sql select number除法汇总和功能

Question

I have this schema

and I want to have a query to calculate the cost per consultant per hour per month. In other words, a consultant has a salary per month, I want to divide the amount of the salary between the hours that he/she worked that month.

SELECT
    concat_ws(' ', consultants.first_name::text, consultants.last_name::text) as name,
    EXTRACT(MONTH FROM tasks.init_time) as task_month,
    SUM(tasks.finish_time::timestamp::time - tasks.init_time::timestamp::time) as duration,
    EXTRACT(MONTH FROM salaries.payment_date) as salary_month,
    salaries.payment
FROM consultants
INNER JOIN tasks ON consultants.id = tasks.consultant_id
INNER JOIN salaries ON consultants.id = salaries.consultant_id
WHERE EXTRACT(MONTH FROM tasks.init_time) = EXTRACT(MONTH FROM salaries.payment_date)
GROUP BY (consultants.id, EXTRACT(MONTH FROM tasks.init_time), EXTRACT(MONTH FROM salaries.payment_date), salaries.payment);

It is not possible to do this in the select

salaries.payment / SUM(tasks.finish_time::timestamp::time - tasks.init_time::timestamp::time)

Is there another way to do it? Is it possible to solve it in one query?

Answer 1

Assumptions made for this answer:

The model is not entirely clear to me, so I am assuming the following:

• you are using PostgreSQL
• salaries.date is defined as a date column that stores the day when a consultant was paid
• tasks.init_time and task.finish_time are defined as timestamp storing the data & time when a consultant started and finished work on a specific task.

---

Your join on only the month is wrong as far as I can tell. For one, because it would also include months from different years, but more importantly because this would lead to a result where the same row from salaries appeared several times. I think you need to join on the complete date:

FROM consultants c 
  JOIN tasks t ON c.id = t.consultant_id
  JOIN salaries s ON c.id = s.consultant_id 
                 AND t.init_time::date = s.payment_date --<< here

If my assumptions about the data types are correct, the cast to a timestamp and then back to a time is useless and wrong. Useless because you can simply subtract to timestamps and wrong because you are ignoring the actual date in the timestamp so (although unlikely) if init_time and finish_time are not on the same day, the result is wrong.

So the calculation of the duration can be simplified to:

t.finish_time - t.init_time

To get the cost per hour per month, you need to convert the interval (which is the result when subtracting one timestamp from another) to a decimal indicating the hours, you can do this by extracting the seconds from the interval and then dividing that by 3600, eg

extract(epoch from sum(t.finish_time - t.init_time)) / 3600)

If you divide the sum of the payments by that number you get your cost per hour per month:

SELECT concat_ws(' ', c.first_name, c.last_name) as name,
       to_char(s.payment_date, 'yyyy-mm') as salary_month,
       extract(epoch from sum(t.finish_time - t.init_time)) / 3600 as worked_hours, 
       sum(s.payment) / (extract(epoch from sum(t.finish_time - t.init_time)) / 3600) as cost_per_hour
FROM consultants c 
  JOIN tasks t ON c.id = t.consultant_id
  JOIN salaries s ON c.id = s.consultant_id AND t.init_time::date = s.payment_date
GROUP BY c.id, to_char(s.payment_date, 'yyyy-mm') --<< no parentheses!
order by name, salary_month;

As you want the report broken down by month you should convert the month into something that contains the year as well. I used to_char() to get a string with only year and month. You also need to remove salaries.payment from the group by clause.

You also don't need the "payment month" and "salary month" because both will always be the same as that is the join condition.

And finally you don't need the cast to ::text for the name columns because they are most certainly defined as varchar or text anyway.

The sample data I made up for this: http://sqlfiddle.com/#!15/ae0c9

---

Somewhat unrelated, but: You should also not put the column list of the group by in parentheses. Putting a column list in parentheses in Postgres creates an anonymous record which is something completely different then having multiple columns. This is also true for the columns in the select list.

Answer 2

If at all the target is putting it in one query, then just confirming, have you tried to achieve it using CTEs? Like

    ;WITH cte_pymt
    AS
    (
    //Your existing query 1
    )
    SELECT <your required data> FROM cte_pymt