fix: accept STANDALONE_TABLE_FUNCTION_ENTRY in bindInQueryCall (VECTOR extension fix)

[chiangchenghsin-hash]

Summary

Fixes CREATE_VECTOR_INDEX returning Binder exception: CREATE_VECTOR_INDEX is not a table or algorithm function when called from the Node.js API.

Bug

The VECTOR extension registers CREATE_VECTOR_INDEX as a STANDALONE_TABLE_FUNCTION_ENTRY (type 26). However, bindInQueryCall() in bind_in_query_call.cpp only matches TABLE_FUNCTION_ENTRY (type 23) — missing the standalone variant entirely.

This causes the error on ALL database types (:memory: and file-based) when calling CREATE_VECTOR_INDEX from Node.js. The CLI shell is unaffected because it goes through parser/visitor/standalone_call_rewriter.cpp which handles the rewrite path differently.

Fix

Add case CatalogEntryType::STANDALONE_TABLE_FUNCTION_ENTRY: to the switch statement in bindInQueryCall(). One line change.

Verification

• Windows 11 x64, MSVC 19.50, LadybugDB 0.18.3
• Node.js API: conn.querySync("CALL CREATE_VECTOR_INDEX(...)") → ✅ HNSW index created and queried successfully
• FTS (CREATE_FTS_INDEX) continues to work correctly (unaffected by this change)
• ALGO functions (page_rank, scc, etc.) continue to work correctly

Related

• SEGFAULT (0xC0000005) on Windows x64 when executing LOAD VECTOR LadybugDB/ladybug#499 (Windows VECTOR SEGFAULT — this fix addresses the remaining VECTOR API issue)