison False < True means 'NYC' will appear before other areas.

Writing a class likely has other benefits too, eg it lets you write obj.time instead of the less-descriptive lst[6] , and there may be other methods you can write which simplify other parts of your code.