Add test coverage to the lexer

[jansorg]

This PR adds test coverage to the lexer.
It adds a parameterized lexer test to automatically execute lexer tests for src/test/data/ParserTest/*.nix. I prefer this approach over keeping a long list of @Test methods in sync with the files src/test/data/ParserTest/*.nix.

I'm reusing the set of Nix files of the ParserTest instead of copying all to another directory. Please let me know if you'd like a copy instead or if I should rename ParserTest to something else.

The first commit adds the test and the needed infrastructure to make it work. The base lexer test class is relying on the test execution policy, which is now set up with the build system.
The other commit adds the reference data using the current lexer implementation.

My work is a courtesy of https://github.com/monogon-dev and a preparation for a PR to #39, which will need changes to the lexer.

[JojOatXGME]

Looks good. I am not sure if these tests were strictly necessary since ParsingTest also checks the tokens, but I am happy to merge this PR.

It adds a parameterized lexer test to automatically execute lexer tests for src/test/data/ParserTest/*.nix.

It seems that with this approach using a parameterized test, IntelliJ's UI does not support jumping from the node in the test report directly to the test data. Anyway, your approach still seems favorable.

[JojOatXGME]

Interesting, I am wondering if this could also replace ParsingTest.getTestDataPath() and ParsingFailTest.getTestDataPath().

nix-idea/src/test/java/org/nixos/idea/lang/ParsingTest.java

Lines 154 to 157 in a3e8ac0

	@Override
	protected String getTestDataPath() {
	return "src/test/testData";
	}

[jansorg]

That should work, I think. As far as I understand this is only working because it's using the current working dir (i.e. the root project directory) as default.

[jansorg]

Thanks for merging!
In my experience separate tests for the lexer are very helpful when you're changing the .flex file without touching the parser .bnf file. A clean list of tokens instead of a nested tree of nodes and tokens is easier to understand in such a case.