Remove RemoveConvert (To avoid float conversion to integer)

[MaSchueler]

I ran into an issue with query generation in Entity Framework Core 2.1.

I have a query

var q = KostenSchweissen.Where(k => k.Grenze >= ewd)
  .OrderBy(k => k.Grenze)
  .FirstOrDefault();

with KostenSchweissen.Grenze being an integer, but ewd being a float.

So this looks for the first line with a "Grenze" greater or equal the value that is passed down.
This ran fine previously (EF Core 1.x, not sure about 2.0), but now it's generating the query (from SQL Server Profiler):

exec sp_executesql N'SELECT TOP(1) [k].[ID], [k].[Grenze], [k].[PreisEdelstahl], [k].[PreisNormalstahl]
FROM [KostenSchweissen] AS [k]
WHERE [k].[Grenze] >= @__ewd_0
ORDER BY [k].[Grenze]',N'@__ewd_0 smallint',@__ewd_0=17

where ewd was 17.1. So this completely ignores that ewd is a float, obviously returning wrong costs. Casting to a double or float in the query doesn't work, the only workaround that I could come up with is

    var q = KostenSchweissen.Where(k => k.Grenze * 1.0 >= ewd)
      .OrderBy(k => k.Grenze)
      .FirstOrDefault();

which is obviously not nice since it involves DB-side calculation. The query is not performance critical in my case, but I'm still wondering: Is this a bug, or considered a feature? If so, why? Isn't this at least a breaking change that should have been announced somewhere? Or am I just to blind to find it?

I have tried to find any hint in the EF documentation - release notes, breaking changes, forums, Stackoverflow... but couldn't find anything.

I would be thankful for any hint as to relevant documentation, best practices or whatever else could be helpful for this problem. And if this is the wrong format for asking a question like this, please forgive me.

Further technical details

EF Core version: 2.1.2
Database Provider: Microsoft.EntityFrameworkCore.SqlServer
Operating system: Windows 10 Pro, 1803
IDE: Visual Studio 2017 15.8.9

[MaSchueler]

Thanks for accepting this as a bug. However, 3.0.0 sounds like a long time away. Any suggestions what I can do in the meantime? Do I manually have to look for all queries where this might happen? And am I the only one who compares floats to ints?

[JonPSmith]

Hi @MaSchueler,

I have a problem in the same area that you have (see #14604). Seems like the team isn't going to fix this before version 3 and I wanted to suggest a workaround that I am going to use, i.e. DbFunctions.

I have used DbFunctions before and its a great way to replace part of a query with your own SQL - see this section of one of my articles where I use a DbFunction.

I hope that helps.

[smitpatel]

This has been fixed.