squirreling 0.12.22 → 0.12.23

package/README.md

@@ -141,7 +141,7 @@ Squirreling mostly follows the SQL standard. The following features are supporte
 - `SELECT` statements with `DISTINCT`, `WHERE`, `ORDER BY`, `LIMIT`, `OFFSET`
 - `WITH` clause for Common Table Expressions (CTEs)
 - Subqueries in `SELECT`, `FROM`, `WHERE`, and correlated subqueries
-- `JOIN` operations: `INNER JOIN`, `LEFT JOIN`, `RIGHT JOIN`, `FULL JOIN`, `CROSS JOIN`, `POSITIONAL JOIN`, `LATERAL VIEW [OUTER] EXPLODE(...)`
+- `JOIN` operations: `INNER JOIN`, `LEFT JOIN`, `RIGHT JOIN`, `FULL JOIN`, `CROSS JOIN`, `POSITIONAL JOIN`, `LATERAL VIEW [OUTER] EXPLODE(...)`, with `ON` or `USING (col, ...)` conditions
 - `GROUP BY` and `HAVING` clauses
 - Set operations: `UNION`, `UNION ALL`, `INTERSECT`, `INTERSECT ALL`, `EXCEPT`, `EXCEPT ALL`
 - Expressions: `CASE`, `CAST`, `BETWEEN`, `IN`, `LIKE`, `IS NULL`, `IS NOT NULL`
@@ -160,7 +160,7 @@ Squirreling mostly follows the SQL standard. The following features are supporte
 - Math: `ABS`, `SIGN`, `CEIL`, `FLOOR`, `ROUND`, `MOD`, `RAND`, `RANDOM`, `LN`, `LOG10`, `EXP`, `POWER`, `SQRT`
 - Trig: `SIN`, `COS`, `TAN`, `COT`, `ASIN`, `ACOS`, `ATAN`, `ATAN2`, `DEGREES`, `RADIANS`, `PI`
 - Date: `CURRENT_DATE`, `CURRENT_TIME`, `CURRENT_TIMESTAMP`, `DATE_DIFF`, `DATEDIFF`, `DATE_PART`, `DATE_TRUNC`, `EPOCH`, `EXTRACT`, `INTERVAL`
-- Json: `JSON_VALUE`, `JSON_QUERY`, `JSON_EXTRACT`, `JSON_OBJECT`, `JSON_ARRAY_LENGTH`, `JSON_VALID`, `JSON_TYPE`
+- Json: `JSON_VALUE`, `JSON_QUERY`, `JSON_EXTRACT`, `JSON_OBJECT`, `JSON_ARRAY_LENGTH`, `JSON_VALID`, `JSON_TYPE`, `JSON_KEYS`
 - Array: `ARRAY_LENGTH`, `ARRAY_POSITION`, `ARRAY_CONTAINS`, `ARRAY_SORT`, `ARRAY_APPEND`, `ARRAY_CONCAT`, `LEN`, `CARDINALITY`, `SIZE`
 - Table functions: `UNNEST`, `EXPLODE`, `JSON_EACH`
 - Regex: `REGEXP_SUBSTR`, `REGEXP_EXTRACT`, `REGEXP_REPLACE`, `REGEXP_MATCHES`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "squirreling",
-  "version": "0.12.22",
+  "version": "0.12.23",
   "description": "Squirreling Async SQL Engine",
   "author": "Hyperparam",
   "homepage": "https://hyperparam.app",

package/src/ast.d.ts

@@ -210,6 +210,7 @@ export interface JoinClause extends AstBase {
   table: string
   alias?: string
   on?: ExprNode
+  using?: string[]
   fromFunction?: FromFunction
 }
 

package/src/expression/evaluate.js

@@ -332,7 +332,7 @@ export async function evaluateExpr({ node, row, rowIndex, rows, context }) {
         return values[lower] + (values[upper] - values[lower]) * (pos - lower)
       }
 
-      if (funcName === 'JSON_ARRAYAGG' || funcName === 'ARRAY_AGG') {
+      if (funcName === 'JSON_ARRAYAGG' || funcName === 'ARRAY_AGG' || funcName === 'LIST') {
         const allValues = await evaluateAll(argNode, filteredRows, context)
         if (node.distinct) {
           /** @type {SqlPrimitive[]} */
@@ -516,6 +516,25 @@ export async function evaluateExpr({ node, row, rowIndex, rows, context }) {
       return typeof value
     }
 
+    if (funcName === 'JSON_KEYS') {
+      let value = args[0]
+      if (value == null) return null
+      if (typeof value === 'string') {
+        try {
+          value = JSON.parse(value)
+        } catch {
+          throw new ArgValueError({
+            ...node,
+            message: 'invalid JSON string',
+            hint: 'Argument must be valid JSON.',
+            rowIndex,
+          })
+        }
+      }
+      if (typeof value !== 'object' || value === null || Array.isArray(value) || value instanceof Date) return null
+      return Object.keys(value)
+    }
+
     if (funcName === 'JSON_ARRAY_LENGTH') {
       let arr = args[0]
       if (arr == null) return null

package/src/parse/joins.js

@@ -222,13 +222,26 @@ export function parseJoins(state) {
     const tableTok = expect(state, 'identifier')
     const tableAlias = parseTableAlias(state)
 
-    // Parse ON condition (not for POSITIONAL joins)
+    // Parse ON condition or USING column list (not for POSITIONAL joins)
     /** @type {ExprNode | undefined} */
     let condition
+    /** @type {string[] | undefined} */
+    let using
     if (joinType !== 'POSITIONAL') {
-      expect(state, 'keyword', 'ON')
-      condition = parseExpression(state)
-      expectNoAggregate(condition, 'JOIN ON')
+      if (match(state, 'keyword', 'USING')) {
+        expect(state, 'paren', '(')
+        using = []
+        while (true) {
+          const colTok = expect(state, 'identifier')
+          using.push(colTok.value)
+          if (!match(state, 'comma')) break
+        }
+        expect(state, 'paren', ')')
+      } else {
+        expect(state, 'keyword', 'ON')
+        condition = parseExpression(state)
+        expectNoAggregate(condition, 'JOIN ON')
+      }
     }
 
     joins.push({
@@ -236,6 +249,7 @@ export function parseJoins(state) {
       table: tableTok.value,
       alias: tableAlias,
       on: condition,
+      using,
       positionStart: tok.positionStart,
       positionEnd: tableTok.positionEnd,
     })

package/src/plan/columns.js

@@ -153,6 +153,13 @@ export function extractColumns({ select, parentColumns }) {
   if (sourceAlias !== undefined) visibleLateralAliases.push(sourceAlias)
   for (const join of select.joins) {
     collectColumnsFromExpr(join.on, identifiers)
+    // USING columns are equi-join keys on both sides; keep them in every
+    // table's needed set so projection pushdown can't prune the join key.
+    if (join.using) {
+      for (const col of join.using) {
+        for (const [, set] of perTable) set?.add(col)
+      }
+    }
     const joinAlias = join.alias ?? join.table
     if (join.fromFunction) {
       /** @type {IdentifierNode[]} */

package/src/plan/plan.js

@@ -463,7 +463,10 @@ function planJoin({ left, joins, leftTable, ctePlans, cteColumns, perTableColumn
     if (join.joinType === 'POSITIONAL') {
       plan = { type: 'PositionalJoin', leftAlias: currentLeftTable, rightAlias: rightTable, left: plan, right: rightScan }
     } else {
-      const keys = join.on && extractEquiKeys({ condition: join.on, leftTable: currentLeftTable, rightTable })
+      // `USING (cols)` desugars to an equi-condition `left.col = right.col` per
+      // column, which routes through the hash-join path like any other ON.
+      const condition = join.on ?? (join.using && buildUsingCondition(join.using, join))
+      const keys = condition && extractEquiKeys({ condition, leftTable: currentLeftTable, rightTable })
       if (keys) {
         /** @type {HashJoinNode} */
         const hashJoin = {
@@ -484,7 +487,7 @@ function planJoin({ left, joins, leftTable, ctePlans, cteColumns, perTableColumn
           joinType: join.joinType,
           leftAlias: currentLeftTable,
           rightAlias: rightTable,
-          condition: join.on,
+          condition,
           left: plan,
           right: rightScan,
         }
@@ -623,6 +626,37 @@ function normalizeIdentifiers(node, sourceColumns) {
   return node
 }
 
+/**
+ * Builds the join condition for a `JOIN ... USING (cols)` clause: an AND of
+ * `col = col` equalities using unprefixed identifiers. The hash-join path
+ * evaluates the left key against the left row and the right key against the
+ * right row, so each unqualified name resolves unambiguously on its own side.
+ *
+ * @param {string[]} using - shared column names from the USING clause
+ * @param {{ positionStart: number, positionEnd: number }} pos - position info for the synthesized exprs
+ * @returns {ExprNode | undefined}
+ */
+function buildUsingCondition(using, pos) {
+  const { positionStart, positionEnd } = pos
+  /** @type {ExprNode | undefined} */
+  let condition
+  for (const col of using) {
+    /** @type {ExprNode} */
+    const eq = {
+      type: 'binary',
+      op: '=',
+      left: { type: 'identifier', name: col, positionStart, positionEnd },
+      right: { type: 'identifier', name: col, positionStart, positionEnd },
+      positionStart,
+      positionEnd,
+    }
+    condition = condition === undefined
+      ? eq
+      : { type: 'binary', op: 'AND', left: condition, right: eq, positionStart, positionEnd }
+  }
+  return condition
+}
+
 /**
  * Splits a join ON expression into equi-key pairs and a residual predicate so
  * the planner can route AND-of-equis (with optional range/inequality

package/src/types.d.ts

@@ -129,7 +129,7 @@ export interface UserDefinedFunction {
   arguments: FunctionSignature
 }
 
-export type AggregateFunc = 'COUNT' | 'COUNTIF' | 'SUM' | 'AVG' | 'MIN' | 'MAX' | 'ARRAY_AGG' | 'JSON_ARRAYAGG' | 'STDDEV_SAMP' | 'STDDEV_POP' | 'MEDIAN' | 'PERCENTILE_CONT' | 'APPROX_QUANTILE' | 'STRING_AGG'
+export type AggregateFunc = 'COUNT' | 'COUNTIF' | 'SUM' | 'AVG' | 'MIN' | 'MAX' | 'ARRAY_AGG' | 'LIST' | 'JSON_ARRAYAGG' | 'STDDEV_SAMP' | 'STDDEV_POP' | 'MEDIAN' | 'PERCENTILE_CONT' | 'APPROX_QUANTILE' | 'STRING_AGG'
 
 export type RegExpFunction = 'REGEXP_SUBSTR' | 'REGEXP_EXTRACT' | 'REGEXP_REPLACE' | 'REGEXP_MATCHES'
 

package/src/validation/functions.js

@@ -11,7 +11,7 @@ export const niladicFuncs = ['CURRENT_DATE', 'CURRENT_TIME', 'CURRENT_TIMESTAMP'
  * @returns {name is AggregateFunc}
  */
 export function isAggregateFunc(name) {
-  return ['COUNT', 'COUNTIF', 'SUM', 'AVG', 'MIN', 'MAX', 'ARRAY_AGG', 'JSON_ARRAYAGG', 'STDDEV_SAMP', 'STDDEV_POP', 'MEDIAN', 'PERCENTILE_CONT', 'APPROX_QUANTILE', 'STRING_AGG'].includes(name)
+  return ['COUNT', 'COUNTIF', 'SUM', 'AVG', 'MIN', 'MAX', 'ARRAY_AGG', 'LIST', 'JSON_ARRAYAGG', 'STDDEV_SAMP', 'STDDEV_POP', 'MEDIAN', 'PERCENTILE_CONT', 'APPROX_QUANTILE', 'STRING_AGG'].includes(name)
 }
 
 /**
@@ -179,8 +179,10 @@ export const FUNCTION_SIGNATURES = {
   JSON_ARRAY_LENGTH: { min: 1, max: 1, signature: 'array' },
   JSON_VALID: { min: 1, max: 1, signature: 'value' },
   JSON_TYPE: { min: 1, max: 1, signature: 'value' },
+  JSON_KEYS: { min: 1, max: 1, signature: 'value' },
   JSON_ARRAYAGG: { min: 1, max: 1, signature: 'expression' },
   ARRAY_AGG: { min: 1, max: 1, signature: 'expression' },
+  LIST: { min: 1, max: 1, signature: 'expression' },
 
   // Array functions
   ARRAY_LENGTH: { min: 1, max: 2, signature: 'array[, dimension]' },

package/src/validation/keywords.js

@@ -3,7 +3,7 @@ export const KEYWORDS = new Set([
   'HAVING', 'ORDER', 'ASC', 'DESC', 'NULLS', 'LIMIT', 'OFFSET', 'AS', 'ALL',
   'DISTINCT', 'TRUE', 'FALSE', 'NULL', 'LIKE', 'IN', 'EXISTS', 'BETWEEN',
   'CASE', 'WHEN', 'THEN', 'ELSE', 'END', 'JOIN', 'INNER', 'LEFT', 'RIGHT',
-  'FULL', 'OUTER', 'CROSS', 'POSITIONAL', 'LATERAL', 'VIEW', 'ON', 'INTERVAL', 'DAY', 'MONTH', 'YEAR',
+  'FULL', 'OUTER', 'CROSS', 'POSITIONAL', 'LATERAL', 'VIEW', 'ON', 'USING', 'INTERVAL', 'DAY', 'MONTH', 'YEAR',
   'HOUR', 'MINUTE', 'SECOND', 'FILTER', 'WITHIN',
   'UNION', 'INTERSECT', 'EXCEPT',
 ])
@@ -17,7 +17,7 @@ export const RESERVED_KEYWORDS = new Set([
   'EXISTS', 'CASE', 'WHEN', 'THEN', 'ELSE', 'END', 'INTERVAL',
   'GROUP', 'BY', 'HAVING', 'ORDER', 'LIMIT', 'OFFSET',
   'AS', 'ALL', 'DISTINCT',
-  'JOIN', 'INNER', 'LEFT', 'RIGHT', 'FULL', 'OUTER', 'ON',
+  'JOIN', 'INNER', 'LEFT', 'RIGHT', 'FULL', 'OUTER', 'ON', 'USING',
   'UNION', 'INTERSECT', 'EXCEPT',
 ])
 
@@ -30,6 +30,6 @@ export const RESERVED_AFTER_COLUMN = new Set([
 // Keywords that cannot be used as table aliases
 export const RESERVED_AFTER_TABLE = new Set([
   'WHERE', 'GROUP', 'HAVING', 'ORDER', 'LIMIT', 'OFFSET', 'JOIN', 'INNER',
-  'LEFT', 'RIGHT', 'FULL', 'CROSS', 'ON', 'POSITIONAL', 'LATERAL',
+  'LEFT', 'RIGHT', 'FULL', 'CROSS', 'ON', 'USING', 'POSITIONAL', 'LATERAL',
   'UNION', 'INTERSECT', 'EXCEPT',
 ])