Replace sqlparser-rs with DuckDB native parser via C++ FFI

- Add C++ FFI layer using DuckDB's Parser class directly
- Remove sqlparser-rs dependency from Cargo.toml
- Add detection and rejection of non-decomposable aggregates (COUNT DISTINCT, MEDIAN, PERCENTILE) with clear error messages
- Expand aggregate function support to all DuckDB aggregates (STDDEV, MEDIAN, STRING_AGG, etc.)
- Add tests for DuckDB-specific functions (date functions, COALESCE, array functions, cast syntax)
- Update README to remove "experimental" label
- Document non-decomposable aggregate limitation

Author: nicosuave

.claude/settings.local.json

@@ -1,7 +1,11 @@
 {
   "permissions": {
     "allow": [
-      "WebFetch(domain:arxiv.org)"
+      "WebFetch(domain:arxiv.org)",
+      "WebSearch",
+      "WebFetch(domain:docs.rs)",
+      "WebFetch(domain:github.com)",
+      "WebFetch(domain:pganalyze.com)"
     ],
     "deny": [],
     "ask": []

CMakeLists.txt

@@ -11,6 +11,7 @@ set(LOADABLE_EXTENSION_NAME ${TARGET_NAME}_loadable_extension)
 
 project(${TARGET_NAME})
 include_directories(src/include)
+include_directories(include)
 
 # Detect Rust target based on platform
 execute_process(
@@ -82,7 +83,10 @@ corrosion_import_crate(MANIFEST_PATH "${CMAKE_CURRENT_SOURCE_DIR}/yardstick-rs/C
 # Include Rust library headers
 include_directories(${CMAKE_CURRENT_SOURCE_DIR}/yardstick-rs/include)
 
-set(EXTENSION_SOURCES src/yardstick_extension.cpp)
+set(EXTENSION_SOURCES
+    src/yardstick_extension.cpp
+    src/yardstick_parser_ffi.cpp
+)
 
 build_static_extension(${TARGET_NAME} ${EXTENSION_SOURCES})
 build_loadable_extension(${TARGET_NAME} " " ${EXTENSION_SOURCES})

LIMITATIONS.md

@@ -15,13 +15,30 @@ Implementation of Julian Hyde's "Measures in SQL" paper (arXiv:2406.00251).
 - `AT (VISIBLE)` - uses visible WHERE clause filters
 - Multiple measures in same view
 - Arithmetic with AGGREGATE results (ratios, percentages, differences)
-- SUM, COUNT, MIN, MAX, AVG aggregations
+- All DuckDB aggregate functions (SUM, COUNT, AVG, MIN, MAX, STDDEV, MEDIAN, etc.)
 - Derived measures: `revenue - cost AS MEASURE profit` expands to `SUM(revenue) - SUM(cost)`
 - Multi-fact JOINs: measures from different views can be queried together in a single JOIN
 
 ## Known Limitations
 
-### 1. No Window Function Measures
+### 1. Non-Decomposable Aggregates
+
+```sql
+-- NOT SUPPORTED with AGGREGATE(): COUNT(DISTINCT), MEDIAN, PERCENTILE, MODE
+CREATE VIEW v AS
+SELECT region, COUNT(DISTINCT customer_id) AS MEASURE unique_customers
+FROM orders;
+
+-- Direct query works fine:
+SELECT region, unique_customers FROM v;
+
+-- But AGGREGATE() fails (cannot re-aggregate):
+SEMANTIC SELECT region, AGGREGATE(unique_customers) FROM v;  -- ERROR
+```
+
+Non-decomposable aggregates like COUNT(DISTINCT), MEDIAN, and PERCENTILE cannot be re-aggregated. Query these views directly without AGGREGATE().
+
+### 2. No Window Function Measures
 
 ```sql
 -- NOT SUPPORTED: Window functions in measure definitions

README.md

@@ -1,6 +1,6 @@
 # Yardstick
 
-An experimental DuckDB extension implementing Julian Hyde's "Measures in SQL" paper ([arXiv:2406.00251](https://arxiv.org/abs/2406.00251)).
+A DuckDB extension implementing Julian Hyde's "Measures in SQL" paper ([arXiv:2406.00251](https://arxiv.org/abs/2406.00251)).
 
 ## What is this?
 
@@ -107,7 +107,7 @@ SELECT
 FROM table;
 ```
 
-Yardstick automatically handles the grouping. Supported aggregations: `SUM`, `COUNT`, `AVG`, `MIN`, `MAX`
+Yardstick automatically handles the grouping. All DuckDB aggregate functions are supported.
 
 ### Querying Measures
 
@@ -152,6 +152,7 @@ The extension will be at `build/release/extension/yardstick/yardstick.duckdb_ext
 See [LIMITATIONS.md](LIMITATIONS.md) for known issues and workarounds.
 
 Key limitations:
+- Non-decomposable aggregates (COUNT DISTINCT, MEDIAN, PERCENTILE) cannot use AGGREGATE()
 - Window function measures not supported
 
 ## References