Making sense of initialization order in Scala

on YouTube ↗31m

" summary

Demonstrates that Scala’s trait linearization and compiler-generated trait setters cause fields to be observed with default values or discarded values in surprising cases, with concrete examples that print zeros instead of expected constants; shows Dotty’s alternative where Dotty uses per-field init methods so fields can be final and better optimized; evaluates workarounds: lazy vals cost synchronization and can deadlock and early definitions can break assumptions; proposes forcing lazy-like init only during construction and reports Dotty experiments that exposed real code breakage.

AI-written summary · about

" chapters (6)

1. 0:00Start
2. 6:07TRICKY EXAMPLES
3. 9:09SPECIFICATION: LINEARIZATION ORDER
4. 22:28ALTERNATIVE: EARLY DEFINITIONS
5. 25:06DOES NOT SOLVE THE PROBLEM OF ACCESSING PARTIALLY OR NOT INITIALIZED FIELDS
6. 25:42A CRAZY IDEA

" related talks

• 45mScala 3, Type Inference and You!Guillaume Martres · Scala World 2019
• 49mFrom DOT to DottyMartin Odersky · Voxxed Days CERN 2017
• 30mDotty and types: the story so farGuillaume Martres · Typelevel Summit Oslo 2016
• 39mA few tips on modelling things in ScalaMateusz Kubuszok · Scala Love in the City 2021
• 45mDesign Patterns in ScalaPavel Fatin · JavaDay 2013
• 32mSingle Serving Values with Linear Types in ScalaJames Earl Douglas · Scale By the Bay 2021
• 01h 04mKeynote - What to Leave ImplicitMartin Odersky · Scala Days Copenhagen 2017
• 51mOpening Keynote: New Functional Constructs in Scala 3Martin Odersky · Scale By The Bay 2018