Method scan_collection

nodedb/src/data/executor/scan_normalize.rs:29–49

Universal scan: reads from the correct engine for `collection` and returns `(doc_id, msgpack_bytes)` pairs in standard msgpack map format. Routing order: 1. KV engine (if collection has KV entries) 2. Columnar storage (timeseries memtable or plain/spatial engine) 3. Sparse/document engine (default) All results are normalized to standard msgpack maps so callers (aggregate, join, sort, filter) nev

(
        &self,
        tid: u64,
        collection: &str,
        limit: usize,
    )

    /// All results are normalized to standard msgpack maps so callers
    /// (aggregate, join, sort, filter) never need engine-specific code.
    pub fn scan_collection(
        &self,
        tid: u64,
        collection: &str,
        limit: usize,
    ) -> crate::Result<Vec<(String, Vec<u8>)>> {
        // 1. KV engine
        let kv_docs = self.scan_kv(tid, collection, limit);
        if !kv_docs.is_empty() {
            return Ok(kv_docs);
        }

        // 2. Columnar memtable
        let col_docs = self.scan_columnar(tid, collection, limit);
        if !col_docs.is_empty() {
            return Ok(col_docs);
        }

        // 3. Sparse/document engine (schemaless + strict)
        self.scan_sparse(tid, collection, limit)
    }

    /// Scan KV engine entries → standard msgpack.
    /// Injects the `key` field directly into the msgpack map — no JSON roundtrip.