// the take
A native XLL add-in that embeds DuckDB directly inside the Excel process, exposing worksheet ranges to SQL through an xlrange() table function and supporting parameter binding, async execution, and external DuckDB/Parquet queries. It's aimed at Excel-bound analysts who want real SQL over spreadsheet data without VBA or a separate database, and positions itself as a more capable alternative to xlDuckDB.
No .NET runtime dependency — it's a plain C XLL, avoiding the VSTO/.NET-for-Office weight most Excel add-ins carry. duckdb.dll loads dynamically at runtime, so upgrading the engine is a file swap, not a recompile. xlrange() has real configuration (sample size, all_varchar, strict header validation, explicit per-column type overrides) instead of naive first-row-wins inference, plus column projection pushdown since 1.7.0. Parameter binding (both auto-incremented ? and positional $n) lets formulas stay parameterized instead of string-concatenating SQL.
xlrange has no row filter pushdown — DuckDB's C API doesn't expose pushed filters to table functions — so every query scans the full input range regardless of the WHERE clause, which caps how far this scales on large sheets. Dynamic PIVOT with bound parameters hits a documented internal error ("Attempted to dereference unique_ptr that is NULL"), a crash-adjacent bug rather than a clean unsupported-feature message, on a fairly common BI pattern. Results have to fit Excel's grid: capped rows/columns, 32,767-char strings, no LIST/STRUCT/MAP/UNION/VARIANT output, and BIGINT/HUGEINT/DECIMAL silently lose precision converting to Excel doubles. Only 13 stars with one apparent maintainer, a MinGW-w64 toolchain that requires hand-editing XLCALL.H before it even compiles, and 'Stable' status backed by a single bundled test workbook rather than CI.
We read the README, the directory tree and a key source file in November 2026. Nobody pays to be listed. How we judge →