Linq continuous date takewhile

Question

Main problem: how do i group elements by their Date, only if continuous and only if they match some properties?

Details:

Given this type of object:

public class MyObj
{
    public DateTime Date { get; set; }
    public string ConditionalValue1 { get; set; }
    public int ConditionalValue2 { get; set; }
    public int OtherValue { get; set; }

    //for test purposes only, i would like to avoid doing this
    public bool CompareFields(MyObj toBeCompared)
    {
        return ConditionalValue1 == toBeCompared.ConditionalValue1 &&
               ConditionalValue2 == toBeCompared.ConditionalValue2; 
    }
}

I have a list of them, populated as follow:

//list is ordered for semplicity, but it won't be by default
//list is a result of a group by too
var myList = new[]
{
    new MyObj { Date = new DateTime(2009, 10, 20), ConditionalValue1 = "A", ConditionalValue2 = 1, OtherValue = 1 },
    new MyObj { Date = new DateTime(2009, 10, 21), ConditionalValue1 = "A", ConditionalValue2 = 1, OtherValue = 2 },
    new MyObj { Date = new DateTime(2009, 10, 22), ConditionalValue1 = "B", ConditionalValue2 = 1, OtherValue = 3 },
    new MyObj { Date = new DateTime(2009, 10, 23), ConditionalValue1 = "A", ConditionalValue2 = 2, OtherValue = 4 },
    new MyObj { Date = new DateTime(2009, 10, 24), ConditionalValue1 = "A", ConditionalValue2 = 2, OtherValue = 5 },
    new MyObj { Date = new DateTime(2009, 10, 25), ConditionalValue1 = "A", ConditionalValue2 = 1, OtherValue = 6 },
    //Note the hole for day 26
    new MyObj { Date = new DateTime(2009, 10, 27), ConditionalValue1 = "A", ConditionalValue2 = 1, OtherValue = 7},
};

Wanted (pseudocode):

List = 
  { ListOfObjects }, //First has date = "2009/10/20", last has = "2009/10/21" 
  { ListOfObjects }, //First has date = "2009/10/22", last has = "2009/10/22" //reason: doesn't match Value1
  { ListOfObjects }, //First has date = "2009/10/23", last has = "2009/10/24" 
  { ListOfObjects }, //First has date = "2009/10/25", last has = "2009/10/25" //reason: doesn't match Value2
  { ListOfObjects }  //First has date = "2009/10/27", last has = "2009/10/27" //reason: there is no "2009/10/26" object

Tests so far:

//i pick the first element, to be compared with the next ones
var firstInList = myList.First();
//take while date is consequent and fields are compared
var partialList = myList
    .TakeWhile(
        (obj, index) => obj.Date == firstInList.Date.AddDays(index) &&
                        obj.CompareFields(firstInList)
    );

This is partially working, as of course it will return the correct results only for the first matching elements.

I could Skip(count) the list until 0 elements, but I would like to use group by, using an IEqualityComparer instead of method CompareFields.

Answer 1

You may be able to use Enumerable.GroupBy. However, I have a feeling it might loop through the entire set. If so, you could always create your own SequencedGroupBy extension method.

Warning: won't work with parallel extensions (AsParallel)

DateTime lastGroupDate = DateTime.MinValue;
DateTime lastDate = DateTime.MinValue;

IEnumerable<IGrouping<DateTime, MyObj>> groups = myList
    .GroupBy(o =>
        {
            if (lastGroupDate != DateTime.MinValue &&
                o.Date.AddDays(-1) == lastDate)
            {
                lastDate = o.Date;
            }
            else
            {
                lastGroupDate = o.Date;
                lastDate = o.Date;
            }

            return lastGroupDate;
        }
    );

foreach(IGrouping<DateTime, MyObj> grouping in groups)
{
    // grouping.Key == the grouped date / first in sequence

    foreach(MyObj obj in grouping)
    {
        // obj.Date == actual date
    }
}