Test typescript types using tsd

[hildjj]

This does not yet have full coverage over all of the types in peg.d.ts;
I want to see if this approach seems viable to the typescript experts.

[hildjj]

Note to future self: to generate the package-lock.json file with an older npm (to keep lockfileVersion: 1):

docker run --rm -it -v $PWD:/app --workdir /app node:12.22.1 npm install --package-lock-only

[hildjj]

Fixes #155

[Mingun]

It seems good, but I'm prefer to split one commit into several. In the current state it changes too much weakly connected things. I suggest to split it into:

• update dependencies
• change format of package-lock.json
• switch jest from command-line switches to the config file
• fix SyntaxError inheritance
• add tsd and the stuff
• update minified files (you can include this step into commit that changed them)

in any order which would build on each step

[Mingun]

"rule.fail" also has a location.

Suggested change

	switch (event.type) {
	case "rule.enter":
	expectType<peggy.LocationRange>(event.location);
	break;
	case "rule.match":
	expectType<peggy.LocationRange>(event.location);
	expectType<any>(event.result);
	break;
	}
	expectType<peggy.LocationRange>(event.location);
	if (event.type === "rule.match") {
	expectType<any>(event.result);
	}

[Mingun]

Why second test doesn't take a grammarSource?

Suggested change

	const p2 = peggy.generate(src, { output: "source" });
	const p2 = peggy.generate(src, { output: "source", grammarSource: "src" });
	});

[hildjj]

Willfix, with something not a string.

[Mingun]

Suggested change

	function add(typ:string): void {
	function add(typ: string): void {

[Mingun]

It is necessary to check types of node fields? Hasn't the whole node test checked them out yet?

[hildjj]

What I'm trying to achieve is that we'll notice if we break backward-compatibility in the future.

[Mingun]

Makes sense

[Mingun]

You forgot to check a name, if that even needed:

Suggested change

	expectType<peggy.LocationRange>(node.location);
	expectType<peggy.LocationRange>(node.nameLocation);
	expectType<peggy.LocationRange>(node.location);
	expectType<string>(node.name);
	expectType<peggy.LocationRange>(node.nameLocation);

[Mingun]

If I understand correctly, that construction mean to run visit(grammar) and assert types with snapshot that was taken on the first run. However that snapshot should be committed and it seems it is not committed

[hildjj]

This is a mistake, left over from a previous attempt to use dts-jest. Will remove.

[hildjj]

@Mingun I think i fixed everything, but I'm interested in your opinion about being able to detect backward-compat breaks.

[hildjj]

@Mingun I think i fixed everything, but I'm interested in your opinion about being able to detect backward-compat breaks.

nm, just saw your comment. ok to merge?

[Mingun]

LGTM. One minor note: is it necessary to generate declaration files? Automatically-generated files seems very useless because almost all types inferred as any. Besides, we already has our own file.

[hildjj]

I agree that we don't need to generate .d.ts files. They will have more interesting things in them once we start documenting our interfaces jsdoc-style, but they're never going to generate something close enough to what we currently have to be backward-compatible.