I'm trying to move an ASP.NET MVC 5 application to MVC 6 (ASP.NET vNext).
My existing application uses EF 6 and has some EF code first migrations. I want to port these migrations as well, because I want to be able to continue using the existing databases with the existing migrations already applied, without breaking the upgrade path.
It seems that the "new" EF Core migrations (k ef migration add ...) have quite the different format compared to the "old" EF 6 migrations:
No more Designer partial class
No more RESX file
Is there already an "official" way to upgrade these migrations? Is it supported at all? Will it be supported in the future?
There isn't really a way to do this yet. In general we strongly recommend against trying to upgrade an EF6 application to EF7 yet. EF7 is still very much pre-release.
We will have some guidance on how to do this when we get closer to RTM. We may provide some tools to help or just some guidance on how to do it.
Date: 13/April/2016
As of now it is not advisable to use EF7 in production because of following limitaions
"Because of the fundamental changes in EF7 we do not recommend attempting to move an EF6.x application to EF7 at this stage. If you want to move to EF7 to make use of new features, then make sure you are aware of the limitations of EF7 before you start. You should view the move from EF6.x to EF7 as a “port” rather than an “upgrade” ".
One area of RC1 that has a number of outstanding issues is our query pipeline. The majority of issues result in an exception when you attempt to execute a LINQ query that contains a particular pattern. There are often ways to workaround these issues by expressing the same query using different patterns, or evaluating parts of the query client-side. We try to include these workarounds in the issue, when they are available.
In RC1, there are a number of scenarios where EF7 performs slower than EF6
More Readings:https://blogs.msdn.microsoft.com/dotnet/2015/11/18/entity-framework-7-rc1-available/
Related
I am wondering about designing a EF core model of the SAP B1 database. I realize that this is a major undertaking, but I aim to start small.
This is mainly as an exercise, which may lead to something in the future.
My main issue right now, is to ensure that the DbContext is read-only.
I have found several suggestions, like this one:
How to make Entity Framework Data Context Readonly
However, that is for the Entity Framework "standard" not core, and the accepted solution does not seem to be possible with EF Core.
So, aside from using a login without write permissions, how would I go about making a read-only DbContext?
I am using EF Core 5.0.
The project currently I'm working on, it has been changed to fluent migrator from EF migrations. What are the benefits of fluent migrator over ef migrations? Is it really worth using over EF migrations?
EF migrations are inherently code-first - you write model first, run ef commands to generate automatic migration and then update the database. No matter how sophisticated, automatic migrations are always problematic. First, things like column renames, dropping unused columns are always problematic. Additionally, if you are using F# record types or C# POCO objects, then to facilitate migrations, you often have to decorate your plain DB entities with migration specific attributes which are not desirable.
Second, EF migrations are not easily package-able as a standalone console app. Packaging migrations into a separate executable app are always a better idea as the consumer of your application would not have worry about knowing the specific commands like entity framework migration commands. Packages like FluentMigrator and DbUp makes it very easy to package into executable. However, it depends on the application needs. For example, if you are building off-the-shelf application like open source Wordpress style blogging engine and if your audience is not well versed with .net core, then migration as a dedicated utility is helpful.
You might want to choose to run ef migrations programmatically but that again is an anti-pattern as in container world, multiple container might yield race condition and needs special care, thus always better to have separate console project for this.
Is it possible to deploy / install Entity Framework Core 2.0 alongside traditional Entity Framework 6? Is it fully possible, or possible but with some hang-ups, or not possible? Is this documented somewhere? I think I've seen they said they designed it to be side-by-side, but I'm having a hard time fully confirming this. Thanks.
The official documentation says the following:
It is possible to use EF Core and EF6 in the same application. EF
Core and EF6 have the same type names that differ only by namespace,
so this may complicate code that attempts to use both EF Core and EF6
in the same code file.
If you are porting an existing application that has multiple EF
models, then you can selectively port some of them to EF Core, and
continue using EF6 for the others.
This means of course that you can install both EF6 and EF Core in the same project. I have done this in a few simple cases myself and it was working ok.
How can I scaffold EF Core directly from a Visual Studio SQL Server Database Project?
Solutions such as the following are preferred:
scaffold-dbcontext -connection "provider=ssdtproject, name=myprojectname.sqlproj"
scaffold-dbcontext -ddl "ssdtprojectoutput.sql"
scaffold-dbcontext -ssdtschema "ssdtproject.dacpac"
maintained-third-party-tool myprojectname.sqlproj -EfModelGenerationParameters
That's the whole question. What follows is my situation in more detail, so that you may be able to offer alternate solutions:
Although MS acknowledges EF Core is still not production-ready, it's also now 3-4 years since EF 6 progress ceased, and EF Core is the only LINQ code-similar path forward with NETCore compatibility. Thus begins the saga titled "So you're going to be using EF Core."
This part is opinionated, but to me (based on 25+ years of enterprise software design and development experience) Code-First is an absolute non-starter. It's fine for small week-one application concepts, but there's no reasonable pattern/process/practice that I can see to integrate constraints, views, etc. Without views designed-in, real business apps end up with devs repeating logic fundamentals in LINQ expressions all over the place, littering the code with static fields to support LINQ-to-SQL queries, confusing micro-combinatory patterns using LinqKit, etc. Without constraints we end up with ten times the defensive code requirements to handle runtime errors, rapidly blossoming unit and integration tests, and demo failures become the norm. Either our object-oriented experts need to become SQL experts or the converse, and we drastically increase the difficulty of finding and properly-compensating engineers. All of these issues I pointed out in a detailed conversation four years ago with Rowan Miller (who recently left the EF team, which doesn't bode well for near-term solutions).
Model-First (the visual .edmx designer in prior EF versions) is obviously off the table, since the MS solution to this was to claim Code-First really IS Model-First, and wash their hands of it. Consequently a truly neutral, let's call it "Contract-First" for clarity, approach doesn't exist in EF Core.
So, that rant (sorry, frustrated) brings me to Database-First, and thus Scaffold-DbContext. Our DB Schema is currently a revision-controlled Visual Studio SQL Server Database Project. Aside some known issues with this, it also seems ridiculous to have to take our DB schema (currently our single-point-of-truth), rebuild a live database from it, and then back-generate code from the live database, all as part of our build process just to verify database type alignment. I'd like instead to be able to simply detect changes and regenerate my DbContext and related Entities directly from the Database Project.
SSDT Database Projects seem to make Database-like objects available in many of the UIs where normally database connections are required. That makes me think it may be a short walk to use the database schema as a source for existing tools. For example, use a metadata provider in a connection string, make a simple modification to the EF Core code, etc.
SQL Sharpener "generate[s] at design-time using SQL files as the source-of-truth (such as those found in an SSDT project", and was recommended as a solution to this problem for previous versions of EF, but it does not support EF Core.
SQLite and SQL Server Compact Toolbox just added support for generating EF Models directly from .DACPAC, but it appears to depend on the EntityFramework Reverse POCO Code First Generator for that functionality, which prominently lists "Support EF Core" on its TODO List. The primary contributor of this project confirms that incompatibility.
Help?
I was struggling with the same until I ran into the sublime EF Core Power Tools extension for Visual Studio. Its reverse engineering tool sounds just like what you need.
https://marketplace.visualstudio.com/items?itemName=ErikEJ.EFCorePowerTools
I've searched but haven't found much on this topic. Has anyone seen a concise "order of operations" for the DBContext, including stuff like instantiation, validation, saving, etc? I'm mainly curious because while using LINQPad with my EF 4.3 Code First context, it runs two queries before any of my own. One has to do with the migration history which I understand is because we're using EF Migrations and it's attempting to see if it needs to auto-update, and another is for the EdmMetadata table which I don't understand since it sounds like that's not necessary if we're using EFMigrations.
My lack of understanding of why the one query is called makes me wonder if there are other parts of the DBContext's lifecycle that I am unaware of that might be useful to understand/override. Thanks in advance!
The query for EdmMetadata is for compatibility reasons. EF 4.1 did not have migrations. To be able to figure out whether model changed or not the EdmMetadata table was used. Let's say you developed an app using EF 4.1 app and you moved to EF 4.3 but you did not change the model. In this case there is no need to run migrations, throw exceptions or touch the database since your model has not changed. The call to the EdmMetadata table is just to be able to handle this situation gracefully and avoid throwing or touching the database if it is not needed.
I don't think there were any other changes of this kind in EF 4.3.