gilch
gilch
The new REPL and incremental lexer from #62 have much improved read-time error messages.
I feel like there's still room for improvement in the reader error messages. It might need a trace decorator like the compiler.
I have at least covered all of my lines now. That only means they were *run* by the tests, not that the tests actually checked everything (but at least they...
I'm also considering full type annotations and CrossHair contracts. That's probably about the best we can do short of rewriting it in Idris and proving theorems.
The REPL already does this, but you can still see the traceback for SyntaxErrors when running a main module, and maybe when compiling one.
The way the REPL suppresses these makes them incompatible with doctests. It's easy enough to work around, but not ideal.
I'm not sure if Clojure's behavior is the best target then. Besides the mentioned namespace expansion of symbols, Clojure can also syntax-quote vectors, maps, and sets, which MAL doesn't have...
I found a fairly simple-looking Scheme quasiquote implementation (in Scheme) in [Alan Bawden's "Quasiquotation in Lisp"](http://repository.readscheme.org/ftp/papers/pepm99/bawden.pdf), which the author asserts is correct. It shouldn't be too difficult to port it...
Chicken Scheme's [quasiquote test cases.](https://github.com/sergi/chicken-core/blob/39b1a129cddb4f39279b39396556112a444d3267/tests/syntax-tests.scm#L932)
We might end up going the other way with this. See #1324.