@quereus/quereus 0.7.3 → 0.7.4

package/src/planner/building/select-aggregates.ts

@@ -1,318 +0,0 @@
-import type * as AST from '../../parser/ast.js';
-import type { RelationalPlanNode, ScalarPlanNode } from '../nodes/plan-node.js';
-import type { PlanningContext } from '../planning-context.js';
-import { AggregateNode } from '../nodes/aggregate-node.js';
-import { FilterNode } from '../nodes/filter.js';
-import { SortNode, type SortKey } from '../nodes/sort.js';
-import { type Projection } from '../nodes/project-node.js';
-import { RegisteredScope } from '../scopes/registered.js';
-import { ColumnReferenceNode } from '../nodes/reference.js';
-import { buildExpression } from './expression.js';
-import { QuereusError } from '../../common/errors.js';
-import { StatusCode } from '../../common/types.js';
-import { CapabilityDetectors } from '../framework/characteristics.js';
-import { Scope } from '../scopes/scope.js';
-
-/**
- * Processes GROUP BY, aggregates, and HAVING clauses
- */
-export function buildAggregatePhase(
-	input: RelationalPlanNode,
-	stmt: AST.SelectStmt,
-	selectContext: PlanningContext,
-	aggregates: { expression: ScalarPlanNode; alias: string }[],
-	hasAggregates: boolean,
-	projections: Projection[]
-): {
-	output: RelationalPlanNode;
-	aggregateScope?: RegisteredScope;
-	needsFinalProjection: boolean;
-	preAggregateSort: boolean;
-	aggregateNode?: RelationalPlanNode;
-	groupByExpressions?: ScalarPlanNode[];
-} {
-		const hasGroupBy = stmt.groupBy && stmt.groupBy.length > 0;
-
-	// If there is a HAVING clause but the SELECT contains **no aggregate functions**
-	// AND **no GROUP BY**, we can safely treat the HAVING predicate as a regular filter
-	// that runs *before* the aggregation (i.e. between the source and the AggregateNode).
-	// This avoids the "missing column context" problem where the predicate refers to columns
-	// that are not available after the AggregateNode (only grouping columns and
-	// aggregate results are exposed). This behaviour is compatible with SQLite –
-	// GROUP BY with a primary-key guarantees one row per group so the semantics
-	// are unchanged.
-	const shouldPushHavingBelowAggregate = Boolean(stmt.having && !hasAggregates && !hasGroupBy);
-
-	if (!hasAggregates && !hasGroupBy) {
-		return { output: input, needsFinalProjection: false, preAggregateSort: false };
-	}
-
-	// ---------------------------------------------------------------------------
-	// Build HAVING predicate as *pre-aggregate* filter when appropriate
-	// ---------------------------------------------------------------------------
-	let currentInput: RelationalPlanNode = input;
-	if (shouldPushHavingBelowAggregate) {
-		// Build the predicate using the *pre-aggregate* scope because all columns
-		// are still available here.
-		const havingExpr = buildExpression(selectContext, stmt.having as AST.Expression, true);
-		currentInput = new FilterNode(selectContext.scope, currentInput, havingExpr);
-	}
-
-	// After (optional) early HAVING filter we continue with the existing pipeline
-	// ----------------------------------------------------------------------------
-	// Handle pre-aggregate sorting for ORDER BY without GROUP BY
-	const preAggregateSort = Boolean(hasAggregates && !hasGroupBy && stmt.orderBy && stmt.orderBy.length > 0);
-	currentInput = handlePreAggregateSort(currentInput, stmt, selectContext, hasAggregates, !!hasGroupBy);
-
-	// Validate aggregate/non-aggregate mixing
-	validateAggregateProjections(projections, hasAggregates, !!hasGroupBy);
-
-	// Build GROUP BY expressions
-	const groupByExpressions = stmt.groupBy ?
-		stmt.groupBy.map(expr => buildExpression(selectContext, expr, false)) : [];
-
-	// Create AggregateNode
-	const aggregateNode = new AggregateNode(selectContext.scope, currentInput, groupByExpressions, aggregates);
-	currentInput = aggregateNode;
-
-	// Create aggregate output scope
-	const aggregateOutputScope = createAggregateOutputScope(
-		selectContext.scope,
-		currentInput,
-		groupByExpressions,
-		aggregates
-	);
-
-	// Handle HAVING clause *after* aggregation only when we did not already push
-	// it below the AggregateNode.
-	if (stmt.having && !shouldPushHavingBelowAggregate) {
-		currentInput = buildHavingFilter(currentInput, stmt.having, selectContext, aggregateOutputScope, aggregates, groupByExpressions);
-	}
-
-	// Determine if final projection is needed
-	const needsFinalProjection = checkNeedsFinalProjection(projections);
-
-	return {
-		output: currentInput,
-		aggregateScope: aggregateOutputScope,
-		needsFinalProjection,
-		preAggregateSort,
-		aggregateNode,
-		groupByExpressions
-	};
-}
-
-/**
- * Handles pre-aggregate sorting for special cases
- */
-function handlePreAggregateSort(
-	input: RelationalPlanNode,
-	stmt: AST.SelectStmt,
-	selectContext: PlanningContext,
-	hasAggregates: boolean,
-	hasGroupBy: boolean
-): RelationalPlanNode {
-	// Special handling for ORDER BY with aggregates but no GROUP BY
-	if (hasAggregates && !hasGroupBy && stmt.orderBy && stmt.orderBy.length > 0) {
-		// Apply ORDER BY before aggregation
-		const sortKeys: SortKey[] = stmt.orderBy.map(orderByClause => {
-			const expression = buildExpression(selectContext, orderByClause.expr);
-			return {
-				expression,
-				direction: orderByClause.direction,
-				nulls: orderByClause.nulls
-			};
-		});
-
-		return new SortNode(selectContext.scope, input, sortKeys);
-	}
-
-	return input;
-}
-
-/**
- * Validates that aggregate and non-aggregate projections don't mix inappropriately
- */
-function validateAggregateProjections(
-	projections: Projection[],
-	hasAggregates: boolean,
-	hasGroupBy: boolean
-): void {
-	if (projections.length > 0 && hasAggregates && !hasGroupBy) {
-		throw new QuereusError(
-			'Cannot mix aggregate and non-aggregate columns in SELECT list without GROUP BY',
-			StatusCode.ERROR
-		);
-	}
-}
-
-/**
- * Creates a scope that includes the aggregate output columns
- */
-function createAggregateOutputScope(
-	parentScope: Scope,
-	aggregateNode: RelationalPlanNode,
-	groupByExpressions: ScalarPlanNode[],
-	aggregates: { expression: ScalarPlanNode; alias: string }[]
-): RegisteredScope {
-	const aggregateOutputScope = new RegisteredScope(parentScope);
-	const aggregateAttributes = aggregateNode.getAttributes();
-
-	// Register GROUP BY columns
-	groupByExpressions.forEach((expr, index) => {
-		const attr = aggregateAttributes[index];
-		aggregateOutputScope.registerSymbol(attr.name.toLowerCase(), (exp, s) =>
-			new ColumnReferenceNode(s, exp as AST.ColumnExpr, expr.getType(), attr.id, index));
-	});
-
-	// Register aggregate columns by their aliases
-	aggregates.forEach((agg, index) => {
-		const columnIndex = groupByExpressions.length + index;
-		const attr = aggregateAttributes[columnIndex];
-		aggregateOutputScope.registerSymbol(agg.alias.toLowerCase(), (exp, s) =>
-			new ColumnReferenceNode(s, exp as AST.ColumnExpr, agg.expression.getType(), attr.id, columnIndex));
-	});
-
-			// Register source columns for HAVING clause access
-	// Start after GROUP BY and aggregate columns
-	const sourceColumnStartIndex = groupByExpressions.length + aggregates.length;
-	for (let i = sourceColumnStartIndex; i < aggregateAttributes.length; i++) {
-		const attr = aggregateAttributes[i];
-		// Only register if not already registered (avoid conflicts with GROUP BY columns)
-		const symbolName = attr.name.toLowerCase();
-		const existingSymbols = aggregateOutputScope.getSymbols();
-		const alreadyRegistered = existingSymbols.some(([key]) => key === symbolName);
-		if (!alreadyRegistered) {
-			aggregateOutputScope.registerSymbol(symbolName, (exp, s) =>
-				new ColumnReferenceNode(s, exp as AST.ColumnExpr, attr.type, attr.id, i));
-		}
-	}
-
-	return aggregateOutputScope;
-}
-
-/**
- * Builds HAVING filter clause
- */
-function buildHavingFilter(
-	input: RelationalPlanNode,
-	havingClause: AST.Expression,
-	selectContext: PlanningContext,
-	aggregateOutputScope: RegisteredScope,
-	aggregates: { expression: ScalarPlanNode; alias: string }[],
-	groupByExpressions: ScalarPlanNode[]
-): RelationalPlanNode {
-	const aggregateAttributes = input.getAttributes();
-
-	// Create a hybrid scope that first tries the aggregate output scope,
-	// then falls back to the original source scope for column resolution
-	const hybridScope = new RegisteredScope();
-
-	// Copy all symbols from aggregate output scope
-	for (const [symbolKey, callback] of aggregateOutputScope.getSymbols()) {
-		hybridScope.registerSymbol(symbolKey, callback);
-	}
-
-	// For any source columns not already registered, register them with
-	// references to the source table
-	const sourceInput = input.getRelations()[0]; // The AggregateNode's source
-	const sourceAttributes = sourceInput.getAttributes();
-
-	sourceAttributes.forEach((sourceAttr, sourceIndex) => {
-		const symbolName = sourceAttr.name.toLowerCase();
-		const existingSymbols = hybridScope.getSymbols();
-		const alreadyRegistered = existingSymbols.some(([key]) => key === symbolName);
-
-		if (!alreadyRegistered) {
-			hybridScope.registerSymbol(symbolName, (exp, s) =>
-				new ColumnReferenceNode(s, exp as AST.ColumnExpr, sourceAttr.type, sourceAttr.id, sourceIndex));
-		}
-	});
-
-	// Build HAVING expression with the hybrid scope
-	const havingContext: PlanningContext = {
-		...selectContext,
-		scope: hybridScope,
-		aggregates: aggregates.map((agg, index) => {
-			const columnIndex = groupByExpressions.length + index;
-			const attr = aggregateAttributes[columnIndex];
-			return {
-				expression: agg.expression,
-				alias: agg.alias,
-				columnIndex,
-				attributeId: attr.id
-			};
-		})
-	};
-
-	const havingExpression = buildExpression(havingContext, havingClause, true);
-
-	return new FilterNode(hybridScope, input, havingExpression);
-}
-
-/**
- * Checks if a final projection is needed for complex expressions
- */
-function checkNeedsFinalProjection(projections: Projection[]): boolean {
-	if (projections.length === 0) {
-		return false;
-	}
-
-	// Check if any of the projections are complex expressions (not just column refs)
-	return projections.some(proj => {
-		// If it's not a simple ColumnReferenceNode, we need final projection
-		return !CapabilityDetectors.isColumnReference(proj.node);
-	});
-}
-
-/**
- * Builds final projections for the complete SELECT list in aggregate context
- */
-export function buildFinalAggregateProjections(
-	stmt: AST.SelectStmt,
-	selectContext: PlanningContext,
-	aggregateOutputScope: RegisteredScope,
-	aggregateNode: RelationalPlanNode,
-	aggregates: { expression: ScalarPlanNode; alias: string }[],
-	groupByExpressions: ScalarPlanNode[]
-): Projection[] {
-	const finalProjections: Projection[] = [];
-	const aggregateAttributes = aggregateNode.getAttributes();
-
-	// Build context with aggregates so buildFunctionCall can resolve aggregate references
-	const aggregatesContext = aggregates.map((agg, index) => {
-		const columnIndex = groupByExpressions.length + index;
-		const attr = aggregateAttributes[columnIndex];
-		return {
-			expression: agg.expression,
-			alias: agg.alias,
-			columnIndex,
-			attributeId: attr.id
-		};
-	});
-
-	for (const column of stmt.columns) {
-		if (column.type === 'column') {
-			// Re-build the expression in the context of the aggregate output
-			const finalContext: PlanningContext = {
-				...selectContext,
-				scope: aggregateOutputScope,
-				aggregates: aggregatesContext
-			};
-			const scalarNode = buildExpression(finalContext, column.expr, true);
-
-			let attrId: number | undefined = undefined;
-			if (CapabilityDetectors.isColumnReference(scalarNode)) {
-				attrId = scalarNode.attributeId;
-			}
-
-			finalProjections.push({
-				node: scalarNode,
-				alias: column.alias || (column.expr.type === 'column' ? column.expr.name : undefined),
-				attributeId: attrId
-			});
-		}
-	}
-
-	return finalProjections;
-}

package/src/planner/building/select-compound.ts

@@ -1,119 +0,0 @@
-import type * as AST from '../../parser/ast.js';
-import type { RelationalPlanNode } from '../nodes/plan-node.js';
-import type { PlanningContext } from '../planning-context.js';
-import type { CTEScopeNode } from '../nodes/cte-node.js';
-import type { Scope } from '../scopes/scope.js';
-import { SetOperationNode } from '../nodes/set-operation-node.js';
-import { SortNode, type SortKey } from '../nodes/sort.js';
-import { LimitOffsetNode } from '../nodes/limit-offset.js';
-import { LiteralNode } from '../nodes/scalar.js';
-import { RegisteredScope } from '../scopes/registered.js';
-import { ColumnReferenceNode } from '../nodes/reference.js';
-import { buildExpression } from './expression.js';
-// Import will be added after refactoring select.ts
-import { QuereusError } from '../../common/errors.js';
-import { StatusCode } from '../../common/types.js';
-
-/**
- * Builds a compound SELECT statement (UNION, INTERSECT, EXCEPT)
- */
-export function buildCompoundSelect(
-	stmt: AST.SelectStmt,
-	contextWithCTEs: PlanningContext,
-	cteNodes: Map<string, CTEScopeNode>,
-	buildSelectStmt: (ctx: PlanningContext, stmt: AST.SelectStmt, parentCTEs?: Map<string, CTEScopeNode>) => RelationalPlanNode
-): RelationalPlanNode {
-	if (!stmt.compound) {
-		throw new QuereusError('buildCompoundSelect called without compound clause', StatusCode.INTERNAL);
-	}
-
-	// Build left side by cloning the statement without compound and stripping ORDER BY/LIMIT/OFFSET that belong to outer query
-	const { compound: _outerCompound, orderBy: outerOrderBy, limit: outerLimit, offset: outerOffset, ...leftCore } = stmt;
-
-	// Also strip ORDER BY/LIMIT/OFFSET from the right side - they should only apply to the final compound result
-	const { orderBy: _rightOrderBy, limit: _rightLimit, offset: _rightOffset, ...rightCore } = stmt.compound.select;
-
-	const leftPlan = buildSelectStmt(contextWithCTEs, leftCore as AST.SelectStmt, cteNodes) as RelationalPlanNode;
-	const rightPlan = buildSelectStmt(contextWithCTEs, rightCore as AST.SelectStmt, cteNodes) as RelationalPlanNode;
-
-	// Expand DIFF as (A EXCEPT B) UNION (B EXCEPT A)
-	let setNode: RelationalPlanNode;
-	if (stmt.compound.op === 'diff') {
-		const leftMinusRight = new SetOperationNode(contextWithCTEs.scope, leftPlan, rightPlan, 'except');
-		const rightMinusLeft = new SetOperationNode(contextWithCTEs.scope, rightPlan, leftPlan, 'except');
-		setNode = new SetOperationNode(contextWithCTEs.scope, leftMinusRight, rightMinusLeft, 'union');
-	} else {
-		setNode = new SetOperationNode(contextWithCTEs.scope, leftPlan, rightPlan, stmt.compound.op);
-	}
-
-	// After set operation, apply ORDER BY / LIMIT / OFFSET from the *outer* (original) statement
-	let input: RelationalPlanNode = setNode;
-
-	// Build scope for output columns
-	const setScope = createSetOperationScope(input);
-	const selectContext: PlanningContext = { ...contextWithCTEs, scope: setScope };
-
-	// Apply outer modifiers
-	input = applyOuterOrderBy(input, outerOrderBy, selectContext);
-	input = applyOuterLimitOffset(input, outerLimit, outerOffset, selectContext);
-
-	return input;
-}
-
-/**
- * Creates a scope for set operation output columns
- */
-function createSetOperationScope(setNode: RelationalPlanNode): RegisteredScope {
-	const setScope = new RegisteredScope();
-	const attrs = setNode.getAttributes();
-
-	setNode.getType().columns.forEach((c, i: number) => {
-		const attr = attrs[i];
-		// Ensure column has a name - use attribute name as fallback
-		const columnName = c.name || attr.name;
-		if (!columnName) {
-			throw new QuereusError(`Column at index ${i} has no name in set operation`, StatusCode.ERROR);
-		}
-		setScope.registerSymbol(columnName.toLowerCase(), (exp: AST.Expression, s: Scope) =>
-			new ColumnReferenceNode(s, exp as AST.ColumnExpr, c.type, attr.id, i));
-	});
-
-	return setScope;
-}
-
-/**
- * Applies ORDER BY clause from outer compound statement
- */
-function applyOuterOrderBy(
-	input: RelationalPlanNode,
-	outerOrderBy: AST.OrderByClause[] | undefined,
-	selectContext: PlanningContext
-): RelationalPlanNode {
-	if (outerOrderBy && outerOrderBy.length > 0) {
-		const sortKeys: SortKey[] = outerOrderBy.map((ob) => ({
-			expression: buildExpression(selectContext, ob.expr),
-			direction: ob.direction,
-			nulls: ob.nulls,
-		}));
-		return new SortNode(selectContext.scope, input, sortKeys);
-	}
-	return input;
-}
-
-/**
- * Applies LIMIT and OFFSET clauses from outer compound statement
- */
-function applyOuterLimitOffset(
-	input: RelationalPlanNode,
-	outerLimit: AST.Expression | undefined,
-	outerOffset: AST.Expression | undefined,
-	selectContext: PlanningContext
-): RelationalPlanNode {
-	if (outerLimit || outerOffset) {
-		const literalNull = new LiteralNode(selectContext.scope, { type: 'literal', value: null });
-		const limitExpr = outerLimit ? buildExpression(selectContext, outerLimit) : literalNull;
-		const offsetExpr = outerOffset ? buildExpression(selectContext, outerOffset) : literalNull;
-		return new LimitOffsetNode(selectContext.scope, input, limitExpr, offsetExpr);
-	}
-	return input;
-}

package/src/planner/building/select-context.ts

@@ -1,85 +0,0 @@
-import type * as AST from '../../parser/ast.js';
-import type { PlanningContext } from '../planning-context.js';
-import type { CTEScopeNode } from '../nodes/cte-node.js';
-import type { Scope } from '../scopes/scope.js';
-import { RegisteredScope } from '../scopes/registered.js';
-import { ParameterScope } from '../scopes/param.js';
-import { ColumnReferenceNode } from '../nodes/reference.js';
-import { buildWithClause } from './with.js';
-
-/**
- * Helper function to get the non-parameter ancestor scope.
- * This ensures table/column scopes don't inherit from ParameterScope,
- * preventing parameter resolution ambiguity in MultiScope.
- */
-export function getNonParamAncestor(scope: Scope): Scope {
-	return (scope instanceof ParameterScope) ? scope.parentScope : scope;
-}
-
-/**
- * Builds context with CTEs if present
- */
-export function buildWithContext(
-	ctx: PlanningContext,
-	stmt: AST.SelectStmt,
-	parentCTEs: Map<string, CTEScopeNode> = new Map()
-): {
-	contextWithCTEs: PlanningContext;
-	cteNodes: Map<string, CTEScopeNode>;
-} {
-	// Start with parent CTEs - either from parameter or from context
-	const cteNodes: Map<string, CTEScopeNode> = new Map(parentCTEs.size > 0 ? parentCTEs : (ctx.cteNodes ?? new Map()));
-	let contextWithCTEs = ctx;
-
-	if (stmt.withClause) {
-		const newCteNodes = buildWithClause(ctx, stmt.withClause);
-		// Merge parent CTEs with new ones (new ones take precedence)
-		for (const [name, node] of newCteNodes) {
-			cteNodes.set(name, node);
-		}
-
-		// Create a new scope that includes the CTEs
-		const cteScope = createCTEScope(cteNodes, ctx);
-		contextWithCTEs = { ...ctx, scope: cteScope, cteNodes, cteReferenceCache: ctx.cteReferenceCache };
-	} else if (cteNodes.size > 0) {
-		// No WITH clause but we have parent CTEs, create scope for them
-		const cteScope = createCTEScope(cteNodes, ctx);
-		contextWithCTEs = { ...ctx, scope: cteScope, cteNodes, cteReferenceCache: ctx.cteReferenceCache };
-	}
-
-	return { contextWithCTEs, cteNodes };
-}
-
-/**
- * Creates a scope that includes CTE references
- * CRITICAL: Uses stable input attribute IDs only, ignoring any projection output scopes
- * that might cause attribute ID collisions in correlated subqueries
- */
-function createCTEScope(
-	cteNodes: Map<string, CTEScopeNode>,
-	ctx: PlanningContext
-): RegisteredScope {
-	const cteScope = new RegisteredScope(getNonParamAncestor(ctx.scope));
-
-	// Register each CTE in the scope
-	for (const [cteName, cteNode] of cteNodes) {
-		// CRITICAL: Use only the stable input attributes from the CTE definition
-		// Do NOT use any projection output attributes that might have fresh IDs
-		const attributes = cteNode.getAttributes();
-		const columnTypes = cteNode.getType().columns;
-
-		// Only register columns that are stable input attributes
-		// This prevents scope pollution from projection output attributes
-		columnTypes.forEach((col, i) => {
-			if (i < attributes.length) {
-				const attr = attributes[i];
-				// Register CTE columns with qualified names to avoid collisions
-				const qualifiedColumnName = `${cteName}.${col.name.toLowerCase()}`;
-				cteScope.registerSymbol(qualifiedColumnName, (exp, s) =>
-					new ColumnReferenceNode(s, exp as AST.ColumnExpr, col.type, attr.id, i));
-			}
-		});
-	}
-
-	return cteScope;
-}