Fix vect parsing with newline before comma

[pfitzseb]

Fixes #59.

[oscardssmith]

Looks good. Merge?

[pfitzseb]

I'll merge this tomorrow morning, unless Chis chimes in and says this is the wrong fix :)

The test is currently a bit awkward, because SyntaxNode/Expr conversion can't handle unexpected ,s. Would like to fix that.

[c42f]

Looks good thanks. Let's add such error cases to the tests as s-expressions along with the happy cases.

[c42f]

The test is currently a bit awkward, because SyntaxNode/Expr conversion can't handle unexpected

Ah yeah I didn't see this comment before I wrote the review. Precisely :)

[pfitzseb]

Tests look a lot nicer now. We're now just emitting an ErrorVal when encountering unknown node kinds, which may be a bad idea, but at least allows us to recover from the corrupted tree we produce further down :P

[pfitzseb]

This good to go, or do you think emitting ErrorVals instead of actually erroring is going to be an issue further down the line?

[c42f]

The FIXME with ErrorVal makes me a bit uncomfortable (does seem wrong). But this is still an improvement so I think we can merge and fix that later.