python - builing new column in sqlalchemy - Stack Overflow

admin管理员组

文章数量:1328031

Given this table containing buying information of customer_id "xxxxxx". How can we write a script that computes the column nb_bmw_cars that are to the number of rows where car_type is "BMW" BUT in the interval ] date - 365 days: date - 1 day [ for each date. (in sqlalchemy).

customer_id	date	car_type	nb_bmw_cars
xxxxxx	2022-11-22	BMW	-
xxxxxx	2023-08-24	BMW	1
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	MERCEDES	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-10-31	BMW	9
xxxxxx	2024-10-31	BMW	9
xxxxxx	2024-10-31	BMW	9

Given this table containing buying information of customer_id "xxxxxx". How can we write a script that computes the column nb_bmw_cars that are to the number of rows where car_type is "BMW" BUT in the interval ] date - 365 days: date - 1 day [ for each date. (in sqlalchemy).

customer_id	date	car_type	nb_bmw_cars
xxxxxx	2022-11-22	BMW	-
xxxxxx	2023-08-24	BMW	1
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	MERCEDES	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2023-12-21	BMW	2
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-02-07	BMW	6
xxxxxx	2024-10-31	BMW	9
xxxxxx	2024-10-31	BMW	9
xxxxxx	2024-10-31	BMW	9

To understand what is nb_bmw_cars, if you take for example "2024-10-31", it has 9 because there is 9 transactions of car_type BMW before the date (2024-10-31) .

• python
• sql
• sqlalchemy

Share Improve this question Follow edited Jan 18 at 7:17 redj asked Jan 17 at 22:01 redjredj 9011 silver badge77 bronze badges 4

• 1 Can you post the script you have tried so far? – Ian Wilson Commented Jan 17 at 22:44
• @IanWilson test_alias = aliased(test) query = select(test.c.customer_id, test.c.date, test.c.car_type, func.count().filter( and_( test_alias.c.date.between( test.c.date - timedelta(days=365), test.c.date - timedelta(days=1) ), test.c.car_type == 'BMW' ) ).label('nb_bmw_cars') ).outerjoin(test_alias, and_(test_alias.c.customer_id == test_alias.c.customer_id) ).group_by(test.c.customer_id, test.c.date, test.c.car_type) – redj Commented Jan 18 at 9:39
• 1 Sorry, I meant put the code you have tried into the question itself. It doesn't format well in comments. – Ian Wilson Commented Jan 18 at 18:48
• @IanWilson i've just added the code that worked for me in the answers section. – redj Commented Jan 21 at 10:59

Add a comment  | 

2 Answers 2

Sorted by: Reset to default 1

For the date 2024-10-31 , nb_bmw_cars should be 7 instead of 9(as mentioned in question) since there are dates(2022-11-22,2023-08-24) older than (2024-10-31) -1 and (2024-10-31) - 365 days which is the criteria mentioned in question.Same holds true for other dates as well.

To calculate the logic you can just self join the table with itself, where a date is between 365 days and -1 and filter the car_type as BMW

In case you want to calculate for different customer ids, you would need to join based on customer id.I have commented it here as in the sample data given it is not required as all customer ids are same.

select_query = text("""
   select t1.customer_id,  t1.date, t1.car_type, 
    (select COUNT(*) FROM test AS t2
       --where  t2.customer_id = t1.customer_id
       where t2.date between date(t1.date, '-365 day') and date(t1.date, '-1 day')
       and t2.car_type = 'BMW'
    ) as nb_bmw_cars
from test AS t1
""")

result = session.execute(select_query).fetchall() 

Output

xxxxxx | 2022-11-22 | BMW | 0
xxxxxx | 2023-08-24 | BMW | 1
xxxxxx | 2023-12-21 | BMW | 1
xxxxxx | 2023-12-21 | BMW | 1
xxxxxx | 2023-12-21 | MERCEDES | 1
xxxxxx | 2023-12-21 | BMW | 1
xxxxxx | 2023-12-21 | BMW | 1
xxxxxx | 2024-02-07 | BMW | 5
xxxxxx | 2024-02-07 | BMW | 5
xxxxxx | 2024-02-07 | BMW | 5
xxxxxx | 2024-10-31 | BMW | 7
xxxxxx | 2024-10-31 | BMW | 7
xxxxxx | 2024-10-31 | BMW | 7

It is not clear whether you want to update or select, in case you wanted update here is an equivalent

update = text("""update test
set nb_bmw_cars = (select COUNT(*) from test AS t2
    where t2.customer_id = test.customer_id
      and t2.car_type = 'BMW'
      and t2.date BETWEEN DATE(test.date, '-365 day') AND DATE(test.date, '-1 day'))""")

Conceptually it is the same, but here is the Sqlalchemy ORM version of it.

Sample data

data = [
    {'customer_id': 'xxxxxx', 'date': date(2022, 11, 22), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 8, 24), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 12, 21), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 12, 21), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 12, 21), 'car_type': 'MERCEDES'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 12, 21), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2023, 12, 21), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 2, 7), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 2, 7), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 2, 7), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 10, 31), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 10, 31), 'car_type': 'BMW'},
    {'customer_id': 'xxxxxx', 'date': date(2024, 10, 31), 'car_type': 'BMW'}
]

-- connections etc

for each_row in all_rows:
  interval_start = each_row.date - timedelta(days=365)
  interval_end = each_row.date - timedelta(days=1)
  count = connection.execute(
    select(func.count()).where(and_(
        test.columns.customer_id == 
        each_row.customer_id,test.columns.car_type == 
        'BMW',test.columns.date.between(interval_start, 
         interval_end)))).scalar()
  connection.execute(
    test.update().where(and_(
        test.columns.customer_id == each_row.customer_id, 
        test.columns.date == each_row.date 
        )).values(nb_bmw_cars=count) )

results = connection.execute(select(test.columns.customer_id, test.columns.date, test.columns.car_type, test.columns.nb_bmw_cars)).fetchall()
for each in results:
    print(each)

Output

('xxxxxx', datetime.date(2022, 11, 22), 'BMW', 0)
('xxxxxx', datetime.date(2023, 8, 24), 'BMW', 1)
('xxxxxx', datetime.date(2023, 12, 21), 'BMW', 1)
('xxxxxx', datetime.date(2023, 12, 21), 'BMW', 1)
('xxxxxx', datetime.date(2023, 12, 21), 'MERCEDES', 1)
('xxxxxx', datetime.date(2023, 12, 21), 'BMW', 1)
('xxxxxx', datetime.date(2023, 12, 21), 'BMW', 1)
('xxxxxx', datetime.date(2024, 2, 7), 'BMW', 5)
('xxxxxx', datetime.date(2024, 2, 7), 'BMW', 5)
('xxxxxx', datetime.date(2024, 2, 7), 'BMW', 5)
('xxxxxx', datetime.date(2024, 10, 31), 'BMW', 7)
('xxxxxx', datetime.date(2024, 10, 31), 'BMW', 7)
('xxxxxx', datetime.date(2024, 10, 31), 'BMW', 7)

Thanks for your answers, I post this one here in sqlalchemy. It worked for me it is based on a self join like @samhita mentioned in the comment above.

table_cars_alias_1 = aliased(table_cars)
table_cars_alias_2 = aliased(table_cars)


select_query = (
    select(
        table_cars_alias_1.c.customer_id,
        table_cars_alias_1.c.date,
        table_cars_alias_1.c.type_car,
        func.sum(
            case(
                (table_cars_alias_2.c.type_car == 'BMW', 1),
                else_=0
            )
        ).label("nb_cars_bmw"),
        func.sum(case(
            (table_cars_alias_2.c.type_car == 'MERCEDES', 1)
            ,else_=0 )
        ).label("nb_cars_mercedes")
    ).select_from(
        table_cars_alias_1.join(
            table_cars_alias_2,
            and_(
                table_cars_alias_1.c.customer_id == table_cars_alias_2.c.customer_id,
                table_cars_alias_2.c.date.between(
                    func.date(table_cars_alias_1.c.date) - text("interval '365 days'"),
                    func.date(table_cars_alias_1.c.date) - text("interval '1 days'")
                )
            )
        )
    )
    .group_by(
        table_cars_alias_1.c.customer_id,
        table_cars_alias_1.c.date,
        table_cars_alias_1.c.type_car,
    )
)

本文标签： pythonbuiling new column in sqlalchemyStack Overflow