linq Group By 并取 first 和 lats 记录

Question

I am dealing with probably simple yet difficult problem for me. I am taking photos of parked cars and save them by car's plate and timestamp it. Same car could be photographed several times during the day.

Sample Data

Plate	Brand	Model	InsertDate
99AA111	Tesla	S	2022-01-17 04:00:00
99AA111	Tesla	S	2022-01-17 04:30:00
99AA111	Tesla	S	2022-01-17 05:00:00
59TA3312	Nissan	Skyline	2022-01-17 04:00:00
59TA3312	Nissan	Skyline	2022-01-17 04:30:00
129EA512	Subaru	Impreza	2022-01-17 03:30:00

What i am trying to achieve is;

Plate	Brand	Model	FirstPhotoDate	SecondPhotoDate
99AA111	Tesla	S	2022-01-17 04:00:00	2022-01-17 04:30:00
99AA111	Tesla	S	2022-01-17 05:00:00	-
59TA3312	Nissan	Skyline	2022-01-17 04:00:00	2022-01-17 04:30:00
129EA512	Subaru	Impreza	2022-01-17 03:30:00	-

I have came up with;

var groupedResult = resultList.GroupBy(f => f.Plate).Select(f => new ResultListView
{
    Plate = f.Key,
    FirstDate = f.Min(f => f.InsertDate),
    SecondDate = f.Max(f => f.InsertDate),
    Brand = f.FirstOrDefault().Brand,
    Model = f.FirstOrDefault().Model,
    TimeDifference = (f.Max(f => f.InsertDate) - f.Min(f => f.InsertDate)).TotalMinutes
});

But as it's obvious in the code, it only gives me the first and the last record but not as I expected. I am trying to group by plates and if same plate goes more than once, match it with the next one. if only there is one photo or three photo, second date should be null.

Of course different logic's could be applied but I believe this is more clear way.

I thougt that looping inside grouped result. Take first and the second and skip two. But this is not I am looking for.

Thank you for your help.

Answer 1

Well, having (let use named tuple to demo):

var resultList = new (string Plate, string Brand, string Model, DateTime InsertDate)[] {
  ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 04,00,00)),
  ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 04,30,00)),
  ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 05,00,00)),
  ("59TA3312", "Nissan", "Skyline", new DateTime(2022,01,17, 04,00,00)),
  ("59TA3312", "Nissan", "Skyline", new DateTime(2022,01,17, 04,30,00)),
  ("129EA512", "Subaru", "Impreza", new DateTime(2022,01,17, 03,30,00)),
};

You can GroupBy twice :

var groupedResult = resultList
  .GroupBy(item => item.Plate)
  .SelectMany(group => group
     .OrderBy(item => item.InsertDate)
     .Select((item, index) => (item: item, index: index / 2))
     .GroupBy(pair => pair.index)
     .Select(g => (
        Plate: g.First().item.Plate,
        Brand: g.First().item.Brand,
        Model: g.First().item.Model,
        FirstPhotoDate: g.First().item.InsertDate,
        SecondPhotoDate: (g.Count() == 1 ? null : (DateTime?)(g.Last().item.InsertDate))
     ))
   );

Let's have a look:

string report = string.Join(Environment.NewLine, groupedResult
  .Select(r => $"{r.Plate,-8} : {r.Brand,-6} : {r.Model,-7} : {r.FirstPhotoDate} : {r.SecondPhotoDate}"));

Console.Write(report);

Outcome:

99AA111  : Tesla  : S       : 17.01.2022 4:00:00 : 17.01.2022 4:30:00
99AA111  : Tesla  : S       : 17.01.2022 5:00:00 : 
59TA3312 : Nissan : Skyline : 17.01.2022 4:00:00 : 17.01.2022 4:30:00
129EA512 : Subaru : Impreza : 17.01.2022 3:30:00 : 

Answer 2

Here's the cleanest approach I can think of:

var resultList = new (string Plate, string Brand, string Model, DateTime InsertDate)[]
{
    ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 04,00,00)),
    ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 04,30,00)),
    ("99AA111",  "Tesla",  "S",       new DateTime(2022,01,17, 05,00,00)),
    ("59TA3312", "Nissan", "Skyline", new DateTime(2022,01,17, 04,00,00)),
    ("59TA3312", "Nissan", "Skyline", new DateTime(2022,01,17, 04,30,00)),
    ("129EA512", "Subaru", "Impreza", new DateTime(2022,01,17, 03,30,00)),
};

var query =
    from r in resultList
    group (DateTime?)r.InsertDate by new { r.Plate, r.Brand, r.Model } into grs
    from bgrs in grs.OrderBy(x => x).Buffer(2)
    select new
    {
        grs.Key.Plate,
        grs.Key.Brand,
        grs.Key.Model,
        FirstPhotoDate = bgrs.First(),
        SecondPhotoDate = bgrs.Skip(1).Any() ? bgrs.Last() : null,
    };

I used the Buffer operator from Microsoft's System.Interactive NuGet.

That gives me: