benchmark: use a rocksdb version of the dataset (#23)

Use a rocksdb version of the dataset created from the preprocessed
HashMap. This loads almost instantly, has peak memory usage starting
around 75 MB, and takes about 220 MB on disk.

Cargo.toml

@@ -20,3 +20,4 @@ rayon = "1.10.0"
 serde_json = "1.0.133"
 sled = "0.34.7"
 bincode = "1.3.3"
+rocksdb = "0.22.0"

src/main.rs

@@ -12,6 +12,7 @@ use std::fs::metadata;
 use streaming_iterator::StreamingIterator;
 use tree_sitter::{Parser, Query, QueryCursor};
 use walkdir::{DirEntry, WalkDir};
+use rocksdb;
 
 fn read_contents_file(file_path: &str) -> HashMap<String, Vec<(String, String)>> {
     let mut package_map = HashMap::new();
@@ -32,13 +33,13 @@ fn read_contents_file(file_path: &str) -> HashMap<String, Vec<(String, String)>>
 }
 
 fn main() {
-    let db = sled::open("sled_db").expect("Failed to open sled database");
+    let db = rocksdb::DB::open_default("rocksdb").expect("Failed to open RocksDB database");
 
     // for (key, value) in read_contents_file("Contents-amd64-noble") {
     //     let value_bytes: Vec<u8> = bincode::serialize(&value).expect("Failed to serialize value");
-    //     db.insert(key, value_bytes).expect("Failed to insert into sled database");
+    //     db.put(key, value_bytes).expect("Failed to insert into RocksDB database");
     // }
-    // db.flush().expect("Failed to flush sled database");
+    // db.flush().expect("Failed to flush RocksDB database");
     // return;
 
     // let mut file_counts: Vec<_> = db.iter().map(|(file, packages)| (file, packages.len())).collect();