Hello everyone, I have read this article about performance of `Koin` Ref: <https://proandroiddev.com...

# koin

[Sam]

Hello everyone,
I have read this article about performance of Koin
Ref: https://proandroiddev.com/how-dagger-hilt-and-koin-differ-under-the-hood-c3be1a2959d7

Im curious, How can I calculate the performance of Koin in Runtime like this articles?
Thanks.

[eurycea]

Have you looked into the project that the article references? https://github.com/Sloy/android-dependency-injection-performance

[arnaud.giuliani]

Take in mind that resolving a Koin instance key is around 0.01 ms. All around is a mater of creating the right instances.

[Sam]

Thanks @eurycea @arnaud.giuliani.

@arnaud.giuliani You mean this repo is creating the whole keys in the same time => take more time for startup.

[arnaud.giuliani]

at startup Koin is just creating all the optimized index key to retrieve a definition instance

[arnaud.giuliani]

it just take a few ms, everything is lazy by default

[Sam]

oh I see the startup time / setup koin <https://github.com/Sloy/android-dependency-injection-performance> over 0,01 ms. Maybe it is not implement correct way.

Some people argue that Koin is not good for startup time , so I would like to check this information.
Do you have any best practices link or repository to test the performance in runtime?

[arnaud.giuliani]

over 0,01 ms. Maybe it is not implement correct way.

this is the hashmap resolution. There is also a bit around stuff to check before. A resolved instance is around clearly less 0.05 < 0.1 ms. But it scales. requesting in hashmap is effective

[arnaud.giuliani]

startup time has been heavily reworked in 3.1.x, as the indexes are more quickly computed

[arnaud.giuliani]

but yes, we need to declare things at your startup to be able to fetch them

[arnaud.giuliani]

The benchmark doesn’t help really compare the real stuff finally. What’s your time when you “inject” with Dagger by calling your builder. And finally … does it really matter when you request your instance at 0.05 ms?

[Sam]

oh yeah!, I think its doesn’t matter when I request a instance in 0,05ms.

My project is really big, so Im using Koin 3.x now,
How do I calculate the timing of these cases such as startup time, factory, singleton/ inject etc?

I need data to report from my research and make a decision to continue using Koin. Some team mems are trusting the above article to switch to Hilt.

Thanks.

[arnaud.giuliani]

ok, intersting

[arnaud.giuliani]

How do I calculate the timing of these cases such as startup time, factory, singleton/ inject etc?

you can activate logger debug, and track times. androidLogger(DEBUG) in your startKoin block

[Sam]

Thanks @arnaud.giuliani

[Mani]

Hi @Sam
Were you able to calculate the results, we are also discussing about adding Hilt or Koin in our project, but not yet decided.

[arnaud.giuliani]

In terms of performances, Koin is resolving as fast as Dagger/Hilt is pushing. Instance access is around 0.05 ms

[arnaud.giuliani]

With Koin you remove all the kapt side and DI config is very minimal