Linq query: MAX in WHERE - linq

I have a long query that returns
Item {
DateTime entryDate
.....
}
I like to combine the result of this query with another table
Value {
DateTime date,
double value
}
such that if entryDate >= CUTOFF, then take the value on CUTOFF, else take the value on entryDate. In other words, I'd want to achieve:
SELECT Item.*, Value.value WHERE
MIN( Item.entryDate, CUTOFF ) == Value.date
Excuse my syntax, but that's the idea.
EDIT: After some trial and error, I came up with this linq-to-sql query:
from iValue in Values
join iItem in ... (long query)
let targetDate = iItem.EntryDate > CUTOFF ? iItem.EntryDate : CUTOFF
where iValue.Date == targetDate
select new
{
iItem,
targetDate,
iValue
}
Thanks for your help.

yourLongQuery.Where(y => y.Item.entryDate == Value.date || CUTOFF == Value.date)
.Select(x => new {
entrydate = (x.Item.entryDate < CUTOFF ? x.Item.entryDate : CUTOFF),
/*rest of x.Item properties here */ ,
x.Value.date,
x.Value.value
});
Filter the query, Combine the two items into one item and modify the first item

Given that you've returned your data from your Item query and that the Value table is relatively small then this is a nice way to go:
var lookup = values.ToLookup(v => v.date, v => v.value);
var query =
from i in items
let c = i.entryDate < CUTOFF ? i.entryDate : CUTOFF
let v = lookup[c].FirstOrDefault()
select new
{
Item = i,
Value = v,
};
The ToLookup extension is very useful and often overlooked.

such that if entryDate >= CUTOFF, then take the value on CUTOFF, else
take the value on entryDate. In other words, I'd want to achieve:
SELECT Item.*, Value.value WHERE MAX( Item.entryDate, CUTOFF ) ==
Value.date
This is contradictory - if entryDate >= CUTOFF then take the value on CUTOFF imples that you want MIN(Item.entryDate, CUTOFF), not MAX.
Having said that, you just want to select the value.Value that matches each item of your query. Each item should look up the relevant value which matches your MAX (or, I believe, MIN) statement.
query.Select(item =>
{
var matchingValue = Values.Single(v =>
v.date == Min(item.entryDate, CUTOFF));
return new { item, matchingValue.value };
});
This will return an IQueryable of anonymous { Item, double } objects.
If you require this to be a executed as a single SQL statement you'll need to do some refactoring. A good start is to swap matchingValue into a single statement.
query.Select(item => new { item, context.Values.Single(v =>
v.date == Min(item.entryDate, CUTOFF)).value });
I don't have a Visual Studio in front of me to confirm, but I am not sure that the Math.Min function is mapped in LINQ-to-SQL. Let's assume it's not.
query.Select(item => new { item, context.Values.Single(v =>
v.date == (item.entryDate < CUTOFF ? item.entryDate : CUTOFF)).value });
I believe that will resolve to a single query if you execute it with a .ToList() but can't confirm until I have some tools in front of me. Test it with SQL profiler to be sure.

Related

How to get two sums from a linq query

Summary
I have a list of Transactions. Using Linq, I want to get a sum of the Cost and sum of the Quantity from this list in one query.
Grouping
My first thought is to use grouping - but I don't really have a key that I want to group on, I want just one group with the results from the whole list. So, I happen to have a property called "Parent" that will be the same for all of the transactions, so I'm using that to group on:
var totalCostQuery =
(from t in Transactions
where t.Status != GeneralStoreTransactionStatus.Inactive &&
(t.Type == GeneralStoreTransactionType.Purchase ||
t.Type == GeneralStoreTransactionType.Adjustment)
group t by t.Parent into g
select new
{
TotalCost = g.Sum(t => t.Cost.GetValueOrDefault()),
TotalQuantity = g.Sum(t => t.Quantity.GetValueOrDefault())
});
Grouping by t.Parent seems like it could be wrong. I really don't want to group at all, I just want the sum of t.Quantity and sum of t.Cost.
Is that the correct way to get a sum of two different properties or can it be done in a different way.
Assuming this is Linq to SQL or Entity Framework, you can do that:
var totalCostQuery =
(from t in Transactions
where t.Status != GeneralStoreTransactionStatus.Inactive &&
(t.Type == GeneralStoreTransactionType.Purchase ||
t.Type == GeneralStoreTransactionType.Adjustment)
group t by 1 into g
select new
{
TotalCost = g.Sum(t => t.Cost),
TotalQuantity = g.Sum(t => t.Quantity)
});
Note that you don't need to use GetValueOrDefault, null values will be ignored in the sum.
EDIT: not sure this works with Linq to NHibernate though...
Note that if you're using Linq to objects, the solution above won't be efficient, because it will enumerate each group twice (once for each sum). In that case you can use Aggregate instead:
var transactions =
from t in Transactions
where t.Status != GeneralStoreTransactionStatus.Inactive &&
(t.Type == GeneralStoreTransactionType.Purchase ||
t.Type == GeneralStoreTransactionType.Adjustment)
select t;
var total =
transactions.Aggregate(
new { TotalCost = 0.0, TotalQuantity = 0 },
(acc, t) =>
{
TotalCost = acc.TotalCost + t.Cost.GetValueOrDefault(),
TotalQuantity = acc.TotalQuantity + t.Quantity.GetValueOrDefault(),
});

Better way to check resultset from LINQ projection than List.Count?

Is there a better way to check if a LINQ projection query returns results:
IList<T> TList = db.Ts.Where(x => x.TId == 1).ToList(); // More canonical way for this?
if (TitleList.Count > 0)
{
// Result returned non-zero list!
string s = TList.Name;
}
You can use Any(), or perhaps more appropriately to your example, SingleOrDefault(). Note that if you are expecting more than one result and plan to use all of them, then it doesn't really save anything to use Any() instead of converting to a List and checking the length. If you don't plan to use all the results or you're building a larger query that might change how the query is performed then it can be a reasonable alternative.
var item = db.Ts.SingleOrDefault( x => x.TId == 1 );
if (item != null)
{
string s = item.Name;
...
}
or
var query = db.Ts.Where( x => x.Prop == "foo" );
if (query.Any())
{
var moreComplexQuery = query.Join( db.Xs, t => t.TId, x => x.TId );
...
}

Row number in LINQ

I have a linq query like this:
var accounts =
from account in context.Accounts
from guranteer in account.Gurantors
where guranteer.GuarantorRegistryId == guranteerRegistryId
select new AccountsReport
{
recordIndex = ?
CreditRegistryId = account.CreditRegistryId,
AccountNumber = account.AccountNo,
}
I want to populate recordIndex with the value of current row number in collection returned by the LINQ. How can I get row number ?
Row number is not supported in linq-to-entities. You must first retrieve records from database without row number and then add row number by linq-to-objects. Something like:
var accounts =
(from account in context.Accounts
from guranteer in account.Gurantors
where guranteer.GuarantorRegistryId == guranteerRegistryId
select new
{
CreditRegistryId = account.CreditRegistryId,
AccountNumber = account.AccountNo,
})
.AsEnumerable() // Moving to linq-to-objects
.Select((r, i) => new AccountReport
{
RecordIndex = i,
CreditRegistryId = r.CreditRegistryId,
AccountNumber = r.AccountNo,
});
LINQ to objects has this builtin for any enumerator:
http://weblogs.asp.net/fmarguerie/archive/2008/11/10/using-the-select-linq-query-operator-with-indexes.aspx
Edit: Although IQueryable supports it too (here and here) it has been mentioned that this does unfortunately not work for LINQ to SQL/Entities.
new []{"aap", "noot", "mies"}
.Select( (element, index) => new { element, index });
Will result in:
{ { element = aap, index = 0 },
{ element = noot, index = 1 },
{ element = mies, index = 2 } }
There are other LINQ Extension methods (like .Where) with the extra index parameter overload
Try using let like this:
int[] ints = new[] { 1, 2, 3, 4, 5 };
int counter = 0;
var result = from i in ints
where i % 2 == 0
let number = ++counter
select new { I = i, Number = number };
foreach (var r in result)
{
Console.WriteLine(r.Number + ": " + r.I);
}
I cannot test it with actual LINQ to SQL or Entity Framework right now. Note that the above code will retain the value of the counter between multiple executions of the query.
If this is not supported with your specific provider you can always foreach (thus forcing the execution of the query) and assign the number manually in code.
Because the query inside the question filters by a single id, I think the answers given wont help out. Ofcourse you can do it all in memory client side, but depending how large the dataset is, and whether network is involved, this could be an issue.
If you need a SQL ROW_NUMBER [..] OVER [..] equivalent, the only way I know is to create a view in your SQL server and query against that.
This Tested and Works:
Amend your code as follows:
int counter = 0;
var accounts =
from account in context.Accounts
from guranteer in account.Gurantors
where guranteer.GuarantorRegistryId == guranteerRegistryId
select new AccountsReport
{
recordIndex = counter++
CreditRegistryId = account.CreditRegistryId,
AccountNumber = account.AccountNo,
}
Hope this helps.. Though its late:)

LINQ BuildContainsExpression With OR conditions

I'm trying to get the following SQL query to work in LINQ:
Select id from table1 where id in (1,2) or canceledId in (1,2)
I'm using BuildContainsExpression to achieve the "IN" condition, but I can't figure out how to implement the "or" condition.
My shot in the dark is as follows:
var identifiers = new List<int> {1,2};
var query = (from t in Context.Table1
select t);
var query =
query.Where(BuildContainsExpression<Table1, int>(t => t.Id, identifiers));
if (showCanceled)
{
var expression = query.Where(BuildContainsExpression<Table1, int>(t => t.CanceledId.Value, identifiers)).Expression;
Expression.Or(expression, transactionsQuery.Expression);
}
But I get the following exception:
The binary operator Or is not defined for the types 'System.Linq.IQueryable1[Table1]' and 'System.Linq.IQueryable1[Table1]'..
Any ideas? -Am I in the right direction?
Thanks,
Nir.
You are appending your OR in the wrong place. What you are doing now is effectively something like this:
(from t in Context.Table1
where identifiers.Contains(t.Id)
select t)
OR
(where identifiers.Contains(t.CanceledId))
The second problem is that the BuildContainsExpression method you use, returns a lambda expression, something that looks like this:
t => t.Id == 1 || t.Id == 2 || ...
You can't change this expression once it's generated. However, that's what you want because you'd like to have something like this:
t => t.Id == 1 || t.Id == 2 || ... || t.CanceledId == 1 || t.CanceledId == 2 || ...
You can't simply take the body of this lambda expression and or it together with another expression because it depends on the parameter t.
So what you can do is the following:
// Overload of BuildContainsExpression.
private static Expression<Func<T, bool>> BuildOtherContainsExpression<T>(
ParameterExpression p, Expression field1, Expression field2, int[] values)
{
var eq1 = values.Select(v => Expression.Equal(field1, Expression.Constant(v)));
var eq2 = values.Select(v => Expression.Equal(field2, Expression.Constant(v)));
var body = eq1.Aggregate((acc, equal) => Expression.Or(acc, equal));
body = eq2.Aggregate(body, (acc, equal) => Expression.Or(acc, equal));
return Expression.Lambda<Func<T, bool>>(body, p);
}
// Create a parameter expression that represents something of type Table1.
var parameter = Expression.Parameter(typeof(Table1), "t");
// Create two field expressions that refer to a field of the parameter.
var idField = Expression.Property(parameter, "Id");
var canceledIdField = Expression.Property(parameter, "CanceledId");
// And finally the call to this method.
query.Where(BuildContainsExpression<Table1>(
parameter, idField, canceledIdField, identifiers));
Your if statement would now look like this:
if (!showCanceled)
{
// Use original version of BuildContainsExpression.
}
else
{
// Create some expressions and use overloaded version of BuildContainsExpression.
}
I know I'm a bit late to the party here - but I think the original code in the original poster's question was 99% right.
The only wrong was that instead of
Expression.Or
it should have been
Expression.OrElse

How to make a linq Sum return null if the summed values are all null

I have a LINQ query that looks like this...
var duration = Level3Data.AsQueryable().Sum(d => d.DurationMonths);
If all the d.DurationMonths values are null the Sum returns 0. How can I make the Sum return null if all the d.DurationMonths are null? Or do I need to run a separate query first to eliminate this situation before performing the sum?
Along with the previous suggestion for an extension method - you could use a ternary operator...
var duration = Level3Data.AsQueryable().Any(d => d.DurationMonths.HasValue)
? Level3Data.AsQueryable().Sum(d => d.DurationMonths)
: null;
You can use Aggregate to provide custom aggregation code :
var items = Level3Data.AsQueryable();
var duration = items.Aggregate<D,int?>(null, (s, d) => (s == null) ? d.DurationMonths : s + (d.DurationMonths ?? 0));
(assuming the items in Level3Data are of type D)
var outputIndicatorSum = (from OutputIndicatorTable in objDataBaseContext.Output_Indicators
where OutputIndicatorTable.Output_Id == outputId
select (int?)OutputIndicatorTable.Status).Sum();
int outputIndicatorSumReturn = Convert.ToInt32(outputIndicatorSum);
return outputIndicatorSumReturn;
You can explicitly type cast non-nullable varaible into nullable type.
i.e, select (int?)OutputIndicatorTable.Status).Sum();
Using Sum alone, this is impossible. As you indicated in your question, you will need to check for this situation before you call Sum:
var q = Level3Data.AsQueryable();
var duration = q.All(d => d.DurationMonths == null)
? null
: q.Sum(d => d.DurationMonths);
If you would like the result without two queries try:
var duration = Level3Data.AsQueryable().Sum(d => (double?)d.DurationMonths);
If you want zero instead of null as the result of this query use:
var duration = Level3Data.AsQueryable().Sum(d => (double?)d.DurationMonths) ?? 0;

Resources