I have a list of Customers who each have a list of Orders. Each Order has a list of LineItems.
I would like to write a LINQ query that would get me the top 10 customers based on order value (i.e. money spent) and not the total number of orders.
One customer could have 2 orders but could have spent £10,000, but another customer could have 100 orders, and only spent £500.
Right now, I have this which gets me the top 10 customers by the number of orders.
var customers = (from c in _context.Customers where c.SaleOrders.Count > 0
let activeCount = c.SaleOrders.Count(so => so.Status != SaleOrderStatus.Cancelled)
orderby activeCount descending
select c).Take(10);
UPDATE
Thanks to Jon Skeet's comment about doing a double Sum, I wrote the following query which compiles.
var customers = (from c in _context.Customers where c.SaleOrders.Count > 0
let orderSum = c.SaleOrders.Where(so => so.Status != SaleOrderStatus.Cancelled)
.Sum(so => so.LineItems.Sum(li => li.CalculateTotal()))
orderby orderSum descending
select c).Take(10);
But when I run this, I get the following error:
It seems LINQ doesn't recognise my .CalculateTotal() method which sit on my LineItem.cs entity.
The problem you were seeing is that CalculateTotal() is not something that Linq can translate into SQL (which is done at run-time, hence no complier error).
The essential problem here is that Linq doesn't really work on lambdas (Func<>), but actually Expressions (Expression<Func<>>), which is the code in a partial compiled state, which Linq then goes about disassembling and translating into SQL.
So, let assume CalculateTotal is a member function defined like this:
public decimal CalculateTotal()
{
return this.quantity * this.value;
}
We could define that as a local lambda function
Func<LineItem, decimal> CalculateTotal = (li => li.quantity * li.value);
Now, we have a lambda which takes a LineItem and returns a value, which is exactly what Sum() wants, so now we can replace:
.Sum(so => so.LineItems.Sum(li => li.CalculateTotal()))
with
.Sum(so => so.LineItems.Sum(CalculateTotal))
But that will crash, just as it did before, because, as I said, it wants an Expression. So, we give it one:
Expression<Func<LineItem, decimal>> CalculateTotal = (li => li.quantity * li.value);
Related
I want to change code below to be sql translateable because now i get exception.
Basicallly i want list of customers from certain localisation and there could be more than one customer with the same CustomerNumber so i want to take the one that was most recently added.
In other words - distinct list of customers from localisation where "distinct algorithm" works by taking the most recently added customer if there is conflict.
The code below works only if it is client side. I could move Group By and Select after ToListAsync but i want to avoid taking unnecessary data from database (there is include which includes list that is pretty big for every customer).
var someData = await DbContext.Set<Customer>()
.Where(o => o.Metadata.Localisation == localisation)
.Include(nameof(Customer.SomeLongList))
.GroupBy(x => x.CustomerNumber)
.Select(gr => gr.OrderByDescending(x => x.Metadata.DateAdded).FirstOrDefault())
.ToListAsync();
Short answer:
No way. GroupBy has limitation: after grouping only Key and Aggregation result can be selected. And you are trying to select SomeLongList and full entity Customer.
Best answer:
It can be done by the SQL and ROW_NUMBER Window function but without SomeLongList
Workaround:
It is because it is not effective
var groupingQuery =
from c in DbContext.Set<Customer>()
group c by new { c.CustomerNumber } into g
select new
{
g.Key.CustomerNumber,
DateAdded = g.Max(x => x.DateAdded)
};
var query =
from c in DbContext.Set<Customer>().Include(x => x.SomeLongList)
join g in groupingQuery on new { c.CustomerNumber, c.DateAdded } equals
new { g.CustomerNumber, g.DateAdded }
select c;
var result = await query.ToListAsync();
I need to get a calculation on aggregation from linq which I hope someone can help
I have a list of objects that have 3 fields (date, saleprice and productcode) I need to get FOR EACH date (Group by date), the SUM of saleprice
/ COUNT of distinct product code.
I know how I can find the SUM alone but not calculation by another aggregate
It would be easier to answer your question with some sample code and objects. I'll assume, items is your list of objects:
items.GroupBy(obj => obj.Date)
.Select(g => new
{
Date = g.Key.Date,
Aggregate = g.Sum(obj => obj.SalePrice) / g.Select(obj => obj.ProductCode)
.Distinct().Count()
});
I have a model with the following entities:
Game: Id
User: Id
UserRating: Id, UserId, GameId, Value
Users can give a rating for a game.
I want to have a query that returns the top 5 games with the highest rating. The highest rating should be based on the Value, but when there are 2 or more games with the same rating, the count should also be taken into account.
How do express this in a query, return the result list as Game entities, using lambda expressions ?
(using EF 4, MVC, SQL Server).
I assume that by "highest rating" you mean "highest average rating"? It sounds like you want something like:
var query = from rating in db.UserRating
group rating by rating.GameId into ratings
orderby ratings.Average(x => x.Value) descending,
ratings.Count() descending
join game in db.Game on ratings.Key equals game.Id
select game;
var top5Games = query.Take(5);
EDIT: In non-query-expression form, this would be more painful, though still feasible:
var top5Games = db.UserRating
.GroupBy(rating => rating.GameId)
.OrderByDescending(ratings => ratings.Average(x => x.Value))
.ThenByDescending(ratings => ratings.Count())
.Join(db.Game, r => r.Key, g => g.Id, (r, g) => g)
.Take(5);
In this case doing it in the lambda syntax isn't too bad, as you're only getting the game out of the join... but more generally, it becomes nastier. It's definitely worth understanding and using both forms, depending on which is the simpler approach for the query at hand. In particular, when you start doing multiple joins it becomes horrible in lambda syntax.
Note that this won't give the rating for that game... to get that, you'd probably want something like:
select new { Game = game,
RatingAverage = ratings.Average(x => x.Value),
RatingCount = ratings.Count() }
at the end of the query instead of just select game.
Also note that you may want to exclude games which currently don't have enough ratings to be meaningful yet - otherwise a game with a single rating which is "perfect" will always be at the top.
EDIT: Okay, with the final tie-break of "name of game", I'd definitely use the query expression form:
var query = from rating in db.UserRating
join game in db.Game on ratings.Key equals game.Id
select new { rating, game } into pair
group pair by pair.game into pairs
orderby pairs.Average(x => x.rating.Value) descending,
pairs.Count() descending,
pairs.Key.Name
select pairs.Key;
var top5Games = query.Take(5);
I have the following 3 classes(mapped to sql tables).
Places table:
Name(key)
Address
Capacity
Events table:
Name(key)
Date
Place
Orders table:
Id(key)
EventName
Qty
The Places and Events tables are connected through Places.Name = Events.Place, while the Events and Orders tables: Events.Name = Orders.EventName .
The task is that given an event, return the tickets left for that event. Capacity is the number a place can hold and Qty is the number of tickets ordered by someone. So some sort of grouping in the Orders table is needed and then subtract the sum from capacity.
Something like this (C# code sample below)?
Sorry for the weird variable names, but event is a keyword :)
I didn't use visual studio, so I hope that the syntax is correct.
string eventName = "Event";
var theEvent = Events.FirstOrDefault(ev => ev.Name == eventName);
int eventOrderNo = Orders.Count(or => or.EventName == eventName);
var thePlace = Places.FirstOrDefault(pl => pl.Name == theEvent.Place);
int ticketsLeft = thePlace.Capacity - eventOrderNo;
If the Event has multiple places, the last two lines would look like this:
int placesCapacity = Places.Where(pl => pl.Name == theEvent.Place)
.Sum(pl => pl.Capacity);
int ticketsLeft = placesCapacity - eventOrderNo;
On a sidenote
LINQ 101 is a great way to get familiar with LINQ: http://msdn.microsoft.com/en-us/vcsharp/aa336746
I felt like the following should be possible I'm just not sure what approach to take.
What I'd like to do is use the include method to shape my results, ie define how far along the object graph to traverse. but... I'd like that traversal to be conditional.
something like...
dealerships
.include( d => d.parts.where(p => p.price < 100.00))
.include( d => d.parts.suppliers.where(s => s.country == "brazil"));
I understand that this is not valid linq, in fact, that it is horribly wrong, but essentially I'm looking for some way to build an expression tree that will return shaped results, equivalent to...
select *
from dealerships as d
outer join parts as p on d.dealerid = p.dealerid
and p.price < 100.00
outer join suppliers as s on p.partid = s.partid
and s.country = 'brazil'
with an emphasis on the join conditions.
I feel like this would be fairly straight forward with esql but my preference would be to build expression trees on the fly.
as always, grateful for any advice or guidance
This should do the trick:
using (TestEntities db = new TestEntities())
{
var query = from d in db.Dealership
select new
{
Dealer = d,
Parts = d.Part.Where
(
p => p.Price < 100.0
&& p.Supplier.Country == "Brazil"
),
Suppliers = d.Part.Select(p => p.Supplier)
};
var dealers = query.ToArray().Select(o => o.Dealer);
foreach (var dealer in dealers)
{
Console.WriteLine(dealer.Name);
foreach (var part in dealer.Part)
{
Console.WriteLine(" " + part.PartId + ", " + part.Price);
Console.WriteLine
(
" "
+ part.Supplier.Name
+ ", "
+ part.Supplier.Country
);
}
}
}
This code will give you a list of Dealerships each containing a filtered list of parts. Each part references a Supplier. The interesting part is that you have to create the anonymous types in the select in the way shown. Otherwise the Part property of the Dealership objects will be empty.
Also, you have to execute the SQL statement before selecting the dealers from the query. Otherwise the Part property of the dealers will again be empty. That is why I put the ToArray() call in the following line:
var dealers = query.ToArray().Select(o => o.Dealer);
But I agree with Darren that this may not be what the users of your library are expecting.
Are you sure this is what you want? The only reason I ask is, once you add the filter on Parts off of Dealerships, your results are no longer Dealerships. You're dealing in special objects that are, for the most part, very close to Dealerships (with the same properties), but the meaning of the "Parts" property is different. Instead of being a relationship between Dealerships and Parts, it's a filtered relationship.
Or to put it another way, if I pull a dealership out of your results and passed to a method I wrote, and then in my method I call:
var count = dealership.Parts.Count();
I'm expecting to get the parts, not the filtered parts from Brazil where the price is less than $100.
If you don't use the dealership object to pass the filtered data, it becomes very easy. It becomes as simple as:
var query = from d in dealerships
select new { DealershipName = d.Name,
CheapBrazilProducts = dealership.Parts.Where(d => d.parts.Any(p => p.price < 100.00) || d.parts.suppliers.Any(s => s.country == "brazil")) };
If I just had to get the filtered sets like you asked, I'd probably use the technique I mentioned above, and then use a tool like Automapper to copy the filtered results from my anonymous class to the real class. It's not incredibly elegant, but it should work.
I hope that helps! It was an interesting problem.
I know this can work with one single Include. Never test with two includes, but worth the try:
dealerships
.Include( d => d.parts)
.Include( d => d.parts.suppliers)
.Where(d => d.parts.All(p => p.price < 100.00) && d.parts.suppliers.All(s => s.country == "brazil"))
Am I missing something, or aren't you just looking for the Any keyword?
var query = dealerships.Where(d => d.parts.Any(p => p.price < 100.00) ||
d.parts.suppliers.Any(s => s.country == "brazil"));
Yes that's what I wanted to do I think the next realease of Data Services will have the possiblity to do just that LINQ to REST queries that would be great in the mean time I just switched to load the inverse and Include the related entity that will be loaded multiple times but in theory it just have to load once in the first Include like in this code
return this.Context.SearchHistories.Include("Handle")
.Where(sh => sh.SearchTerm.Contains(searchTerm) && sh.Timestamp > minDate && sh.Timestamp < maxDate);
before I tried to load for any Handle the searchHistories that matched the logic but don't know how using the Include logic you posted so in the mean time I think a reverse lookup would be a not so dirty solution