During typechecking, Unison substitutes free variables in an expression by looking them up in an environment populated from a codebase of available definitions. A Unison codebase is a database of term and type definitions, indexed by hashes and names.
A name in the environment can refer to either terms or types, or both (a type name can never be confused with a term name).
Suffix-based name resolution
If the list of segments of a name (''base.List.map'' has the segments [base,List,map]) is a suffix of exactly one fully qualified name in the environment, Unison substitutes that name in the expression with a reference to the definition. For example, the fully qualified name .base.List.map could be referenced via base.List.map, List.map as long as no other definitions end in base.List.map or List.map. This reduces the number of imports needed and cuts down on needing to remember the fully qualified names for definitions. ("Was it base.List.map or util.List.map?")
Hash literals in the program are substituted with references to the definitions in the environment whose hashes they match.
If a free term variable in the program cannot be found in the environment and is not the name of another term in scope in the program itself, or if an free variable matches more than one name (it's ambiguous), Unison tries type-directed name resolution.
Type-directed name resolution
During typechecking, if Unison encounters a free term variable that is not a term name in the environment, Unison attempts type-directed name resolution, which:
- Finds term definitions in the environment whose unqualified name is the same as the free variable.
- If exactly one of those terms has a type that conforms to the expected type of the variable (the type system has always inferred this type already at this point), perform that substitution and resume typechecking.
If name resolution is unable to find the definition of a name, or is unable to disambiguate an ambiguous name, Unison reports an error.