Change TestCase from trait objects to enum

Previously, TestCase was a trait, and features were composed of boxed TestCase
trait objects. This changes it to an enum, since we expect to have only a small
number of TestCase types, and users are not expected to provide new ones.

This also changes the module layout; previously, the scenario and feature
parsers and AST structs were each in their own modules. This makes less sense
when using the enum at the TestCase level. Now all the AST structs are in the
AST module, and the parsers are in the parser module.

src/ast.rs

@@ -0,0 +1,77 @@
+/// A test context is used to pass state between different steps of a test case.
+/// It may also be initialized at the feature level via a Background (TODO)
+pub trait TestContext {
+    fn new() -> Self;
+}
+
+
+pub enum TestCase<C: TestContext> {
+    Background(Scenario<C>),
+    Scenario(Scenario<C>),
+}
+
+impl<C: TestContext> TestCase<C> {
+    pub fn name(&self) -> String {
+        match self {
+            TestCase::Background(s) => s.name.clone(),
+            TestCase::Scenario(s) => s.name.clone()
+        }
+    }
+
+
+    pub fn eval(&self, context: &mut C) -> bool {
+        match self {
+            TestCase::Background(s) => s.eval(context),
+            TestCase::Scenario(s) => s.eval(context)
+        }
+    }
+}
+
+/// A feature is a collection of test cases.
+pub struct Feature<C: TestContext> {
+    pub name: String,
+    pub comment: String,
+    pub test_cases: Vec<TestCase<C>>,
+}
+
+impl<C: TestContext> Feature<C> {
+    pub fn eval(&self) -> (bool, C) {
+        let mut context = C::new();
+
+        for tc in self.test_cases.iter() {
+            let pass = tc.eval(&mut context);
+
+            if !pass {
+                return (false, context);
+            }
+        }
+
+        (true, context)
+    }
+}
+
+pub struct Scenario<TC: TestContext> {
+    pub name: String,
+    pub steps: Vec<Box<Step<TC>>>,
+}
+
+impl<C: TestContext> Scenario<C> {
+    /// Execute a scenario by running each step in order, with mutable access to
+    /// the context.
+    pub fn eval(&self, context: &mut C) -> bool {
+        for s in self.steps.iter() {
+            if !s.eval(context) {
+                return false;
+            }
+        }
+
+        true
+    }
+}
+
+/// A specific step which makes up a scenario. Users should create their own
+/// implementations of this trait, which are returned by their step parsers.
+pub trait Step<C: TestContext> {
+    fn eval(&self, &mut C) -> bool;
+}
+

src/lib.rs

@@ -3,8 +3,8 @@ extern crate combine;
 
 extern crate itertools;
 
-pub mod feature;
-pub mod scenario;
+pub mod parser;
+pub mod ast;
 pub mod parse_utils;