Sure. Sorry, I misspoke. I can achieve the behavior I want with moleculeFlow, but not with launchMolecule. I want it with a StateFlow because I want to remember the last value upon te-subscription AND stop processing the lambda when there are no subscribers

Maybe I want to use moleculeFlow + stateIn with a SharingStarted..

You still want

moleculeFlow

then but need to apply some operator for caching the last item

Yeah

Yeah makes sense. Thanks Jake

Hello there! I am wondering if wire-compiler has any plan to support adding prefixes to the generated Kotlin classes. Since Kotlin/Native does not support framework modularization for Swift/ObjC, wire-generated Kotlin classes can cause name collision with my own classes. When the protobuf schema defines a

User

message, I cannot define another

User

class in my own module without the K/N compiler suffixing an underscore to either of them (

User_

) in the generated ObjC framework. I even have a number protobuf schemas for different endpoints, that leads to multiple underscores (

User___

) in the suffix when they share a same name. That’s why I want to make the naming persistent by specifying a custom prefix for each wire-compiler invocation, or letting it decide one from package names or something.

You could also use @ObjCName as a workaround: https://kotlinlang.org/api/latest/jvm/stdlib/kotlin.native/-obj-c-name/

Thanks, I’ve also considered using @ObjCName. I think I cannot actually workaround the name collision among wire-generated classes though, like the

User___

example above. Is there any way to annotate wire classes?

Hi! I’m new to Kotlin and I’m trying to figure out why this isn’t working. I’m trying to write a repository layer that uses a SQLDelight database to return a flow of data…but it seems like I’m missing a

context

parameter. I want it to use the caller’s context (unless that’s unsafe). I can’t find any examples where this needs to be specified, but it’s not working here.

That operator actually performs the DB queries, do you probably want to pass Dispatchers.IO or equivalent

The callers context is still used, but it's almost guaranteed to not have the correct dispatcher on which you want queries to run so this basically does a withContext using the supplied dispatcher

Ah I see. Why can some callers omit the

context

parameter? Is that passed invisibly through some syntactic sugar in some situations? Or are those calling a different extension?

Is there a way to "preprocess" a BufferredSource before handing it to KotlinX Serialization to remove all invalid line feeds?

Sure. Write a Source that removes them, wrap it around yours, then call buffer to get back to a BufferedSource

Sources and Sinks are meant to be layered/decorated

Do I have to implement

Source

(or

BufferedSource

?) or is there some higher-level api I could work with?

Source is what you want. You read from the underlying one into a Buffer and then process it before writing to your Buffer. Presumably you'd do a loop to find the characters and then issue writes of the bytes that comes before one to your caller

I'm using SQLDelight. Can I use enum constant in the query, so it'll be converted with the adapter?

CREATE TABLE SomeTable(
  enumColumn TEXT AS EnumClass
);


update:
UPDATE SomeTable
SET enumColumn = EnumClass.EnumItem;

I can use argument instead of

EnumClass.EnumItem

, but my query only makes sense with single enum constant.

That's not supported, and supporting would be complex. The SQL parser, which validates the SQL input before the code generator generates Kotlin code, does not look into Kotlin sources, so it does not know if EnumClass a) is a valid Kotlin enum and b) does not know the possible entries. Extreme case: EnumClass could be defined in a dependency.

Got a few SQLDelight questions, I'm following the docs on migrations (https://cashapp.github.io/sqldelight/2.0.0-alpha05/android_sqlite/migrations/) and it says there should be a task

generateSqlDelightSchema

but I'm seeing

generateReleaseMainDatabaseSchema

and

generateDebugMainDatabaseSchema

. "MainDatabase" is the configured name of my database, so that makes sense, but why are there "Release" and "Debug" versions of the task?

Presumably because you are in an Android project which has two variants made of up a release and debug build type

Yes. Do they do anything different if my .sq files are in the main source set?

Nope

Second, I'm migrating from another library, and the database is many schema versions in. Is there a way to specify the starting schema version?

I’m trying to pass

emptyList()

to

.defaultValue("%L", defaultValue)

but it generate an array instead :

public val myList: List<String> = []

which fails. How can I fix it to generate

public val myList: List<String> = emptyList()

instead?

I have an OkHttp Interceptor that makes another request before the original request can proceed (think authentication). Sometimes I make many requests at once, exceeding the

maxRequestsPerHost

in the Dispatcher, meaning the Interceptor cannot make it's request, because

maxRequestsPerHost

is already at limit. No request can finish, because they are waiting at the Interceptor, and I have a deadlock. This seems like a problem that should be pretty common, so what is the proper solution here? • Make the inner request with a different Dispatcher? • Somehow make sure I don't reach my

maxRequestsPerHost

, so there is always room for the inner request? • Move the inner request out of the Interceptor somehow? (Maybe there is some mechanism I don't know about where I can hook into the request before it's counted in

maxRequestsPerHost

)?

Make a blocking call

They don’t count against those limits

Are there any docs you could point me to? I'm making all my calls with Retrofit so I'm not creating the calls myself, Retrofit is.

Or is that just a matter of using blocking APIs of Retrofit (as opposed to

suspend

functions I currently have)?