On what kind of machine are you considering these timings?

In most cases, Ints are not possible for us because we have sealed classes (aka enums with associated values).

Our alternative is interface + multiple implementations + multibinds. We trade compile-time exhaustivity and safety for scalability. You forget a @IntoSet a nobody will notice…

bsky.app/profile/cyri...

I actually wasn't talking about runtime performance. I was mostly referring to codebase scalability, modularization and build times.

Enums with 50+ variants all having different variants fields (which end up as sealed classes in Kotlin) pointing to symbols located in multiple modules in the code.

No I'm not talking about runtime performance. I'm talking about codebase scalability, modularization and build times.
Runtime performance wise, in our case, it's not a problem at all compared to everything else we do 😅

Thanks to our savior: R8

Personal today's example: searching for the longest dependency chains - hence potential critical paths - in a build system-based directed acyclic graph and being 😲 by the result.

It was the theme of the year. Carnival

It is! Not the classical bread of course but it's bread! Italy was there too. Their craft is 🤌

For example, here’s today’s snapshot from our project: only 4 modules are massive in LOC—and surprise, they all build in >15s on CI.
Now that we’ve measured, we can target fixes and optimize! Data drives action.

Too late. I judge you. 😜

I can't agree more. It is so true. A lot of work has been done because it was definitely the opposite back in the days.

4️⃣Educate & guide
Teach through subtle, intuitive cues.

5️⃣Create immersive experiences
Pull users into the moment.