Handle things separated by things differently

I noticed that programs like '{} would parse correctly while '{ } would
expect an inner element. This was because the leading space was actually
part of the element parser, which is a violation of the (as of yet
unspoken) rule that parsers should not parse surrounding whitespace.

Because whitespace whas treated differently from everywhere else and
because this implementation was wrong, I decided to reimplement it,
abstracting the concept of things separated by other things with
optional trailing things. I did this in such a way that surrounding
whitespace is not touched.

src/parser/basic.rs

@@ -3,7 +3,7 @@
 use chumsky::prelude::*;
 use chumsky::text::Character;
 
-use crate::ast::{Ident, Line, Space};
+use crate::ast::{Ident, Line, Separated, Space};
 use crate::span::Span;
 
 pub type Error = Simple<char, Span>;
@@ -58,3 +58,27 @@ pub fn ident() -> EParser<Ident> {
 pub fn local(space: EParser<Space>) -> EParser<Option<Space>> {
     text::keyword("local").ignore_then(space).or_not().boxed()
 }
+
+// This function is more of a utility function. Because of this and to keep the
+// code nicer, I have decided that the rules specified in the `parser` module
+// don't apply to it.
+pub fn separated_by<E: 'static, S1: 'static, S2: 'static>(
+    elem: impl Parser<char, E, Error = Error> + Clone + 'static,
+    separator: impl Parser<char, S1, Error = Error> + 'static,
+    trailing_separator: impl Parser<char, S2, Error = Error> + 'static,
+) -> EParser<Separated<E, S1, S2>> {
+    elem.clone()
+        .then(separator.then(elem).repeated())
+        .then(trailing_separator.or_not())
+        .or_not()
+        .map_with_span(|s, span| match s {
+            Some(((first_elem, last_elems), trailing)) => Separated::NonEmpty {
+                first_elem,
+                last_elems,
+                trailing,
+                span,
+            },
+            None => Separated::Empty(span),
+        })
+        .boxed()
+}