Factory 2.3.0 Almost Here! Help Needed!

[hmlongco]

The next version of Factory is in the develop branch, and brings significant performance enhancements:

• Debug execution speeds are up to 2x faster!
• Production execution speeds are up to 4x faster!

Better yet, the improvements needed are pretty much internal to the library, and as such I don't believe anyone should see any breaking changes related to this update.

That said, the changes are significant. Factory has 100% code coverage and everything passes with flying colors, but I'd like to make sure there are no problems before I push it to main and Cocoapods.

So I'd like the more adventurous of you to do me a favor: Try pulling 2.3.0 (develop) and trying it out in your existing iOS and Mac and other applications. If you find a problem, don't hesitate to open an issue and let me know.

I'd appreciate it a lot.

Until then, enjoy!

[hmlongco]

Note that there is a potential breaking change related to issue #146. See that thread and the CHANGELOG for more.

[hmlongco]

The following Medium article explains the changes in more detail.
https://michaellong.medium.com/high-peformance-dictionary-keys-370dd0b525ac

[doozMen]

Hei an overdue test that was on my todo for a long time. First of all when you mentioned I performance boost I was wondering how you achieved that and what would be the bottle neck. As we discussed before I have a large design system with over 1000+ tokens that make up a design system for over 30 EU newspapers. This is in production and we rely for swift (and in the future Kotlin too) on Factory to resolve the keys to its correct values. So far a big success!

So you can imagine if you talk performance that this interests me. I read your article on medium and did the test with linking to develop in our project.

First the test to develop. All worked just fine and I just needed to change String into StaticString at some points but that was it!

After reading the article I due have a question about the dictionary performance that I will try to explain with how I understood it. So lets go. A dictionary lookup with a hash is fast when it can use the hash and does not have to use equate? In your change you use the type of the factory registration as a hash, not the key which is the property name. So if I understood correctly if 2 properties have the same type the hash would be the same?

extension Container {
  var foo: Factory<String> {
    self { "foo" }
  }
  var bar: Factory<String> {
    self { "bar" }
 }
}

So in the above code the performance increase compared to 2.2 would be less, but still significant, compared to 2.3 because in this case it will have to use compare?

[hmlongco]

How often do you override your 1,000 strings? Factory uses dictionary lookups primarily to see if any registrations or caching have occurred. Given the above code, both of those would probably be dictionary lookup misses with no string comparisons needed.

2.3.0 should still be much faster since key creation is much faster, key hashing does't have to hash a unicode-based string, and key comparison doesn't need to compare a unicode-based string.

Keep in mind that these things are still incredibly fast. I typically have to do millions of lookups to get variations in timing that differ in hundredths of a second.

[doozMen]

Nice good to know! And almost never they get updated. Just in a brand switching and subtheming context. So it happens in a scrollview for a article that needs to be styled breaking for example.
And just to be clear they are not strings that was just an example.

[doozMen]

Do you have a time in mind for the release of 2.3?

[hmlongco]

Probably push this weekend. Day job has been a little hectic.

[hmlongco]

Attempting a pod trunk register email so I can push new changes to Factory and when I click on the email link I get an Internal Server Error. Trying again gives "An unhandled lowlevel error occurred. The application logs may have details."

Cocoapod's status page reports no incidents... but also has no way to report an incident.