todo

rust-whisper-baked/src/main.rs

@@ -84,51 +84,64 @@ fn channel(flags: rust_whisper_lib::Flags) {
 }
 
 struct Destutterer {
-   prev: Option<String>,
+   prevs: Vec<Word>,
 }
 
 fn new_destutterer() -> Destutterer {
-   Destutterer{prev: None}
+   Destutterer{prevs: vec![]}
 }
 
 impl Destutterer {
    fn step(&mut self, next: String) -> String {
-      let next = next.trim().to_string();
       if next.len() == 0 {
          return next;
       }
-      match &self.prev {
+
-         None => {
+      let nexts = Word::from_string(next.clone());
-            self.prev = Some(next.clone());
+
-            next
+      let mut n = self.prevs.len().clamp(0, nexts.len());
-         },
-         Some(prev) => {
-            let without_trailing_punctuation = {
-               let mut next = next.clone();
-               while next.ends_with("?") || next.ends_with(".") {
-                  next = next[..next.len()-1].to_string();
-               }
-               next
-            };
-            let trailing_punctuation = next[without_trailing_punctuation.len() ..].to_string();
-            let next = without_trailing_punctuation.clone();
-            let next = {
-               let mut n = prev.len().clamp(0, next.len());
       while n > 0 {
-                  if prev[prev.len() - n..] == next[..n] {
+         let prev_s = Word::to_comparable_string(self.prevs[self.prevs.len() - n..].to_vec());
+         let next_s = Word::to_comparable_string(nexts[..n].to_vec());
+         if prev_s == next_s {
             break;
          }
          n -= 1;
       }
-               next[n..].to_string()
+      self.prevs = nexts.clone();
-            };
+      Word::to_string(nexts[n..].to_vec())
-            if next.len() == 0 {
-               return "".to_string();
    }
-            self.prev = Some(without_trailing_punctuation);
-            next + &trailing_punctuation
-         },
 }
+
+#[derive(Clone)]
+struct Word {
+   raw: String,
+}
+
+impl Word {
+   fn from_string(s: String) -> Vec<Word> {
+      let mut result = vec![];
+      for word in s.split(" ") {
+         let word = word.trim();
+         if word.len() > 0 {
+            result.push(Word{raw: word.to_string()});
+         }
+      }
+      result
+   }
+
+   fn to_comparable_string(v: Vec<Word>) -> String {
+      v.iter()
+         .map(|x| x.raw.chars().filter(|c| c.is_ascii_alphanumeric()).collect())
+         .collect::<Vec<String>>()
+         .join(" ")
+   }
+
+   fn to_string(v: Vec<Word>) -> String {
+      v.iter()
+         .map(|x| x.raw.clone())
+         .collect::<Vec<String>>()
+         .join(" ")
    }
 }
 
@@ -137,11 +150,19 @@ mod tests {
    use super::*;
 
    #[test]
-   fn test_destutterer() {
+   fn test_destutterer_punctuation() {
+      let mut w = new_destutterer();
+      assert_eq!("a, b. c? d!".to_string(), w.step("a, b. c? d!".to_string()));
+      assert_eq!("e! f g".to_string(), w.step("d, e! f g".to_string()));
+      assert_eq!("hij".to_string(), w.step("f g hij".to_string()));
+   }
+
+   #[test]
+   fn test_destutterer_letters() {
       let mut w = new_destutterer();
       assert_eq!("a b c d e".to_string(), w.step("a b c d e".to_string()));
       assert_eq!("f g".to_string(), w.step(" c d e f g".to_string()));
       assert_eq!("h i j".to_string(), w.step("f g h i j ".to_string()));
-      assert_eq!("fghij".to_string(), w.step("fghij".to_string()));
+      assert_eq!("a g h i j".to_string(), w.step("a g h i j".to_string()));
    }
 }

todo.yaml

@@ -1,5 +1,7 @@
 todo:
-- need to overlap without ANY puctuation, which i can do by breaking into words
+- overlap without stop words
 - rust-whisper warn when transcription time ~ input time
 scheduled: []
-done: []
+done:
+- todo: need to overlap without ANY puctuation, which i can do by breaking into words
+  ts: Tue Jan  2 18:23:00 MST 2024