10.10.1  Overview

The Mythryl type system differs from that of older languages like C++ and Java in two important ways:

• Type inference: The Mythryl compiler uses unification-driven inference to deduce the types of variables and functions wherever possible. Consequently it is rarely actually necessary for the programmer to explicitly specify types outside of API declarations. This dramatically improves code compactness and readability.
• Parametric Type Polymorphism: Many functions and sumtypes do not actually care about the types of some of their arguments.

  Older languages nevertheless require them to be declared with specific types, limiting re-use of such functions and sumtypes.

  Mythryl’s Hindley-Milner type system not only allows such values to be explicitly declared as don’t-cares, but can also almost always automatically compute a most general type for such functions and datastructures, substantially improving re-usability of both functions and sumtypes.

In C++ and Java it is routine to use type casts to defeat the compiler type checker; this is frequently necessary in order to work around the limitations of their type systems.

The enhanced expressiveness of the Mythryl type system means it is almost never necessary to defeat the compiler typechecker in this way; in fact, the language does not even define a type cast operator for doing so. This leads to code which is enormously more typesafe, robust and maintainable than similar code written in legacy languages. It is routine for C programs to dump core when first run after significant maintenance edits; Mythryl programs typically run correctly again the first time they compile.