tags:

views:

82

answers:

2

Hello! I'm currently doing a summer job and I have to extend an existing program.

My boss asked me to make a tool for our clients so they can see how much their employees cost, per month. But that's not all. The thing is that a company can have one or more 'societies', or subcompanies. We want know how much an employee costs per society in a company.

These are the table I use:

  • society: a subcompany with a people_id which contains the name, etc. of the society
  • timesheet: timesheet entries that contain person and society information
  • people: all people or contacts in the database
  • salarystate: contains the salary for a person for a specific month
  • overhead: overhead cost for a person for a specific month (note that date is a string (!) formatted like this: YYYY-MM-DD)

This query works, but it takes very long time to execute. Is there a way to make it faster?

I select the year and month, get the name of the employee (worker) and the name of the society. Then I select the sum of the minutes the employee has worked (for a specific society). And finally I calculate the cost by checking his salary for that month and the overhead for that month.

SELECT
    YEAR(TS.assigndate) AS timesheet_year,
    MONTH(TS.assigndate) AS timesheet_month,
    CONCAT(TP.name, ' ', TP.firstname) AS worker,
    CONCAT(SP.name, ' ', SP.firstname) AS society,
    (
        SELECT
            SUM(timeunits) AS minutes
        FROM timesheet
        WHERE
            people_id = TP.id AND
            society_id = S.id AND
            MONTH(assigndate) = timesheet_month
    ) AS minutes,
    (
        SELECT (minutes / 60)
    ) AS hours,
    (
        SELECT(OO.hourtarif + SS.hourtarif) AS cost
        FROM salarystate SS, overhead OO
        WHERE
            people_id = TP.id AND
            YEAR(OO.date) = timesheet_year AND
            MONTH(OO.date) = timesheet_month AND
            CONVERT(SUBSTRING(SS.month FROM 1 FOR 4), UNSIGNED) = timesheet_year AND
            CONVERT(SUBSTRING(SS.month, -2), UNSIGNED) = timesheet_month
    ) AS cost,
    (
        SELECT (hours * cost)
    ) AS total_cost
FROM timesheet TS, society S, people SP, people TP
WHERE
    S.id = TS.society_id AND
    SP.id = S.people_id AND
    TP.id = TS.people_id
GROUP BY timesheet_year, timesheet_month, worker, society; 
+2  A: 
  1. create Date dimension table http://www.sqlserversavvy.com/2008/03/t-sql-script-to-create-date-dimension.html
  2. add columns that related with date columns to this table
  3. group by with dimDate
MahmutHAKTAN
So I just create a temporary table with the hours and the cost and then perform my calculations on those tables?
gillis
+1  A: 

Now I use temporary tables it goes fast as hell :). This is the code now, if you're interested:

CREATE TEMPORARY TABLE IF NOT EXISTS people_hours (
    people_id INTEGER NOT NULL,
    society_id INTEGER NOT NULL,
    year INTEGER NOT NULL,
    month INTEGER NOT NULL,
    hours DOUBLE NOT NULL,
    PRIMARY KEY(people_id, society_id, year, month)
);

CREATE TEMPORARY TABLE IF NOT EXISTS people_cost (
    people_id INTEGER NOT NULL,
    year INTEGER NOT NULL,
    month INTEGER NOT NULL,
    cost DOUBLE NOT NULL,
    PRIMARY KEY(people_id, year, month)
);

TRUNCATE people_hours;
TRUNCATE people_cost;

INSERT INTO people_hours (people_id, society_id, year, month, hours)
SELECT
    p.id as people_id,
    s.id as society_id,
    YEAR(t.assigndate) as year,
    MONTH(t.assigndate) as month,
    SUM(t.timeunits)/60 as hours
FROM people p, society s, timesheet t
WHERE
    t.society_id = s.id AND
    t.people_id = p.id
GROUP BY year, month, people_id, society_id;

INSERT INTO people_cost (people_id, year, month, cost)
SELECT
    p.id as people_id,
    YEAR(o.date) as cost_year,
    MONTH(o.date) as cost_month,
    SUM(o.hourtarif + s.hourtarif) as cost
FROM people p, salarystate s, overhead o
WHERE
    s.people_id = p.id AND
    CONVERT(SUBSTRING(s.month FROM 1 FOR 4), UNSIGNED) = YEAR(o.date) AND
    CONVERT(SUBSTRING(s.month, -2), UNSIGNED) = MONTH(o.date)
GROUP BY cost_year, cost_month, people_id;

SELECT 
    h.year,
    h.month,
    h.society_id,
    h.hours,
    c.cost,
    (h.hours * c.cost) AS total_cost,
    CONCAT(p.name, ' ', p.firstname) AS employee,
    CONCAT(ps.name, ' ', ps.firstname) AS society
FROM people_hours h, people_cost c, people p, people ps, society s
WHERE
    h.society_id = s.id AND
    h.people_id = p.id AND
    h.people_id = c.people_id AND
    s.people_id = ps.id AND
    h.year = c.year AND
    h.month = c.month
ORDER BY h.year, h.month, h.people_id, h.society_id;
gillis
Good to see you have a solution, the problem with the original query is that the subquery calculations had to be executed per row because they are part of the select, if you were in a situation where you couldn't create temp tables you could have rewritten the query so that you joined against the sub query moving the where part of your subquery into the 'on' condition of the join
ejrowley