Code review and perfomance tips

[Горобец Дмитрий]

Hello, everybody.

I'm working on one project where I need to render cities on the page in 1-4 columns like this. I'm rendering cities using 'regroup' filter:

A                 D       M             U

Aberdeen    Danbury   Macomb    Ukiah

Akron     Delano   Mandan   Urbana

Manhattan

B        H     P     W

Babylon    Haledon       Pace       Waipio

Bangor    Hanover                   Walker

C   I       R

Calhoun   Indiana          Raleigh

Calipatria   Irondale         Ridge

Cheval Riverdale

Main part of code looks:

def get_columns(location):
    qs = Location.objects.filter(parent=location, active=True)
    qs = qs.annotate(first_char=Upper(Substr('name', 1, 1)))
    locations = qs.values()

    first_chars = qs.distinct('first_char').order_by('first_char')
    first_chars = first_chars.values_list('first_char', flat=True)

    first_chars_count = len(first_chars)
    num_of_columns = 4 if (first_chars_count / 4) >= 1 else first_chars_count
    chars_per_column = math.ceil(first_chars_count / num_of_columns)

    start = 0
    stop = chars_per_column

    for i in range(num_of_columns):
        yield [location for location in locations 
               if location['first_char'] in first_chars[start:stop]]

        start += chars_per_column
        stop += chars_per_column

It can be rewritten using QuerySet filter, but this approach adds more queries in database:

def get_columns(location):
    qs = Location.objects.filter(parent=location, active=True)
    qs = qs.annotate(first_char=Upper(Substr('name', 1, 1)))

    first_chars = qs.distinct('first_char').order_by('first_char')
    first_chars = first_chars.values_list('first_char', flat=True)

    first_chars_count = len(first_chars)
    num_of_columns = 4 if (first_chars_count / 4) >= 1 else first_chars_count
    chars_per_column = math.ceil(first_chars_count / num_of_columns)

    start = 0
    stop = chars_per_column

    for i in range(num_of_columns):
        yield qs.filter(first_char__in=first_chars[start:stop])

        start += chars_per_column
        stop += chars_per_column

Which is better? Filter using list comprehension or using queryset filter? 

Thanks.