Let's Make a Generic Inference Algorithm - Ryan Cavanaugh, TypeScript Congress 2023

Discover how to create a generic inference algorithm that infers types in TypeScript, considering context, type variance, and more, with a focus on providing a good enough answer even in ambiguous cases.

Key takeaways

The goal is to create a generic inference algorithm that can infer types in TypeScript.
The algorithm should consider the context and relate input types to output types.
Type inference is not always about finding the correct answer, but rather the best answer that makes sense.
The algorithm involves collecting candidates, prioritizing them, and choosing the best one.
The best answer is often the one that contains all the other candidates.
The algorithm should consider type variance and the relationship between input and output types.
Type inference is not always perfect and may require additional information.
The algorithm should be able to handle ambiguous cases and provide a good enough answer.
The algorithm should also consider the motivating examples and the intended behavior of the code.
The use of any and unknown types can be useful in certain situations.
Type inference is a complex process that requires consideration of many factors.
The algorithm should be able to handle multiple type arguments and union types.
The concept of “best” is subjective and may require additional information to determine.
Type inference is not always about finding the most specific type, but rather the best type that makes sense.
The algorithm should be able to handle cases where there are multiple candidates and no clear best answer.
The algorithm should be able to consider the context and relate input types to output types.
Type inference is a trade-off between correctness and practicality.
The algorithm should be able to handle cases where there are no type annotations and inferring types is necessary.
Type inference is a complex process that requires consideration of many factors and is not always perfect.
The algorithm should be able to provide a good enough answer even in the face of ambiguity.
The algorithm should be able to handle cases where there are multiple type arguments and union types.
The use of box and unbox can be useful in certain situations.
Type inference is a complex process that requires consideration of many factors and is not always perfect.

Let's Make a Generic Inference Algorithm - Ryan Cavanaugh, TypeScript Congress 2023

More talks