⚡ Bolt: [performance improvement] Optimize D1 SQL string generation

.jules/bolt.md

@@ -2,3 +2,7 @@
 ## 2026-04-08 - [Performance: Defer Allocation during Traversal]
 **Learning:** During DAG traversals, creating owned variants of identifiers (like `file.to_path_buf()`) *before* checking `visited` HashSets results in heap allocations (O(E)) for every edge instead of every visited node (O(V)). By moving the `&PathBuf` allocation strictly *after* all HashSet `contains` checks using the borrowed reference (`&Path`), we drastically reduce memory churn.
 **Action:** Always check `HashSet::contains` with a borrowed reference *before* creating the owned version required by `HashSet::insert`, especially in performance-critical graph traversal paths.
+
+## 2024-05-30 - [Performance: Direct String Writing over Format]
+**Learning:** Constructing complex SQL queries (or other long strings) with dynamic elements using `format!` or joining intermediate arrays causes multiple unnecessary heap allocations. This directly negatively impacts query generation latency, which can be the bottleneck in the pipeline.
+**Action:** For string construction with loops or multiple elements, always use `std::fmt::Write` (the `write!` macro) into a `String::with_capacity(...)` pre-allocated buffer to drastically reduce heap allocations and memory copies.

crates/flow/src/targets/d1.rs

@@ -214,7 +214,13 @@ impl D1ExportContext {
 
         // Generate cache key from SQL + params
         #[cfg(feature = "caching")]
-        let cache_key = format!("{}{:?}", sql, params);
+        let cache_key = {
+            use std::fmt::Write;
+            let mut key = String::with_capacity(sql.len() + params.len() * 32);
+            key.push_str(sql);
+            let _ = write!(key, "{:?}", params);
+            key
+        };
 
         // Check cache first (only for caching feature)
         #[cfg(feature = "caching")]
@@ -300,40 +306,60 @@ impl D1ExportContext {
         key: &KeyValue,
         values: &FieldValues,
     ) -> Result<(String, Vec<serde_json::Value>), RecocoError> {
-        let mut columns = vec![];
-        let mut placeholders = vec![];
-        let mut params = vec![];
-        let mut update_clauses = vec![];
+        use std::fmt::Write;
+        // Bolt optimization: Pre-allocate capacity and use write! to construct SQL
+        // string without intermediate Vec allocations or format! loops
+        let num_keys = self.key_fields_schema.len();
+        let num_values = self.value_fields_schema.len();
+        let mut sql = String::with_capacity(128 + num_keys * 10 + num_values * 30);
+        let mut params = Vec::with_capacity(num_keys + num_values);
+
+        let _ = write!(sql, "INSERT INTO {} (", self.table_name);
+        let mut first = true;
 
         // Extract key parts - KeyValue is a wrapper around Box<[KeyPart]>
         for (idx, _key_field) in self.key_fields_schema.iter().enumerate() {
             if let Some(key_part) = key.0.get(idx) {
-                columns.push(self.key_fields_schema[idx].name.clone());
-                placeholders.push("?".to_string());
+                if !first {
+                    let _ = write!(sql, ", ");
+                }
+                let _ = write!(sql, "{}", self.key_fields_schema[idx].name);
+                first = false;
                 params.push(key_part_to_json(key_part)?);
             }
         }
 
         // Add value fields
         for (idx, value) in values.fields.iter().enumerate() {
             if let Some(value_field) = self.value_fields_schema.get(idx) {
-                columns.push(value_field.name.clone());
-                placeholders.push("?".to_string());
+                if !first {
+                    let _ = write!(sql, ", ");
+                }
+                let _ = write!(sql, "{}", value_field.name);
+                first = false;
                 params.push(value_to_json(value)?);
-                update_clauses.push(format!(
-                    "{} = excluded.{}",
-                    value_field.name, value_field.name
-                ));
             }
         }
 
-        let sql = format!(
-            "INSERT INTO {} ({}) VALUES ({}) ON CONFLICT DO UPDATE SET {}",
-            self.table_name,
-            columns.join(", "),
-            placeholders.join(", "),
-            update_clauses.join(", ")
-        );
+        let _ = write!(sql, ") VALUES (");
+        for i in 0..params.len() {
+            if i > 0 {
+                let _ = write!(sql, ", ");
+            }
+            let _ = write!(sql, "?");
+        }
+
+        let _ = write!(sql, ") ON CONFLICT DO UPDATE SET ");
+        first = true;
+        for (idx, _) in values.fields.iter().enumerate() {
+            if let Some(value_field) = self.value_fields_schema.get(idx) {
+                if !first {
+                    let _ = write!(sql, ", ");
+                }
+                let _ = write!(sql, "{0} = excluded.{0}", value_field.name);
+                first = false;
+            }
+        }
 
         Ok((sql, params))
     }
@@ -342,22 +368,25 @@ impl D1ExportContext {
         &self,
         key: &KeyValue,
     ) -> Result<(String, Vec<serde_json::Value>), RecocoError> {
-        let mut where_clauses = vec![];
-        let mut params = vec![];
+        use std::fmt::Write;
+        let num_keys = self.key_fields_schema.len();
+        let mut sql = String::with_capacity(32 + self.table_name.len() + num_keys * 20);
+        let mut params = Vec::with_capacity(num_keys);
+
+        let _ = write!(sql, "DELETE FROM {} WHERE ", self.table_name);
+        let mut first = true;
 
         for (idx, _key_field) in self.key_fields_schema.iter().enumerate() {
             if let Some(key_part) = key.0.get(idx) {
-                where_clauses.push(format!("{} = ?", self.key_fields_schema[idx].name));
+                if !first {
+                    let _ = write!(sql, " AND ");
+                }
+                let _ = write!(sql, "{} = ?", self.key_fields_schema[idx].name);
+                first = false;
                 params.push(key_part_to_json(key_part)?);
             }
         }
 
-        let sql = format!(
-            "DELETE FROM {} WHERE {}",
-            self.table_name,
-            where_clauses.join(" AND ")
-        );
-
         Ok((sql, params))
     }