model-adapter | 模型适配器

It's caused by: https://github.com/emberjs/data/compare/v2.14.3...v2.14.4

workaround: app/serializers/application.js

Found the problem. The objects did turn out usable and the loop was not a fault with the code I posted. It was the fault with an Angular component ng-multiselect-dropdown, which I was trying to get the data for (it called out the method that queried the place list). After a little research, it turned out the component didn't support dynamic content and it or Angular itself kept repeating the GET-request because of that.

After some workarounds, the component now accepts dynamic content and I can access the place objects just fine with the code I posted before.

Let's try to focus a single relationship, since there's a lot going on in the question you've posted. We'll look at second-resource.

It looks like Mirage is sending back second_resource_ids under the attributes key of the first_resource primary data in the JSON:API payload. That tells me Mirage thinks second_resource_ids is an attribute of first_resource, when in fact it's a relationship.

Assuming your Models & Relationships are setup correctly, you need to tweak the way you're creating data in Mirage.

If you take a look at the Associations section of the Defining Routes guide, you'll see this message:

Mirage's database uses camelCase for all model attributes, including foreign keys (e.g. authorId in the example above)

Right now, you're doing this:

In Javascript, there is no typings. Your variables are dynamically typed when they are assigned a value, or when an operation is performed on them.

In Typescript, you have the possibility to type your variables. But once the compilation is made, the types are removed. They are just here to enforce some behavior pre-compilation, and to help you with intellisense.

But to answer you :

1 - The form model should depend on the data model. If you want to edit the informations and send them back, they should have the same structure.

For that, you have two options, template-driven (which relies on your model), and reactive (which relies on the implementation you make) forms. In your case, you should use the template-driven approach.

2 - If the payload sent doesn't differ from the payload received (model-wise), then you should not have different models. So no, you should not create classes or interfaces that are specific to the form.

3 - It should be transparent, you should not have to do anything.

There is no exact need to create another adapter if you will handle requests at server side the same way. All you really need is to create model which corresponds to payload from server. You will need to create another one adapter only if something differs.

DB migrations are always a complicated piece in app development. That said, there are pros and cons which will at some point make you wipe out the entire DB and start from scratch on major updates.

The good thing is that you should be able to check if the column exist or not, either by doing a simple SELECT in the migration.up function and validating if it was a successful query using http://docs.appcelerator.com/platform/latest/#!/api/Titanium.Database.ResultSet-method-isValidRow

You can also "version" your DB by adding a single Alloy.Globals attribute with every change you do to the DB schema. Something like Alloy.Globals.DatabaseVersion = 1 and validate this in your migration.up and migration.down functions, this way you at least know where you are in your current DB schema and where you want to be in your migrated DB schema.

The not-so-good thing is that with every change you will have to write code to manage all the possible uses cases (e.g. a user with version 1 jumps to version 3 without going to 2) so think about this also as you go.