如何优化查询 - Ef core

Question

I am working a asp .net core web API with EF core. I wrote this query. But this take 20-30seconds to execute.

Anyone have idea to improve this query.

var hotels = await _context.Hotels
                    .Where(i => (i.DestinationCode == request.Destination))
                    .Select(i => new HotelListHotelVm
                    {
                        Item1 = i.Item1,
                        Item2 = i.Item2,
                        Item3 = i.Item3,
                        
                        Item4Code = i.Item4Code,
                        Item4Description = i.Item4.TypeDescription,
                        Item5 = i.Item5.Select(x => new HotelListHotelVm.HotelListItem5Vm
                        {
                            Code = x.Item5Code,
                            Description = x.Item5.Description,
                        }).Where(x =>(incomingItem5s.Length > 0 ) ? (incomingItem5s.Contains(x.Code)) : (x.Code != "")),
                        Item6 = i.Item6.Select(x => new HotelListHotelVm.HotelListHotelItem6Vm
                        {
                            Id = x.Id,
                            Item6TypeCode = x.Item6TypeCode,
                            Order = x.Order,
                            Path = x.Path,
                            VisualOrder = x.VisualOrder,
                        }).Take(3),
                        HotelFacilities =  i.Facilities.ToList().Distinct().Take(6).Select(x => new HotelListHotelVm.HotelListFacilityVm {
                            Id = x.Id,
                            FacilityGroupCode = x.FacilityGroupCode,
                            HotelFacilityGroupDescription = x.FacilityGroup.Description,
                            FacilityCode = x.FacilityCode
                        }),
                    })
                    .Where( i => ((incomingItem4.Length > 0 ) ? (incomingItem4.Contains(i.Item4Code)) : (i.Item4Code != ""))   )
                    .OrderByDescending(i => i.Code)
                    .PaginatedListAsync(request.PageNumber, request.PageSize);



                    foreach( var item in hotels.Items){
                        foreach(var facility in item.HotelFacilities){
                            foreach( var fac in  _context.Facilities){
                        
                                if(facility.FacilityCode == fac.Code){
                                    facility.HotelFacilityDescription = fac.Description;
                                }
                            }
                        }
                    }

I f I remove those foreach code, The query takes 8-10s to execute. But I need those foreach codes. Because I need the HotelFacilityDescription

Any suggestion for optimize the query?

Edit The i.Facilities - model

public class HotelFacility 
    {
        // removed some
        public int FacilityCode { get; set; }

        public int FacilityGroupCode { get; set; }
        public FacilityGroup FacilityGroup { get; set; }

        public int HotelCode { get; set; }
        public Hotel Hotel { get; set; }
    }
}

Answer 1

_context.Facilities will be enumerated (ie database will be called) for every iteration of previous loops. The quick fix is to call it ones and store results in variable:

var facilities = _context.Facilities.ToList();
foreach( var item in hotels.Items){
    foreach(var facility in item.HotelFacilities){
        foreach(var fac in facilities){
                        
            if(facility.FacilityCode == fac.Code){
                facility.HotelFacilityDescription = fac.Description;
            }
        }
    }
}

Next improvement can be converting facilities into Dictionary for searching purposes.

Even better approach can be writing query joining with _context.Facilities on database side (but here more info needed).

Answer 2

I've read this a couple times, but it looks like the relationship for Hotel.Facilities is a Facility, so could you not just do:

  HotelFacilities =  i.Facilities.ToList().Distinct().Take(6).Select(x => new HotelListHotelVm.HotelListFacilityVm {
                        Id = x.Id,
                        FacilityGroupCode = x.FacilityGroupCode,
                        HotelFacilityGroupDescription = x.FacilityGroup.Description,
                        FacilityCode = x.FacilityCode,
                        HotelFacilityDescription = x.Description
                    }),

If for some reason Hotel.Facilities is not pointing at a Facility, but is a Many-to-Many HotelFacilityGroup entity to a FacilityGroup, that also contains a FacilityCode, if the associated FacilityGroup has access to a set of Facilities beneath it you could leverage that:

  HotelFacilities =  i.Facilities.ToList().Distinct().Take(6).Select(x => new HotelListHotelVm.HotelListFacilityVm {
                        Id = x.Id,
                        FacilityGroupCode = x.FacilityGroupCode,
                        HotelFacilityGroupDescription = x.FacilityGroup.Description,
                        FacilityCode = x.FacilityCode,
                        HotelFacilityDescription = x.FacilityGroup.Facilities.Where(f => f.Code == x.FacilityCode).Select(f => f.Description).SingleOrDefault()
                    }),

That would avoid the need to go load all of the facilities to resolve that code. Otherwise, if you do need to fetch across all facilities, pre-loading them would be the way to go, but rather than fetching the entire Facility entity I would recommend just the values you need, the Code and the Description. This cuts down on the amount of memory needed and potentially be a faster query:

var facilities = _context.Facilities
    .Select(f => new 
    {
        f.Code,
        f.Description
    }).ToList();

Answer 3

The loop can be eliminated by projecting the value in the LINQ to Entities query.

It would have been quite easy if you had relationship and navigation property like other *Code fields. But as clarified in the comments, there is no such relationship, so you have to resort to old good manual left other join to emulate what navigation property provide automatically, eg

HotelFacilities = i.Facilities.ToList().Distinct().Take(6)
    // left outer join with Facilities
    .SelectMany(x => _context.Facilities
        .Where(f => x.FacilityCode == f.Code).DefaultIfEmpty(),
    (x, x_Facility) => new HotelListHotelVm.HotelListFacilityVm
    {
        Id = x.Id,
        FacilityGroupCode = x.FacilityGroupCode,
        HotelFacilityGroupDescription = x.FacilityGroup.Description,
        FacilityCode = x.FacilityCode,
        HotelFacilityDescription = x_Facility.Description // <--
    }),

Here x_Facility emulates optional reference navigation property x.Facility if existed.

In case you need just single property from the related table, instead of a left join you could also use the original query with single value returning correlated subquery inside the projection, eg

HotelFacilities = i.Facilities.ToList().Distinct().Take(6)
    .Select(x => new HotelListHotelVm.HotelListFacilityVm
    {
        Id = x.Id,
        FacilityGroupCode = x.FacilityGroupCode,
        HotelFacilityGroupDescription = x.FacilityGroup.Description,
        FacilityCode = x.FacilityCode,
        HotelFacilityDescription = _context.Facilities
            .Where(f => x.FacilityCode == f.Code)
            .Select(f => f.Description)
            .FirstOrDefault() // <--
    }),

or even

HotelFacilityDescription = _context.Facilities
    .FirstOrDefault(f => x.FacilityCode == f.Code).Description

All these will eliminate the need of the post loop executiong additional database queries. You can test them and take the one with best performance (#2 and #3 produce one and the same SQL, so it's a matter of taste - the choice is between #1 and #2/3).