vibesql_executor/select/columnar/

mod.rs

//! Columnar execution for high-performance aggregation queries
//!
//! This module implements column-oriented query execution that avoids
//! materializing full Row objects during table scans, providing 8-10x speedup
//! for aggregation-heavy workloads.
//!
//! ## Architecture
//!
//! Instead of:
//! ```text
//! TableScan → Row{Vec<SqlValue>} → Filter(Row) → Aggregate(Row) → Vec<Row>
//! ```
//!
//! We use:
//! ```text
//! TableScan → ColumnRefs → Filter(native types) → Aggregate → Row
//! ```
//!
//! ## Benefits
//!
//! - **Zero-copy**: Work with `&SqlValue` references instead of cloning
//! - **Cache-friendly**: Access contiguous column data instead of scattered row data
//! - **Type-specialized**: Skip SqlValue enum matching overhead for filters/aggregates
//! - **Minimal allocations**: Only allocate result rows, not intermediate data
//!
//! ## Usage
//!
//! This path is automatically selected for simple aggregate queries that:
//! - Have a single table scan (no JOINs)
//! - Use simple WHERE predicates
//! - Compute aggregates (SUM, COUNT, AVG, MIN, MAX)
//! - Don't use window functions or complex subqueries

mod aggregate;
pub mod batch;
mod executor;
pub mod filter;
mod scan;
mod string_ops;

// Auto-vectorized SIMD operations - replaces the `wide` crate dependency
// See simd_ops.rs for documentation on why these patterns are structured this way
pub mod simd_ops;

mod simd_aggregate;
pub mod simd_filter;
mod simd_join;

pub use aggregate::{
    columnar_group_by, columnar_group_by_batch, compute_aggregates_from_batch,
    compute_multiple_aggregates, evaluate_expression_to_column,
    evaluate_expression_with_cached_column, extract_aggregates, AggregateOp, AggregateSource,
    AggregateSpec,
};
pub use batch::{ColumnArray, ColumnarBatch};
pub use executor::execute_columnar_batch;
pub use filter::{
    apply_columnar_filter, apply_columnar_filter_simd_streaming, create_filter_bitmap,
    create_filter_bitmap_tree, evaluate_predicate_tree, extract_column_predicates,
    extract_predicate_tree, ColumnPredicate, PredicateTree,
};
use log;
pub use scan::ColumnarScan;
pub use simd_aggregate::{can_use_simd_for_column, simd_aggregate_f64, simd_aggregate_i64};
pub use simd_filter::{
    simd_create_filter_mask, simd_create_filter_mask_auto, simd_create_filter_mask_packed,
    simd_filter_batch, simd_filter_to_indices,
};
#[cfg(feature = "parallel")]
pub use simd_filter::{simd_create_filter_mask_parallel, simd_filter_batch_parallel};
pub use simd_join::columnar_hash_join_inner;
pub use simd_ops::PackedMask;
use vibesql_storage::Row;
use vibesql_types::SqlValue;

use crate::{errors::ExecutorError, schema::CombinedSchema};

/// Execute a columnar aggregate query with filtering
///
/// This is a simplified entry point for columnar execution that demonstrates
/// the full pipeline: scan → filter → aggregate.
///
/// # Arguments
///
/// * `rows` - Input rows to process
/// * `predicates` - Column predicates for filtering (optional)
/// * `aggregates` - List of (column_index, aggregate_op) pairs to compute
///
/// # Returns
///
/// A single Row containing the computed aggregate values
///
/// # Example
///
/// ```text
/// // Compute SUM(col0), AVG(col1) WHERE col2 < 100
/// let predicates = vec![
///     ColumnPredicate::LessThan {
///         column_idx: 2,
///         value: SqlValue::Integer(100),
///     },
/// ];
/// let aggregates = vec![
///     (0, AggregateOp::Sum),
///     (1, AggregateOp::Avg),
/// ];
///
/// let result = execute_columnar_aggregate(&rows, &predicates, &aggregates)?;
/// ```
///
/// Note: This function provides SIMD-accelerated filtering and aggregation through
/// LLVM auto-vectorization of batch-native operations.
pub fn execute_columnar_aggregate(
    rows: &[Row],
    predicates: &[ColumnPredicate],
    aggregates: &[aggregate::AggregateSpec],
    schema: Option<&CombinedSchema>,
) -> Result<Vec<Row>, ExecutorError> {
    // Early return for empty input
    // SQL standard: COUNT returns 0 for empty input, other aggregates return NULL
    if rows.is_empty() {
        let values: Vec<SqlValue> = aggregates
            .iter()
            .map(|spec| match spec.op {
                aggregate::AggregateOp::Count => SqlValue::Integer(0),
                _ => SqlValue::Null,
            })
            .collect();
        return Ok(vec![Row::new(values)]);
    }

    // Phase 1: Convert to columnar batch for SIMD acceleration
    #[cfg(feature = "profile-q6")]
    let batch_start = std::time::Instant::now();

    let batch = ColumnarBatch::from_rows(rows)?;

    #[cfg(feature = "profile-q6")]
    {
        let batch_time = batch_start.elapsed();
        eprintln!("[PROFILE-Q6]   Phase 1 - Convert to batch: {:?}", batch_time);
    }

    // Phase 2: Apply SIMD-accelerated filtering
    #[cfg(feature = "profile-q6")]
    let filter_start = std::time::Instant::now();

    let filtered_batch =
        if predicates.is_empty() { batch.clone() } else { simd_filter_batch(&batch, predicates)? };

    #[cfg(feature = "profile-q6")]
    {
        let filter_time = filter_start.elapsed();
        eprintln!(
            "[PROFILE-Q6]   Phase 2 - SIMD filter: {:?} ({}/{} rows passed)",
            filter_time,
            filtered_batch.row_count(),
            rows.len()
        );
    }

    // Phase 3: Compute aggregates directly on batch (no row conversion!)
    #[cfg(feature = "profile-q6")]
    let agg_start = std::time::Instant::now();

    // Use batch-native aggregation to avoid to_rows() conversion overhead
    let results = compute_aggregates_from_batch(&filtered_batch, aggregates, schema)?;

    #[cfg(feature = "profile-q6")]
    {
        let agg_time = agg_start.elapsed();
        eprintln!(
            "[PROFILE-Q6]   Phase 3 - Batch-native aggregate: {:?} ({} aggregates)",
            agg_time,
            aggregates.len()
        );
    }

    // Return as single row
    Ok(vec![Row::new(results)])
}

/// Fast single-pass aggregate on rows - avoids batch conversion overhead
///
/// This function performs filtering and aggregation in a single pass over the input rows,
/// without converting to columnar format. It's 3-5x faster than `execute_columnar_aggregate`
/// for queries that come from row-based storage.
///
/// # Use Cases
///
/// Best suited for:
/// - Simple aggregate queries without GROUP BY
/// - When data arrives as Vec<Row> (not native columnar)
/// - TPC-H style queries: SUM(price * discount) WHERE ...
///
/// # Arguments
///
/// * `rows` - Input rows to process
/// * `predicates` - Column predicates for filtering
/// * `aggregates` - Aggregate specifications
///
/// # Returns
///
/// A single Row containing the computed aggregate values
pub fn fast_aggregate_on_rows(
    rows: &[Row],
    predicates: &[ColumnPredicate],
    aggregates: &[aggregate::AggregateSpec],
) -> Result<Vec<Row>, ExecutorError> {
    use aggregate::{AggregateOp, AggregateSource};

    // Early return for empty input
    if rows.is_empty() {
        let values: Vec<SqlValue> = aggregates
            .iter()
            .map(|spec| match spec.op {
                AggregateOp::Count => SqlValue::Integer(0),
                _ => SqlValue::Null,
            })
            .collect();
        return Ok(vec![Row::new(values)]);
    }

    // Initialize accumulators for each aggregate
    struct Accumulator {
        sum_f64: f64,
        sum_i64: i64,
        count: i64,
        min_f64: Option<f64>,
        max_f64: Option<f64>,
        min_i64: Option<i64>,
        max_i64: Option<i64>,
        is_integer: bool,
    }

    let mut accumulators: Vec<Accumulator> = aggregates
        .iter()
        .map(|_| Accumulator {
            sum_f64: 0.0,
            sum_i64: 0,
            count: 0,
            min_f64: None,
            max_f64: None,
            min_i64: None,
            max_i64: None,
            is_integer: true,
        })
        .collect();

    // Single pass: filter and accumulate
    for row in rows {
        // Check all predicates
        let passes_filter = predicates.iter().all(|pred| evaluate_predicate(row, pred));

        if !passes_filter {
            continue;
        }

        // Accumulate values for each aggregate
        for (i, spec) in aggregates.iter().enumerate() {
            let acc = &mut accumulators[i];

            match &spec.source {
                AggregateSource::CountStar => {
                    acc.count += 1;
                }
                AggregateSource::Column(col_idx) => {
                    if let Some(value) = row.get(*col_idx) {
                        if !matches!(value, SqlValue::Null) {
                            acc.count += 1;
                            match value {
                                SqlValue::Integer(v) => {
                                    acc.sum_i64 += v;
                                    acc.sum_f64 += *v as f64;
                                    acc.min_i64 = Some(acc.min_i64.map_or(*v, |m| m.min(*v)));
                                    acc.max_i64 = Some(acc.max_i64.map_or(*v, |m| m.max(*v)));
                                    acc.min_f64 =
                                        Some(acc.min_f64.map_or(*v as f64, |m| m.min(*v as f64)));
                                    acc.max_f64 =
                                        Some(acc.max_f64.map_or(*v as f64, |m| m.max(*v as f64)));
                                }
                                SqlValue::Double(v) => {
                                    acc.is_integer = false;
                                    acc.sum_f64 += v;
                                    acc.min_f64 = Some(acc.min_f64.map_or(*v, |m| m.min(*v)));
                                    acc.max_f64 = Some(acc.max_f64.map_or(*v, |m| m.max(*v)));
                                }
                                SqlValue::Float(v) => {
                                    acc.is_integer = false;
                                    acc.sum_f64 += *v as f64;
                                    acc.min_f64 =
                                        Some(acc.min_f64.map_or(*v as f64, |m| m.min(*v as f64)));
                                    acc.max_f64 =
                                        Some(acc.max_f64.map_or(*v as f64, |m| m.max(*v as f64)));
                                }
                                SqlValue::Bigint(v) => {
                                    acc.sum_i64 += v;
                                    acc.sum_f64 += *v as f64;
                                    acc.min_i64 = Some(acc.min_i64.map_or(*v, |m| m.min(*v)));
                                    acc.max_i64 = Some(acc.max_i64.map_or(*v, |m| m.max(*v)));
                                    acc.min_f64 =
                                        Some(acc.min_f64.map_or(*v as f64, |m| m.min(*v as f64)));
                                    acc.max_f64 =
                                        Some(acc.max_f64.map_or(*v as f64, |m| m.max(*v as f64)));
                                }
                                SqlValue::Numeric(v) => {
                                    acc.is_integer = false;
                                    acc.sum_f64 += v;
                                    acc.min_f64 = Some(acc.min_f64.map_or(*v, |m| m.min(*v)));
                                    acc.max_f64 = Some(acc.max_f64.map_or(*v, |m| m.max(*v)));
                                }
                                _ => {}
                            }
                        }
                    }
                }
                AggregateSource::Expression(expr) => {
                    // For expression aggregates like SUM(a * b), evaluate the expression
                    // This is a simplified evaluator for common binary operations
                    if let Some(value) = eval_simple_expression(row, expr) {
                        acc.count += 1;
                        acc.is_integer = false;
                        acc.sum_f64 += value;
                        acc.min_f64 = Some(acc.min_f64.map_or(value, |m| m.min(value)));
                        acc.max_f64 = Some(acc.max_f64.map_or(value, |m| m.max(value)));
                    }
                }
            }
        }
    }

    // Build result row from accumulators
    let values: Vec<SqlValue> = aggregates
        .iter()
        .zip(accumulators.iter())
        .map(|(spec, acc)| match spec.op {
            AggregateOp::Count => SqlValue::Integer(acc.count),
            AggregateOp::Sum => {
                // SQLite's SUM() preserves integer type for integer inputs
                if acc.count == 0 {
                    SqlValue::Null
                } else if acc.is_integer {
                    SqlValue::Integer(acc.sum_i64)
                } else {
                    SqlValue::Double(acc.sum_f64)
                }
            }
            AggregateOp::Avg => {
                if acc.count == 0 {
                    SqlValue::Null
                } else {
                    SqlValue::Double(acc.sum_f64 / acc.count as f64)
                }
            }
            AggregateOp::Min => {
                if acc.is_integer {
                    acc.min_i64.map(SqlValue::Integer).unwrap_or(SqlValue::Null)
                } else {
                    acc.min_f64.map(SqlValue::Double).unwrap_or(SqlValue::Null)
                }
            }
            AggregateOp::Max => {
                if acc.is_integer {
                    acc.max_i64.map(SqlValue::Integer).unwrap_or(SqlValue::Null)
                } else {
                    acc.max_f64.map(SqlValue::Double).unwrap_or(SqlValue::Null)
                }
            }
        })
        .collect();

    Ok(vec![Row::new(values)])
}

/// Evaluate a column predicate against a row
///
/// Delegates to the canonical comparator in `filter::evaluation`. Issue
/// #5335: this module previously carried a near-duplicate `compare_values`
/// whose incomparable-types catch-all returned `Ordering::Equal`, turning
/// equality/range predicates on temporal columns into tautologies or
/// contradictions. Routing through the filter module keeps a single source
/// of truth for predicate comparison semantics.
fn evaluate_predicate(row: &Row, predicate: &ColumnPredicate) -> bool {
    // Column-to-column comparison needs two values
    if let ColumnPredicate::ColumnCompare { left_column_idx, op, right_column_idx } = predicate {
        return filter::evaluate_column_compare(
            *op,
            row.get(*left_column_idx),
            row.get(*right_column_idx),
        );
    }

    let column_idx = match predicate {
        ColumnPredicate::LessThan { column_idx, .. }
        | ColumnPredicate::GreaterThan { column_idx, .. }
        | ColumnPredicate::GreaterThanOrEqual { column_idx, .. }
        | ColumnPredicate::LessThanOrEqual { column_idx, .. }
        | ColumnPredicate::Equal { column_idx, .. }
        | ColumnPredicate::NotEqual { column_idx, .. }
        | ColumnPredicate::Between { column_idx, .. }
        | ColumnPredicate::Like { column_idx, .. }
        | ColumnPredicate::InList { column_idx, .. } => *column_idx,
        ColumnPredicate::ColumnCompare { .. } => unreachable!(), // Handled above
    };

    match row.get(column_idx) {
        Some(value) => filter::evaluate_predicate(predicate, value),
        None => false,
    }
}

/// Evaluate a simple expression (for expression aggregates)
#[allow(clippy::only_used_in_recursion)]
fn eval_simple_expression(row: &Row, expr: &vibesql_ast::Expression) -> Option<f64> {
    use vibesql_ast::{BinaryOperator, Expression};

    match expr {
        Expression::BinaryOp { left, op, right } => {
            let left_val = eval_simple_expression(row, left)?;
            let right_val = eval_simple_expression(row, right)?;
            match op {
                BinaryOperator::Multiply => Some(left_val * right_val),
                BinaryOperator::Divide => Some(left_val / right_val),
                BinaryOperator::Plus => Some(left_val + right_val),
                BinaryOperator::Minus => Some(left_val - right_val),
                _ => None,
            }
        }
        Expression::ColumnRef(col_id) => {
            // Cannot resolve column names without a schema - return None to skip
            // the fast path and fall back to the columnar execution path which
            // properly handles expression aggregates with schema resolution.
            log::debug!(
                "fast_aggregate_on_rows: ColumnRef '{}' requires schema resolution, skipping fast path",
                col_id.column_canonical()
            );
            None
        }
        Expression::Literal(val) => match val {
            SqlValue::Integer(v) => Some(*v as f64),
            SqlValue::Double(v) => Some(*v),
            SqlValue::Float(v) => Some(*v as f64),
            SqlValue::Bigint(v) => Some(*v as f64),
            SqlValue::Numeric(v) => Some(*v),
            _ => None,
        },
        _ => None,
    }
}

/// Execute a query using columnar processing (AST-based interface)
///
/// This is the entry point for columnar execution that accepts AST expressions
/// and converts them to the columnar execution pipeline.
///
/// # Arguments
///
/// * `rows` - The rows to process
/// * `filter` - Optional WHERE clause expression
/// * `aggregates` - SELECT list aggregate expressions
/// * `schema` - Schema for resolving column names to indices
///
/// # Returns
///
/// Some(Result) if the query can be optimized using columnar execution,
/// None if the expressions are too complex for columnar optimization.
///
/// Note: This function uses LLVM auto-vectorization for vectorized execution.
pub fn execute_columnar(
    rows: &[Row],
    filter: Option<&vibesql_ast::Expression>,
    aggregates: &[vibesql_ast::Expression],
    schema: &CombinedSchema,
    case_sensitive_like: bool,
) -> Option<Result<Vec<Row>, ExecutorError>> {
    log::debug!("  Executing columnar query with {} rows", rows.len());

    // Extract column predicates from filter expression
    let predicates = if let Some(filter_expr) = filter {
        match extract_column_predicates(filter_expr, schema, case_sensitive_like) {
            Some(preds) => {
                log::debug!("    ✓ Extracted {} column predicates for SIMD filtering", preds.len());
                preds
            }
            None => {
                log::debug!("    ✗ Filter too complex for columnar optimization");
                return None; // Too complex for columnar optimization
            }
        }
    } else {
        log::debug!("    No filter predicates");
        vec![] // No filter
    };

    // Extract aggregates from SELECT list
    let agg_specs = match extract_aggregates(aggregates, schema) {
        Some(specs) => {
            log::debug!("    ✓ Extracted {} aggregate operations", specs.len());
            for (i, spec) in specs.iter().enumerate() {
                log::debug!("      Aggregate {}: {:?}", i + 1, spec.op);
            }
            specs
        }
        None => {
            log::debug!("    ✗ Aggregates too complex for columnar optimization");
            return None; // Too complex for columnar optimization
        }
    };

    // Call the simplified interface, passing schema if any aggregates use expressions
    let needs_schema = agg_specs
        .iter()
        .any(|spec| matches!(spec.source, aggregate::AggregateSource::Expression(_)));
    let schema_ref = if needs_schema { Some(schema) } else { None };

    log::debug!("    Executing SIMD-accelerated columnar aggregation");
    Some(execute_columnar_aggregate(rows, &predicates, &agg_specs, schema_ref))
}

#[cfg(test)]
mod tests {
    use vibesql_types::Date;

    use super::*;

    /// Test the full columnar pipeline: filter + aggregation
    #[test]
    fn test_columnar_pipeline_filtered_sum() {
        // Create test data: TPC-H Q6 style query
        // SELECT SUM(l_extendedprice * l_discount)
        // WHERE l_shipdate >= '1994-01-01'
        //   AND l_shipdate < '1995-01-01'
        //   AND l_discount BETWEEN 0.05 AND 0.07
        //   AND l_quantity < 24

        let rows = vec![
            Row::new(vec![
                SqlValue::Integer(10),   // quantity
                SqlValue::Double(100.0), // extendedprice
                SqlValue::Double(0.06),  // discount
                SqlValue::Date(Date::new(1994, 6, 1).unwrap()),
            ]),
            Row::new(vec![
                SqlValue::Integer(25), // quantity (filtered out: > 24)
                SqlValue::Double(200.0),
                SqlValue::Double(0.06),
                SqlValue::Date(Date::new(1994, 7, 1).unwrap()),
            ]),
            Row::new(vec![
                SqlValue::Integer(15), // quantity
                SqlValue::Double(150.0),
                SqlValue::Double(0.05), // discount
                SqlValue::Date(Date::new(1994, 8, 1).unwrap()),
            ]),
            Row::new(vec![
                SqlValue::Integer(20), // quantity
                SqlValue::Double(180.0),
                SqlValue::Double(0.08), // discount (filtered out: > 0.07)
                SqlValue::Date(Date::new(1994, 9, 1).unwrap()),
            ]),
        ];

        // Predicates: quantity < 24 AND discount BETWEEN 0.05 AND 0.07
        let predicates = vec![
            ColumnPredicate::LessThan { column_idx: 0, value: SqlValue::Integer(24) },
            ColumnPredicate::Between {
                column_idx: 2,
                low: SqlValue::Double(0.05),
                high: SqlValue::Double(0.07),
            },
        ];

        // Aggregates: SUM(extendedprice), COUNT(*)
        let aggregates = vec![
            AggregateSpec { op: AggregateOp::Sum, source: AggregateSource::Column(1) }, /* SUM(extendedprice) */
            AggregateSpec { op: AggregateOp::Count, source: AggregateSource::Column(0) }, /* COUNT(*) */
        ];

        let result = execute_columnar_aggregate(&rows, &predicates, &aggregates, None).unwrap();

        assert_eq!(result.len(), 1);
        let result_row = &result[0];

        // Only rows 0 and 2 pass the filter (quantity < 24 AND discount in range)
        // SUM(extendedprice) = 100.0 + 150.0 = 250.0
        assert!(
            matches!(result_row.get(0), Some(&SqlValue::Double(sum)) if (sum - 250.0).abs() < 0.001)
        );
        // COUNT(*) = 2
        assert_eq!(result_row.get(1), Some(&SqlValue::Integer(2)));
    }

    /// Test columnar execution with no filtering
    #[test]
    fn test_columnar_pipeline_no_filter() {
        let rows = vec![
            Row::new(vec![SqlValue::Integer(10), SqlValue::Double(1.5)]),
            Row::new(vec![SqlValue::Integer(20), SqlValue::Double(2.5)]),
            Row::new(vec![SqlValue::Integer(30), SqlValue::Double(3.5)]),
        ];

        let predicates = vec![];
        let aggregates = vec![
            AggregateSpec { op: AggregateOp::Sum, source: AggregateSource::Column(0) },
            AggregateSpec { op: AggregateOp::Avg, source: AggregateSource::Column(1) },
            AggregateSpec { op: AggregateOp::Max, source: AggregateSource::Column(0) },
        ];

        let result = execute_columnar_aggregate(&rows, &predicates, &aggregates, None).unwrap();

        assert_eq!(result.len(), 1);
        let result_row = &result[0];

        // SUM(col0) = 60 (SQLite's SUM() returns REAL)
        assert_eq!(result_row.get(0), Some(&SqlValue::Integer(60)));
        // AVG(col1) = 2.5
        assert!(
            matches!(result_row.get(1), Some(&SqlValue::Double(avg)) if (avg - 2.5).abs() < 0.001)
        );
        // MAX(col0) = 30
        assert_eq!(result_row.get(2), Some(&SqlValue::Integer(30)));
    }

    /// Test columnar execution with empty result set
    #[test]
    fn test_columnar_pipeline_empty_result() {
        let rows =
            vec![Row::new(vec![SqlValue::Integer(100)]), Row::new(vec![SqlValue::Integer(200)])];

        // Filter that matches nothing
        let predicates =
            vec![ColumnPredicate::LessThan { column_idx: 0, value: SqlValue::Integer(50) }];

        let aggregates = vec![
            AggregateSpec { op: AggregateOp::Sum, source: AggregateSource::Column(0) },
            AggregateSpec { op: AggregateOp::Count, source: AggregateSource::Column(0) },
        ];

        let result = execute_columnar_aggregate(&rows, &predicates, &aggregates, None).unwrap();

        assert_eq!(result.len(), 1);
        let result_row = &result[0];

        // SUM of empty set = NULL
        assert_eq!(result_row.get(0), Some(&SqlValue::Null));
        // COUNT of empty set = 0
        assert_eq!(result_row.get(1), Some(&SqlValue::Integer(0)));
    }

    // AST Integration Tests

    use vibesql_ast::{BinaryOperator, Expression};
    use vibesql_catalog::{ColumnSchema, TableSchema};
    use vibesql_types::DataType;

    use crate::schema::CombinedSchema;

    fn make_test_schema() -> CombinedSchema {
        let schema = TableSchema::new(
            "test".to_string(),
            vec![
                ColumnSchema::new("quantity".to_string(), DataType::Integer, false),
                ColumnSchema::new("price".to_string(), DataType::DoublePrecision, false),
            ],
        );
        CombinedSchema::from_table("test".to_string(), schema)
    }

    fn make_test_rows_for_ast() -> Vec<Row> {
        vec![
            Row::new(vec![SqlValue::Integer(10), SqlValue::Double(1.5)]),
            Row::new(vec![SqlValue::Integer(20), SqlValue::Double(2.5)]),
            Row::new(vec![SqlValue::Integer(30), SqlValue::Double(3.5)]),
            Row::new(vec![SqlValue::Integer(40), SqlValue::Double(4.5)]),
        ]
    }

    #[test]
    fn test_execute_columnar_simple_aggregate() {
        let rows = make_test_rows_for_ast();
        let schema = make_test_schema();

        // SELECT SUM(price) FROM test
        let aggregates = vec![Expression::AggregateFunction {
            name: vibesql_ast::FunctionIdentifier::new("SUM"),
            distinct: false,
            args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                "price", false,
            ))],
            order_by: None,
            filter: None,
        }];

        let result = execute_columnar(&rows, None, &aggregates, &schema, false);
        assert!(result.is_some());

        let rows_result = result.unwrap();
        assert!(rows_result.is_ok());

        let result_rows = rows_result.unwrap();
        assert_eq!(result_rows.len(), 1);
        assert_eq!(result_rows[0].len(), 1);

        // Sum should be 1.5 + 2.5 + 3.5 + 4.5 = 12.0
        if let Some(SqlValue::Double(sum)) = result_rows[0].get(0) {
            assert!((sum - 12.0).abs() < 0.001);
        } else {
            panic!("Expected Numeric value for SUM");
        }
    }

    #[test]
    fn test_execute_columnar_with_filter() {
        let rows = make_test_rows_for_ast();
        let schema = make_test_schema();

        // SELECT SUM(price) FROM test WHERE quantity < 25
        let filter = Expression::BinaryOp {
            left: Box::new(Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                "quantity", false,
            ))),
            op: BinaryOperator::LessThan,
            right: Box::new(Expression::Literal(SqlValue::Integer(25))),
        };

        let aggregates = vec![Expression::AggregateFunction {
            name: vibesql_ast::FunctionIdentifier::new("SUM"),
            distinct: false,
            args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                "price", false,
            ))],
            order_by: None,
            filter: None,
        }];

        let result = execute_columnar(&rows, Some(&filter), &aggregates, &schema, false);
        assert!(result.is_some());

        let rows_result = result.unwrap();
        assert!(rows_result.is_ok());

        let result_rows = rows_result.unwrap();
        assert_eq!(result_rows.len(), 1);
        assert_eq!(result_rows[0].len(), 1);

        // Sum of rows where quantity < 25: 1.5 + 2.5 = 4.0
        if let Some(SqlValue::Double(sum)) = result_rows[0].get(0) {
            assert!((sum - 4.0).abs() < 0.001);
        } else {
            panic!("Expected Numeric value for SUM");
        }
    }

    #[test]
    fn test_execute_columnar_multiple_aggregates() {
        let rows = make_test_rows_for_ast();
        let schema = make_test_schema();

        // SELECT SUM(price), COUNT(*), AVG(quantity) FROM test
        let aggregates = vec![
            Expression::AggregateFunction {
                name: vibesql_ast::FunctionIdentifier::new("SUM"),
                distinct: false,
                args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                    "price", false,
                ))],
                order_by: None,
                filter: None,
            },
            Expression::AggregateFunction {
                name: vibesql_ast::FunctionIdentifier::new("COUNT"),
                distinct: false,
                args: vec![Expression::Wildcard],
                order_by: None,
                filter: None,
            },
            Expression::AggregateFunction {
                name: vibesql_ast::FunctionIdentifier::new("AVG"),
                distinct: false,
                args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                    "quantity", false,
                ))],
                order_by: None,
                filter: None,
            },
        ];

        let result = execute_columnar(&rows, None, &aggregates, &schema, false);
        assert!(result.is_some());

        let rows_result = result.unwrap();
        assert!(rows_result.is_ok());

        let result_rows = rows_result.unwrap();
        assert_eq!(result_rows.len(), 1);
        assert_eq!(result_rows[0].len(), 3);

        // Check SUM(price) = 12.0
        if let Some(SqlValue::Double(sum)) = result_rows[0].get(0) {
            assert!((sum - 12.0).abs() < 0.001);
        } else {
            panic!("Expected Numeric value for SUM");
        }

        // Check COUNT(*) = 4
        assert_eq!(result_rows[0].get(1), Some(&SqlValue::Integer(4)));

        // Check AVG(quantity) = (10+20+30+40)/4 = 25.0
        if let Some(SqlValue::Double(avg)) = result_rows[0].get(2) {
            assert!((avg - 25.0).abs() < 0.001);
        } else {
            panic!("Expected Double value for AVG");
        }
    }

    #[test]
    fn test_execute_columnar_unsupported_distinct() {
        let rows = make_test_rows_for_ast();
        let schema = make_test_schema();

        // SELECT SUM(DISTINCT price) FROM test - should return None
        let aggregates = vec![Expression::AggregateFunction {
            name: vibesql_ast::FunctionIdentifier::new("SUM"),
            distinct: true,
            args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                "price", false,
            ))],
            order_by: None,
            filter: None,
        }];

        let result = execute_columnar(&rows, None, &aggregates, &schema, false);
        assert!(result.is_none());
    }

    #[test]
    fn test_execute_columnar_unsupported_complex_filter() {
        let rows = make_test_rows_for_ast();
        let schema = make_test_schema();

        // SELECT SUM(price) FROM test WHERE quantity IN (SELECT ...) - unsupported
        let filter = Expression::ScalarSubquery(Box::new(vibesql_ast::SelectStmt {
            with_clause: None,
            distinct: false,
            select_list: vec![],
            into_table: None,
            into_variables: None,
            from: None,
            where_clause: None,
            group_by: None,
            having: None,
            window_definitions: None,
            order_by: None,
            limit: None,
            offset: None,
            set_operation: None,
            values: None,
        }));

        let aggregates = vec![Expression::AggregateFunction {
            name: vibesql_ast::FunctionIdentifier::new("SUM"),
            distinct: false,
            args: vec![Expression::ColumnRef(vibesql_ast::ColumnIdentifier::simple(
                "price", false,
            ))],
            order_by: None,
            filter: None,
        }];

        let result = execute_columnar(&rows, Some(&filter), &aggregates, &schema, false);
        assert!(result.is_none());
    }
}