Within the check phase of type check, we check each item one at a time (bodies of function expressions are checked as part of the containing function). Inference is used to supply types wherever they are unknown.
By far the most complex case is checking the body of a function. This can be broken down into several distinct phases:
gather: creates type variables to represent the type of each local variable and pattern binding.
main: the main pass does the lion’s share of the work: it determines the types of all expressions, resolves methods, checks for most invalid conditions, and so forth. In some cases, where a type is unknown, it may create a type or region variable and use that as the type of an expression.
In the process of checking, various constraints will be placed on these type variables through the subtyping relationships requested through the
infermodule is in charge of resolving those constraints.
regionck: after main is complete, the regionck pass goes over all types looking for regions and making sure that they did not escape into places where they are not in scope. This may also influence the final assignments of the various region variables if there is some flexibility.
writeback: writes the final types within a function body, replacing type variables with their final inferred types. These final types are written into the
tcx.node_typestable, which should never contain any reference to a type variable.
While type checking a function, the intermediate types for the
expressions, blocks, and so forth contained within the function are
fcx.node_args. These types
may contain unresolved type variables. After type checking is
complete, the functions in the writeback module are used to take the
types from this table, resolve them, and then write them into their
permanent home in the type context
This means that during inferencing you should use
fcx.node_ty() to write/obtain the types of
nodes within the function.
The types of top-level items, which never contain unbound type
variables, are stored directly into the
N.B., a type variable is not the same thing as a type parameter. A
type variable is an instance of a type parameter. That is,
given a generic function
fn foo<T>(t: T), while checking the
foo, the type
ty_param(0) refers to the type
is treated in abstract. However, when
foo() is called,
T will be
substituted for a fresh type variable
N. This variable will
eventually be resolved to some concrete type (which might itself be
a type parameter).
- check 🔒
- entry 🔒
- Type-checking for the rust-intrinsic and platform-intrinsic intrinsics that the compiler exposes.
- region 🔒This file builds up the
ScopeTree, which describes the parent links in the region hierarchy.
- Emit an error when encountering two or more non-zero-sized fields in a transparent enum.
- Emit an error when encountering two or more variants in a transparent enum.
ty::GenericPredicatesin a way suitable to be used in structured suggestions.
- Forbid defining intrinsics in Rust code, as they must always be defined by the compiler.
- Return placeholder code for the given function.
- Given a
DefIdfor an opaque type in return position, find its parent item’s return expressions.
- Return placeholder code for the given associated item. Similar to
ty::AssocItem::suggestion, but appropriate for use as the code snippet of a structured suggestion.