Any reason why retrofit with suspend functions can return Response<MyType> instead of Call<MyType>? I kinda stumbled upon this question because I started looking at the github page for retrofit and actually couldn't find an example of the docs where they show a suspend function. I do see the changelog mentions suspend with Response<> though so I did that and that seemed to work.

Call is an execution mechanism container. It represents the potential of a synchronous or asynchronous request but critically when it has been returned to you nothing has actually been performed.

So what would we be suspending for?

Conversely, suspension is an asynchronous execution mechanism which only resumes when you have the result of an operation. For Retrofit that operation is an HTTP request. Therefore, the return type is the response, or just the response body.

Hi. I've got an existing project with SqlDelight db, in module A. Db already has 19 migrations. I extracted some logic into module B, including one sql model from the module A db. Databases in both modules have the same name and get merged. All works as expected. I would like the model in module B to have its own migrations, so it can be used on its own in the future. Will they get merged with module A db migrations automatically?

Yeah, they do work in simple cases, but currently not with diamond dependencies.

Anvil question: is there a way to pass arguments from the gradle script to the custom code generator? I want to generate things a bit differently in different modules.

How does Moshi look up generated type adapters at runtime? Looking through the source I see it tries to create them from factories, but there are no factories(?) for generated adapters. (I'm curious because I'm trying to implement a similar strategy for something else)

When I insert rows in SQLDelight, what's the easiest way to return the index of the inserted row(s)?

Run another query like

findInsertRowid:
SELECT last_insert_rowid()
;

immediately after in the same transaction

actually now i think you can group that with the query so it's like

insertThing {
  INSERT INTO ...;
  SELECT last_insert_rowid();
}

and if you're on later versions of SQLite or dialects that support it

INSERT INTO ...
RETURNING your_id_column;

7gibf2.jpg

first panel should be complain to alec that the generated insert function doesnt just automatically return the integer id

haha true

Do any of my super simple interceptors look wrong? Getting a crash of

Too many follow-up requests: 21

(full stacktrace in thread, stacktrace mentions both of these interceptors) interceptor 1

class AddOutboundAcceptHeaderInterceptor @Inject constructor() : Interceptor {
  override fun intercept(chain: Interceptor.Chain): Response {
    val request = chain.request().newBuilder().addHeader("Accept", "application/json").build()
    return chain.proceed(request)
  }
}

interceptor 2

class AddOutboundBasicAuthHeaderInterceptor
@Inject
constructor(private val buildTimeValues: BuildTimeValues) : Interceptor {
  override fun intercept(chain: Interceptor.Chain): Response {
    val auth = "Basic ${buildTimeValues.basicAuth.encodeUtf8().base64()}"
    val request = chain.request().newBuilder().addHeader("Authorization", auth).build()
    return chain.proceed(request)
  }
}

Hi, im following the examples on https://cashapp.github.io/sqldelight/2.0.0-alpha05/jvm_mysql/ but having an error “Could not find app.cash.sqldeightmysql dialect2.0.0-alpha05.” . Which repository should i configure? - https://repo.maven.apache.org/maven2/app/cash/sqldeight/mysql-dialect/2.0.0-alpha05/mysql-dialect-2.0.0-alpha05.pom - https://dl.google.com/dl/android/maven2/app/cash/sqldeight/mysql-dialect/2.0.0-alpha05/mysql-dialect-2.0.0-alpha05.pom

Could anyone give me some pointers on debugging SQLDelight (1.5.5) migrations? Our app is using KMM, and this migration adds foreign key constraints to the database for the first time. That's the only thing I see as different from our previous (working) migrations. I'm doing the dance of create new table with the changes, insert from old table, drop old table and rename new table for 5 tables. All but one of the tables silently fails copying the data from the old table, but only on iOS - Android works fine. I've also applied the migration manually outside the app and verified it does what it's supposed to. KMM makes debugging on iOS a challenge, and my log-based debugging hasn't gotten me anywhere other than to verify that, yes, the data is there before the migration and is not there after (by querying directly before migration and again after within the

upgrade

block on

DatabaseConfiguration

). So any pointers on where I should look or any other things to try would be greatly appreciated!

About SQLDelight: I'm using inline value classes for all the id columns of my database. They're all built the same, basically this:

interface LongId : Parcelable { val id: Long } // all implementing this interface

@JvmInline
@Parcelize
value class ProfileId(override val id: Long) : LongId

But I'm forced to provide adapters for all these id classes, and apparently I have to do that for any id column. So with a small db with just 8 tables and a few relations between them, the database creation is already a monster of just providing basically the same adapter in a ton of places. See image, and how practically every required adapter is for an id. Is there a way to "teach" SQLDelight once about how to work with a class? Or that value classes are automatically stored via their underlying types?

No, because it's not true that values are transparently stored. Take a Kotlin Duration, for instance. We don't persist the Double value but instead adapt it to a database type. The same is true of your own value classes. Also multi-field value classes are in KEEP right now.

Just noticed this SqlDelight tool window in Android Studio. What does it do? I tried adding a path to a database file, but nothing happens.

I have a question around SQLDelight and the Flow extension: • I have a view that displays a list of objects. For that I query a Flow<List<T>> and call collectAsState() in my composable and it works fine. • What about if there's a single object that I need to fetch immediately to be able to render my UI, but I also want to get updated if the database changes? Is there still a way to use Flows in that case? I'm still new to kotlin flows so please bear with me :)

kotlinpoet: Is there an option to opt out of the new integer literal formatting like

1_000

? At sqldelight we use integer variables with string interpolation. Now we need to change each interpolation with the new format.

"|notifyQueries(${insert.id}) { emit ->
|      emit("search")
|    }

Is that a test expected value?

yes

Of course we could add a DecimalFormatter. But we still need to change each test.

Yeah. It broke some tests in other projects but that's the problem with relying on the stability of unstable output.

I would probably use a helper function to format the number to match, unless it's possible to reduce the reliance of matching generated code and be more black box.

Yeah, I already start adding a helper formatter to get to work again.

I think the latest kotlinpoet update broke the toTypeName implementation (ksp). I need guidance on where in kotlin-poet a test for this could live.