Question

I came across a colleagues code and thought it was possibly inefficient

bool any = (from c in listDeviceMaxDate
             where c.DeviceKey == m_deviceList[i].deviceKey
             select c).Any();

if (!any)
{
    latestDate = (DateTime)System.Data.SqlTypes.SqlDateTime.MinValue;
}
else
{
    // from the list we have get the lastest max date from the flow table
    DeviceDateTimeItem temp = (from c in listDeviceMaxDate
                                where c.DeviceKey == m_deviceList[i].deviceKey
                                select c).First();

    latestDate = Convert.ToDateTime(temp.dateTimeMax);
}

My first instinct is to store the linq query and then just reference it as needed but then I realised that the First() operator may prevent linq from actually getting ALL the rows which an unconstrained query would do.

How I initially thought about restructuring the code:

var deviceList = from c in listDeviceMaxDate
                            where c.DeviceKey == m_deviceList[i].deviceKey
                            select c;

if (!deviceList.Any())
{
    latestDate = (DateTime)System.Data.SqlTypes.SqlDateTime.MinValue;
}
else
{
    // from the list we have get the lastest max date from the flow table
    DeviceDateTimeItem temp = deviceList.First();

    latestDate = Convert.ToDateTime(temp.dateTimeMax);
}

My question is does the First() call on the second linq query prevent it from returning all results, and as such, is it actually quicker to do it the original way?

Answer 1

It actually depends on what LINQ implementation it is. If it is LINQ-to-Objects (i.e. IEnumerable<T>), then it is basically just enumerating the data whatever it is, and returning the first item if one. So First() is the moral equivalent of:

foreach(var val in sequence) return val;
throw OopsNoData();

and Any() should compare well to:

foreach(var val in sequence) return true;
return false;

(it probably uses a raw iterator in the actual implementation, rather than foreach)

However! If it is LINQ-to-anything-else, all bets are off. LINQ queries (especially IQueryable<T>) are designed to be composable – and I would expect LINQ-to-SQL, for example, to make First() into a select TOP 1 ... TSQL query, and similarly for most other database-backends. So yes, telling it that you only want one row should be helpful. However! I would also expect .Any() to do something very similar, so that shouldn’t (in theory) be a big difference. In a perfect world, it might even use exists(...) in the TSQL, but the world is far from perfect.

The way to find out: attach a sql tracer, and see what the final TSQL was.

---

The ultimate way to do this is simpler:

var obj = someQuery.FirstOrDefault();
if(obj == null) {
   // no match
} else {
   // do something with "obj"
}