Question

Not sure how best to explain this, so bear with me…

I’m writing an application to auto generate a staff rota.

I have a table (tl_RotaDays) which is essentially a calendar which records which staff member was on call that particular day – eg.

Date           Login
---------------------------
02-01-2012     Bob
03-01-2012     Bob 
04-01-2012     Bob 
05-01-2012     Bob 
06-01-2012     Bob 

09-01-2012     Bob 
10-01-2012     Bob 
11-01-2012     Bob 
12-01-2012     Sue 
13-01-2012     Sue 

I want to return the date of the Friday in the last “full” working (Mon-Fri) week each person worked, based on a “full” week being defined as a week in which they worked >=3 days.

As such in the above example, the query for “Bob” should return “13-01-2012”, as he worked the majority of that week, even though Sue worked the end of the week.

I have a query which gets the last date each person worked:

lastEveningOOH = Convert.ToDateTime((from rd in db.tl_rotaDays
                                where rd.ooh == tc.login && ((int)Convert.ToDateTime(rd.date).DayOfWeek >= 1 &&
                                            (int)Convert.ToDateTime(rd.date).DayOfWeek <= 5)
                                select rd.date).Max()),

But the only way I can think to check if the person worked a “full” week is to query the original table against each line, which seems frightfully inefficient.

Answer 1

If dt is the list of dates you can do it in one Linq-Query like this:

DateTime maxWeek =
            dt.Where(d => d.DayOfWeek != DayOfWeek.Saturday && d.DayOfWeek != DayOfWeek.Sunday).GroupBy(
                d => new GregorianCalendar().GetWeekOfYear(d, CalendarWeekRule.FirstFourDayWeek, DayOfWeek.Monday)).Where(
                    g => g.Count() > 2).OrderBy(g => g.Key).Last().Max();

If the purpose is to optimize a SQL query you can check what Linq2Sql generates here and see if not a simple algorithmic approach (look at maximum date => check if full week => look at maximum date – 7 etc.) works best.
Also the above Linq query will possibly be incorrect if the dates span multiple years, but the grouping can be modified to rectify that.