With moshi, how do I generate a decoder if the toplevel is a list?

moshi.adapter<List<Appointment>>(List::class.java).fromJson(response.bodyString())

^ not doing the trick

In SqlDelight, if I define two different queries [for the same fields] on the same table, I end up with generated classes that are identical but with different names. Is there a way to consolidate them or tell the generator to prefer one? [edited]

Use a view. It is very challenging to say

SELECT 'f' as FOO

is equal to

SELECT 'a' as FOO

.

Are there any plans to provide SqlDelight for Kotlin/Wasm?

It depends what do you mean. There is already a Kotlin/JS driver using web workers and SQLite wasm. Kotlin Wasm target itself is not yet supported.

I was actually referring to Kotlin/Wasm target. I wondered if there is any work-in-progress version? Within a few months, with the new Kotlin/Wasm becoming available, I think it makes much more sense to compile Kotlin datalayers to Wasm rather than to JS.

Both make sense

In my case, it would be a performance-based decision. I would expect Kotlin/Wasm to be generally more performing than Kotlin/JS, and have smaller compiled file size.

There's no real plan, but once it's possible (or rather practical) it's definitely something I'd want to implement

SQLDelight/KMP : is there a simple/clean way to catch a specific SQL exception in the commonMain code? (I feel like I have to create my own SqlExceptionMapper with expect/actual but hope there's already something available.)

is there a sqldriver for @Preview?

are you running database operations in your ui composables?

just display data from database

While the models may reach the UI, the queries interfaces probably shouldn't

so AndroidSqliteDriver and Database should be in separate class?

Hello everyone. A question for Anvil: What could I go wrong if I am generating code in my own code generator based on superclass type, not based on annotations? Is there anything in Anvil (like annotation-based caching or something) that could cause issues?

No, that works fine. You can use anything as trigger to generate code.

This is about SQLDelight: Is there a convenient way to generate a couple of default functions for all tables? E.g.

getAll

,

getById(id)

,

deleteById

, etc. I find myself writing them in the sq files for most tables anyway, and that is copy-pasting I'd like to avoid. I know this is predicated on some conditions, like having a atomic

id

property, etc. But again, I find myself writing tables mostly compatible.

If you had Retrofit setup (non suspending) because you have a bunch of places in your code where you already use this... BUT you want your new code to use coroutines, what would you do? I feel like maybe I'd create a new duplicate ApiService and add the suspend modifier and get rid of the Call<> return wrapper type. But keeping two api services in sync seems silly. Am I missing something here for how to use my retrofit apiService in the traditional way + coroutine way?

I know this isn't exactly relating to a square library, but I'm inspired from Mosaic and Molecule. When building out a set of Nodes in Compose is there a meaningful performance difference in the depth of Nodes that get managed? For example if I could centralized state into a single node vs a parent node with 5 children nodes containing that state

Well, a large node tree means a large slot table that each recomposition has to traverse. But as long as you have skippable groups that's probably not a huge deal. A tree of thousands of nodes shouldn't be a problem, but then it becomes what those nodes do. If they're traditional UI do you need to do measure, layout, and draw passes? Those could be expensive and you need all kinds of optimizations for that to be as lazy as possible. Ultimately what you do and it's performance is going to depend on a lot of factors and I wouldn't waste too much time trying to anticipate where the bottleneck is and just go for your idea.

Thanks, that makes sense. I'll give it a shot

Mostly looking for inspiration with this because I don't think there's one true answer, and I don't think this is really Molecule's domain. I'm using Molecule as a presenter to gather state to send to my view (using UDF). User actions on the views are modeled in a

Flow<Intention>

back to the presenter and used to drive certain state changes (or launch coroutines to do work depending on the intention). I also have the concept of

Effects

which are things that happen as side effects to a screen being opened (logging analytics, etc...). The issue is that some Effects end up needing interaction with the user. For example, when a certain screen opens, its Effect starts running a migration under certain conditions. However if that migration fails, I show the user a dialog asking if they want to retry it or skip it. That intention feeds back into the Effect determining if the migration should be retried or not. I've solved that until now by introducing

Sources

which are basically shared mutable state between the Effect and the presenter. But after using it for a few weeks the line between Effect and presenter started blurring. Today it came to a head when I needed to do a combine on two flows in the Effect, and found myself wishing that the Effect used Molecule as well, aside from the fact that Molecule doesn't really fit into what an Effect does. Any suggestions on how to handle this in a Molecule world?

Hi folks I'm using SQLDelight in my KMP app, everything works fine on the local machine but the following task breaks when running in the CI/CD pipeline:

* What went wrong:
Execution failed for task ':shared:verifyCommonMainPTDatabaseMigration'.
> There was a failure while executing work items
   > A failure occurred while executing app.cash.sqldelight.gradle.VerifyMigrationTask$VerifyMigrationAction

      > Error migrating from 1.db, fresh database looks different from migration database:
Deprecated Gradle features were used in this build, making it incompatible with Gradle 9.0.
        /tables[Settings]/columns[Settings.isDarkModeOn]/defaultValue - CHANGED

          BEFORE:
You can use '--warning-mode all' to show the individual deprecation warnings and determine if they come from your own scripts or plugins.

See <https://docs.gradle.org/8.0.1/userguide/command_line_interface.html#sec:command_line_warnings>
198 actionable tasks: 198 executed
            1
          AFTER:
            0
        /tables[Settings]/definition - CHANGED
          BEFORE:
            CREATE TABLE Settings(
                isDarkModeOn INTEGER DEFAULT 1
            )
          AFTER:
            CREATE TABLE Settings(
                isDarkModeOn INTEGER DEFAULT 0
            )

This is how my current GitHub workflow looks like:

name: My Deployment Task
on:
  push:
    branches:
      - main
jobs:
  deployment:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@master
        name: Checking out code...

      - uses: actions/setup-java@v3
        name: Setting up Java 19 environment...
        with:
          distribution: 'corretto'
          java-version: 19

      - name: Building project...
        run: ./gradlew build

      - name: Assembling Release Bundle...
        run: ./gradlew assembleRelease

      - uses: r0adkll/sign-android-release@v1
        name: Signing APK...
        id: signed_apk
        with:
          releaseDirectory: android/build/outputs/apk/release
          signingKeyBase64: ${{ secrets.ANDROID_SIGNING_KEY_BASE64 }}
          alias: ${{ secrets.ANDROID_KEYSTORE_ALIAS }}
          keyStorePassword: ${{ secrets.ANDROID_KEYSTORE_PASSWORD }}
          keyPassword: ${{ secrets.ANDROID_KEY_PASSWORD }}

      - uses: wzieba/Firebase-Distribution-Github-Action@v1
        name: Uploading APK to Firebase App Distribution...
        with:
          appId: ${{ secrets.FIREBASE_ANDROID_APP_ID }}
          serviceCredentialsFileContent: ${{ secrets.GC_CREDENTIAL_FILE_CONTENT }}
          groups: internal
          file: ${{ steps.sign_app.outputs.signed_apk }}

I'm just trying to create a release APK, sign it, and upload it to Firebase App distribution Any thoughts on what might be going wrong here?

I could have sworn that okhttp had a recipes section on the wiki a while back that included recipes for interceptors. Now it seems like the okhttp website has no interceptors on its recipes page... but on its interceptors page it does have an example of writing one, but not in kotlin. I guess what im asking is, are there common kotlin interceptor examples anywhere?

Yeah. Those are the web pages I was talking about in my comment. recipes page has no interceptor recipes. and the interceptor page doesn't have any kotlin snippets. But at least it sounds like i wasn't missing any other examples. 👍

just used wayback machine and can see that the github.io pages above are essentially identical to the wiki. so i guess i mustve been dreaming when i thought there were kotlin specific interceptor samples. ¯\_(ツ)_/¯