web search done right
only consider simhash duplicates if we don't have enough non-duplicates
Mikkel Denkerc4ea75ad2a8a
diff --git a/crates/core/src/collector/top_docs.rs b/crates/core/src/collector/top_docs.rsindex c1cff34a..7e957465 100644--- a/crates/core/src/collector/top_docs.rs+++ b/crates/core/src/collector/top_docs.rs@@ -196,6 +196,7 @@ impl TopSegmentCollector {}}+#[derive(Debug, Clone)]struct ScoredDoc<T: Doc> {doc: T,adjusted_score: f64,@@ -337,7 +338,11 @@ impl<T: Doc> BucketCollector<T> {}}- res.extend(simhash_dups);+ res.extend(+ simhash_dups+ .into_iter()+ .take(self.top_n.saturating_sub(res.len())),+ );res}