use bits instead of strings (#314)

montanalow · web-flow · commit df09b1fd1ca2 · 2022-09-12T15:50:41.000-07:00
diff --git a/pgml-extension/pgml_rust/src/orm/dataset.rs b/pgml-extension/pgml_rust/src/orm/dataset.rs
@@ -30,7 +30,10 @@ impl Dataset {
 
     pub fn distinct_labels(&self) -> u32 {
         let mut v = HashSet::new();
-        self.y.iter().for_each(|i| if !i.is_nan() { v.insert(i.to_string()); });
+        // Treat the f32 values as u32 for std::cmp::Eq. We don't 
+        // care about the nuance of nan equality here, they should
+        // already be filtered out upstream.
+        self.y.iter().for_each(|i| if !i.is_nan() { v.insert(i.to_bits()); });
         v.len().try_into().unwrap()
     }
 }

Original file line number	Diff line number	Diff line change
`@@ -30,7 +30,10 @@ impl Dataset {`
`30`	`30`
`31`	`31`	`pub fn distinct_labels(&self) -> u32 {`
`32`	`32`	`let mut v = HashSet::new();`
`33`		`- self.y.iter().for_each(\|i\| if !i.is_nan() { v.insert(i.to_string()); });`
	`33`	`+ // Treat the f32 values as u32 for std::cmp::Eq. We don't`
	`34`	`+ // care about the nuance of nan equality here, they should`
	`35`	`+ // already be filtered out upstream.`
	`36`	`+ self.y.iter().for_each(\|i\| if !i.is_nan() { v.insert(i.to_bits()); });`
`34`	`37`	`v.len().try_into().unwrap()`
`35`	`38`	`}`
`36`	`39`	`}`