diff --git a/justfile b/justfile
index a6f0b24..4bd1864 100644
--- a/justfile
+++ b/justfile
@@ -3,7 +3,7 @@ test:
     cargo test --features random
 
 watch-test:
-    watchexec -- "reset && just test"
+    watchexec -e rs -rc reset -- just test
 
 build:
     cargo build --all
@@ -18,4 +18,4 @@ watch-doc:
     watchexec -- "just doc && cargo test --all-features --doc"
 
 tree:
-    tree -I target
+    tree -I target
\ No newline at end of file
diff --git a/src/boundary.rs b/src/boundary.rs
new file mode 100644
index 0000000..1df82c8
--- /dev/null
+++ b/src/boundary.rs
@@ -0,0 +1,116 @@
+pub struct Boundary {
+    condition: fn(&str) -> bool,
+    start: usize,
+    len: usize,
+}
+
+impl Boundary {
+    // TODO maybe use graphemes here
+    pub const SPACE: Boundary = Boundary {
+        condition: |s| s.chars().nth(0) == Some(' '),
+        start: 0,
+        len: 1,
+    };
+    pub const HYPHEN: Boundary = Boundary {
+        condition: |s| s.chars().nth(0) == Some('-'),
+        start: 0,
+        len: 1,
+    };
+    pub const UNDERSCORE: Boundary = Boundary {
+        condition: |s| s.chars().nth(0) == Some('_'),
+        start: 0,
+        len: 1,
+    };
+    pub const LOWER_UPPER: Boundary = Boundary {
+        condition: |s| {
+            let mut chars = s.chars();
+            chars.next().map(|c| c.is_lowercase()).unwrap_or(false)
+                && chars.next().map(|c| c.is_uppercase()).unwrap_or(false)
+        },
+        start: 1,
+        len: 0,
+    };
+    pub const ACRONYM: Boundary = Boundary {
+        condition: |s| {
+            let mut chars = s.chars();
+            chars.next().map(|c| c.is_uppercase()).unwrap_or(false)
+                && chars.next().map(|c| c.is_uppercase()).unwrap_or(false)
+                && chars.next().map(|c| c.is_lowercase()).unwrap_or(false)
+        },
+        start: 1,
+        len: 0,
+    };
+
+    pub const fn default_delimiters() -> [Boundary; 3] {
+        [Boundary::SPACE, Boundary::HYPHEN, Boundary::UNDERSCORE]
+    }
+}
+
+// another idea for this algorithm
+// build an array of integers where
+// 0 means no split
+// 1 means the split is left of this char
+// 2 means this character is removed
+// then I can build the word at the end
+fn split<'s>(s: &'s str, boundaries: &[Boundary]) -> Vec<&'s str> {
+    let mut words = Vec::new();
+    let mut last_end = 0;
+    for i in 0..s.len() {
+        for boundary in boundaries {
+            if (boundary.condition)(&s[i..]) {
+                words.push(&s[last_end..i + boundary.start]);
+                last_end = i + boundary.start + boundary.len;
+                break;
+            }
+        }
+    }
+    words.push(&s[last_end..]);
+    words
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn hyphen() {
+        let s = "a-b-c";
+        let v = split(s, &[Boundary::HYPHEN]);
+        assert_eq!(v, vec!["a", "b", "c"]);
+    }
+
+    #[test]
+    fn underscore() {
+        let s = "a_b_c";
+        let v = split(s, &[Boundary::UNDERSCORE]);
+        assert_eq!(v, vec!["a", "b", "c"]);
+    }
+
+    #[test]
+    fn space() {
+        let s = "a b c";
+        let v = split(s, &[Boundary::SPACE]);
+        assert_eq!(v, vec!["a", "b", "c"]);
+    }
+
+    #[test]
+    fn delimiters() {
+        let s = "aaa-bbb_ccc ddd ddd-eee";
+        let v = split(s, &Boundary::default_delimiters());
+        assert_eq!(v, vec!["aaa", "bbb", "ccc", "ddd", "ddd", "eee"]);
+    }
+
+    #[test]
+    fn lower_upper() {
+        let s = "lowerUpperUpper";
+        let v = split(s, &[Boundary::LOWER_UPPER]);
+        assert_eq!(v, vec!["lower", "Upper", "Upper"]);
+    }
+
+    #[test]
+    fn acronym() {
+        let s = "XMLRequest";
+        let v = split(s, &[Boundary::ACRONYM]);
+        assert_eq!(v, vec!["XML", "Request"]);
+    }
+}
diff --git a/src/lib.rs b/src/lib.rs
index 37edc4f..4d8b349 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -186,11 +186,13 @@
 //! This will add two additional cases: Random and PseudoRandom.  You can read about their
 //! construction in the [Case enum](enum.Case.html).
 
+mod boundary;
 mod case;
 mod converter;
 mod pattern;
 mod segmentation;
 
+pub use boundary::Boundary as NewBoundary;
 pub use case::Case;
 pub use converter::Converter;
 pub use pattern::Pattern;
@@ -270,7 +272,8 @@ pub trait Casing<T: AsRef<str>> {
 }
 
 impl<T: AsRef<str>> Casing<T> for T
-where T: ToString
+where
+    T: ToString,
 {
     fn to_case(&self, case: Case) -> String {
         StateConverter::new(self).to_case(case)