Custom TSLint rule to check function parameter assignability (no bivariance)

Question

I just stumbled upon the fact that TypeScript isn't quite strict checking the assignability of functions: https://www.typescriptlang.org/docs/handbook/type-compatibility.html#function-parameter-bivariance

Unfortunately, for some patterns parameter bivariance misses important type checking. So I'm wondering whether it would be possible to build a custom TSLint rule telling me when I'm doing something like this:

interface Base {}
interface BaseEx extends Base { x; }

let fn1: (a: Base) => void;
let fn2: (b: BaseEx) => void;

fn1 = fn2; // TSLint: parameter (a: BaseEx) is not assignable to (b: Base)

However, documentation on creating custom TSLint rules seems rather incomplete, I only found a single example of a purely syntactical check. I would be really happy if you could advise me a resource to learn how to extend TSLint with semantic rules like this one.

Answer 1

When looking to implement a custom rule, the TSLint source code is a useful resource for guidance. The source for all of the built-in rules is available in the TSLint repo . Most of the rules to not require access to type information. However, there are two that do:

• the no-for-in-array rule; and
• the restrict-plus-operands rule.

Those rules use the ProgramAwareRuleWalker , which makes type information available to the rule via the TypeChecker .

There is some information on how the TypeChecker can be used in the TypeScript Compiler API documenation .

If rules that use the ProgramAwareRuleWalker are enabled, TSLint must be run with the --type-check option and a --project must be specified, too.