Skip to main content

Julia Type Severity

In Julia you usually do not need to declare types for every value. The compiler tries to infer types based on how values are used in the code. In many cases this works well and results in efficient compiled code.

Sometimes the compiler cannot determine a precise type, or it ends up using a type that is less efficient. CodeGlass tries to detect these cases and highlights them so you can quickly see which functions use less efficient types.

For this feature we tried to follow the behavior of @code_warntype as closely as possible. We also use the same color scheme.

  • Purple indicates a high type severity.
  • Orange indicates a medium type severity.

These rules are not always perfect. Type inference can be complex, and in some cases the reported severity might not reflect a real performance problem.

The following rules are used to determine the severity level.

High Type Severity​

  • The type is abstract.
  • A union contains an abstract type.
  • A union contains more than three types.
  • (Named) tuples that contain:
    • a union
    • an abstract type
    • a variable length (for example varargs)
      If none of these are present, the tuple is not considered an issue.
  • Any parametric type (such as arrays or pairs) where one of the type parameters:
    • is abstract
    • or is a union.

Medium Type Severity​

  • A union with three or less options where all types are concrete.