Look through the base types of the found type for any which might have been valid subtypes if given conformant type arguments.
Look through the base types of the found type for any which might have been valid subtypes if given conformant type arguments. Examine those for situations where the type error would have been eliminated if the variance were different. In such cases, append an additional explanatory message.
TODO: handle type aliases better.
It can be quite difficult to know which of the many functions called "error" is being called at any given point in the compiler.
It can be quite difficult to know which of the many functions called "error" is being called at any given point in the compiler. To alleviate this I am renaming such functions inside this trait based on where it originated.
Does the positioned line assigned to t1 precede that of t2?
The common situation of making sure nothing is erroneous could be nicer if Symbols, Types, and Trees all implemented some common interface in which isErroneous and similar would be placed.
For errors which are artifacts of the implementation: such messages indicate that the restriction may be lifted in the future.
Given any number of types, alters the name information in the symbols until they can be distinguished from one another: then executes the given code.
Given any number of types, alters the name information in the symbols until they can be distinguished from one another: then executes the given code. The names are restored and the result is returned.
An interface to enable higher configurability of diagnostic messages regarding type errors. This is barely a beginning as error messages are distributed far and wide across the codebase. The plan is to partition error messages into some broad groups and provide some mechanism for being more or less verbose on a selective basis. Possible groups include such examples as
arity errors kind errors variance errors ambiguity errors volatility/stability errors implementation restrictions
And more, and there is plenty of overlap, so it'll be a process.
1.0