yep

I usually try to keep it to more idiomatic Kotlin instead of mimicking Groovy, but I get what you're saying. I use convention plugins for almost everything (although I use Kotlin precompiled script plugins 😬) which results in most of my module's gradle files looking like that screenshot. Maybe I'll check out converting those to Groovy and see if there's any performance increase, although with the config cache most of my builds are super fast nowadays.

there's not likely to be a reason to switch back to Groovy (although it should be slightly faster), however, that is a valid Groovy file as well so we can "just" make our projects look like that too. Plus the IDE will correctly autocomplete everything in there even in Groovy.

Plugging into the Gradle Matrix — John Rodriguez & Tony Robalik There's a part @ 28:10 about Kotlin DSL being meh due to build scripts compiling poorly, which I guess can have a significant impact at scale. With conventions and a strict simple script structure the difference between Groovy and Kotlin should be negligible (in terms of writing and maintaining them).

In a multiplatform project when I use this construct

databaseQueries
            .selectAllItems {
                // mapper
            }
            .asFlow()
            .mapToList()
            .flowOn(dispatcher)

I find that

mapToList

can be concurrently executed by flow collections, and causes

android.database.sqlite.SQLiteBlobTooBigException

by with cursor filled to maximum. Is there any API option to prevent this? I changed the construct to

.asFlow()
            .map { query ->
                mutex.withLock {
                    withContext(dispatcher) {
                        query.executeAsList()
                    }
                }
            }
            .flowOn(dispatcher)

and this fixes the exception. Is this the common way to fix such a problem?

are you certain this solves your problem?

mapToList()

calls the same

executeAsList

function so you should see the same result

the bigger question is: what are you storing in your table?

I store a coordinate list of a map route, which can unfortunately long as 30k :-(

Sounds like you need to page your data

I have exception when

executeAsList

is entered concurrently from different threads. I feel its thread safe, but quickly fills curser window size.

I added a

mutex

to prevent concurrent execution of

executeAsList

If you page your data and only request what’s needed you likely won’t be filling up the cursor as much

even by synchronizing your cursor reads, you aren't actually solving the root problem that your data is larger than what sqlite can handle

I need the coordinates list as a whole always, Is paging the right way to store such a big blob? Or should I store this blob outside of sqlite even?

yea I'd recommend moving it out. this sounds no different than storing potentially large images in your sql table.

For geo data like that your best options are to either store it as a file and save a pointer to that file in the database, or store each coordinate pair along the line as a separate row (similar to what SpatiaLite or a geopackage would do). But unless you need to perform spatial queries on the data, saving it in a separate file is best.

Ok, I will check how this can be done, I guess file access in multiplatform can be tricky.

store each coordinate pair along the line as a separate row (similar to what SpatiaLite or a geopackage would do).

That’s what I thought

selectAllItems

did 🥴

If 30K lat/lng coordinates are living in a single row this sounds like a perfect use case for a foreign key on every lat/lng pair

I’d like to endorse build.gradle.kts as my preference. Being able to discover features by typing

this.

in any scope is fantastic. I’ll convert every cashapp project that they let me.

anywhere in Groovy where you can't type

this

and see the same thing as Kotlin is a bug report that should go to Gradle for a missing

Action

overload that's still using

Closure

👋 was taking a look into molecule and the whole concept it brings. Its quite interesting altho Im wodnering now that you dont have AAC viewmodels how are poeple dealign with config changes? scoping and caching on the data layer? Would be great to hear some alternatives ppl have come up with

To try Compose for iOS I needed to set

isStatic = true

, which results in errors similar to the ones mentioned in this post: https://github.com/cashapp/sqldelight/issues/2512 Can you confirm that right now SQLDelight can't be used in a static XCFramework? Is there explanation why? Does that mean I need to put Compose into a separate XCFramework?

For those using Molecule for KMP/KMM that targets iOS, are you specifying a

MonotonicFrameClock

for iOS?

When using retrofit with coroutines, the suspend functions return the actual data. Is there a guide on how to access headers when using this approach? Any guides would be appreciated.

Trying to write a test for some of my networking code (pure okhttp at this point. no retrofit). id like to write a test with okhttp mockwebserver to simulate what happens when I go into airplane mode in the middle of a request (IOException). I know I can use mockwebserver to send back things like 500s, but is mockwebserver the right place to try to simulate my device losing connectivity in the middle of a request?

yes. take a look at https://square.github.io/okhttp/3.x/mockwebserver/okhttp3/mockwebserver/SocketPolicy.html

about half of those will cause an IOException to be thrown. you probably want

DISCONNECT_AFTER_REQUEST

or one of the

DISCONNECT_DURING_*

ones

NO_RESPONSE

and

STALL_SOCKET_AT_START

are good for testing timeouts

If I have a need to run a JS environment on mobile to interact with a particular JS library, but don't currently need OTA updates, is Zipline still a good choice? How heavy is the OTA machinery if I'm not using it?