@tanstack/db 0.0.23 → 0.0.24

package/dist/esm/query/compiler/index.js

@@ -1,20 +1,25 @@
 import { map, filter, distinct } from "@electric-sql/d2mini";
+import { optimizeQuery } from "../optimizer.js";
 import { compileExpression } from "./evaluators.js";
 import { processJoins } from "./joins.js";
 import { processGroupBy } from "./group-by.js";
 import { processOrderBy } from "./order-by.js";
 import { processSelectToResults } from "./select.js";
-function compileQuery(query, inputs, cache = /* @__PURE__ */ new WeakMap()) {
-  const cachedResult = cache.get(query);
+function compileQuery(rawQuery, inputs, cache = /* @__PURE__ */ new WeakMap(), queryMapping = /* @__PURE__ */ new WeakMap()) {
+  const cachedResult = cache.get(rawQuery);
   if (cachedResult) {
     return cachedResult;
   }
+  const query = optimizeQuery(rawQuery);
+  queryMapping.set(query, rawQuery);
+  mapNestedQueries(query, rawQuery, queryMapping);
   const allInputs = { ...inputs };
   const tables = {};
   const { alias: mainTableAlias, input: mainInput } = processFrom(
     query.from,
     allInputs,
-    cache
+    cache,
+    queryMapping
   );
   tables[mainTableAlias] = mainInput;
   let pipeline = mainInput.pipe(
@@ -30,7 +35,8 @@ function compileQuery(query, inputs, cache = /* @__PURE__ */ new WeakMap()) {
       tables,
       mainTableAlias,
       allInputs,
-      cache
+      cache,
+      queryMapping
     );
   }
   if (query.where && query.where.length > 0) {
@@ -139,7 +145,7 @@ function compileQuery(query, inputs, cache = /* @__PURE__ */ new WeakMap()) {
       })
     );
     const result2 = resultPipeline2;
-    cache.set(query, result2);
+    cache.set(rawQuery, result2);
     return result2;
   } else if (query.limit !== void 0 || query.offset !== void 0) {
     throw new Error(
@@ -153,10 +159,10 @@ function compileQuery(query, inputs, cache = /* @__PURE__ */ new WeakMap()) {
     })
   );
   const result = resultPipeline;
-  cache.set(query, result);
+  cache.set(rawQuery, result);
   return result;
 }
-function processFrom(from, allInputs, cache) {
+function processFrom(from, allInputs, cache, queryMapping) {
   switch (from.type) {
     case `collectionRef`: {
       const input = allInputs[from.collection.id];
@@ -168,7 +174,13 @@ function processFrom(from, allInputs, cache) {
       return { alias: from.alias, input };
     }
     case `queryRef`: {
-      const subQueryInput = compileQuery(from.query, allInputs, cache);
+      const originalQuery = queryMapping.get(from.query) || from.query;
+      const subQueryInput = compileQuery(
+        originalQuery,
+        allInputs,
+        cache,
+        queryMapping
+      );
       const extractedInput = subQueryInput.pipe(
         map((data) => {
           const [key, [value, _orderByIndex]] = data;
@@ -181,6 +193,30 @@ function processFrom(from, allInputs, cache) {
       throw new Error(`Unsupported FROM type: ${from.type}`);
   }
 }
+function mapNestedQueries(optimizedQuery, originalQuery, queryMapping) {
+  if (optimizedQuery.from.type === `queryRef` && originalQuery.from.type === `queryRef`) {
+    queryMapping.set(optimizedQuery.from.query, originalQuery.from.query);
+    mapNestedQueries(
+      optimizedQuery.from.query,
+      originalQuery.from.query,
+      queryMapping
+    );
+  }
+  if (optimizedQuery.join && originalQuery.join) {
+    for (let i = 0; i < optimizedQuery.join.length && i < originalQuery.join.length; i++) {
+      const optimizedJoin = optimizedQuery.join[i];
+      const originalJoin = originalQuery.join[i];
+      if (optimizedJoin.from.type === `queryRef` && originalJoin.from.type === `queryRef`) {
+        queryMapping.set(optimizedJoin.from.query, originalJoin.from.query);
+        mapNestedQueries(
+          optimizedJoin.from.query,
+          originalJoin.from.query,
+          queryMapping
+        );
+      }
+    }
+  }
+}
 export {
   compileQuery
 };

package/dist/esm/query/compiler/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"index.js","sources":["../../../../src/query/compiler/index.ts"],"sourcesContent":["import { distinct, filter, map } from \"@electric-sql/d2mini\"\nimport { compileExpression } from \"./evaluators.js\"\nimport { processJoins } from \"./joins.js\"\nimport { processGroupBy } from \"./group-by.js\"\nimport { processOrderBy } from \"./order-by.js\"\nimport { processSelectToResults } from \"./select.js\"\nimport type { CollectionRef, QueryIR, QueryRef } from \"../ir.js\"\nimport type {\n  KeyedStream,\n  NamespacedAndKeyedStream,\n  ResultStream,\n} from \"../../types.js\"\n\n/**\n * Cache for compiled subqueries to avoid duplicate compilation\n */\ntype QueryCache = WeakMap<QueryIR, ResultStream>\n\n/**\n * Compiles a query2 IR into a D2 pipeline\n * @param query The query IR to compile\n * @param inputs Mapping of collection names to input streams\n * @param cache Optional cache for compiled subqueries (used internally for recursion)\n * @returns A stream builder representing the compiled query\n */\nexport function compileQuery(\n  query: QueryIR,\n  inputs: Record<string, KeyedStream>,\n  cache: QueryCache = new WeakMap()\n): ResultStream {\n  // Check if this query has already been compiled\n  const cachedResult = cache.get(query)\n  if (cachedResult) {\n    return cachedResult\n  }\n\n  // Create a copy of the inputs map to avoid modifying the original\n  const allInputs = { ...inputs }\n\n  // Create a map of table aliases to inputs\n  const tables: Record<string, KeyedStream> = {}\n\n  // Process the FROM clause to get the main table\n  const { alias: mainTableAlias, input: mainInput } = processFrom(\n    query.from,\n    allInputs,\n    cache\n  )\n  tables[mainTableAlias] = mainInput\n\n  // Prepare the initial pipeline with the main table wrapped in its alias\n  let pipeline: NamespacedAndKeyedStream = mainInput.pipe(\n    map(([key, row]) => {\n      // Initialize the record with a nested structure\n      const ret = [key, { [mainTableAlias]: row }] as [\n        string,\n        Record<string, typeof row>,\n      ]\n      return ret\n    })\n  )\n\n  // Process JOIN clauses if they exist\n  if (query.join && query.join.length > 0) {\n    pipeline = processJoins(\n      pipeline,\n      query.join,\n      tables,\n      mainTableAlias,\n      allInputs,\n      cache\n    )\n  }\n\n  // Process the WHERE clause if it exists\n  if (query.where && query.where.length > 0) {\n    // Compile all WHERE expressions\n    const compiledWheres = query.where.map((where) => compileExpression(where))\n\n    // Apply each WHERE condition as a filter (they are ANDed together)\n    for (const compiledWhere of compiledWheres) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return compiledWhere(namespacedRow)\n        })\n      )\n    }\n  }\n\n  // Process functional WHERE clauses if they exist\n  if (query.fnWhere && query.fnWhere.length > 0) {\n    for (const fnWhere of query.fnWhere) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return fnWhere(namespacedRow)\n        })\n      )\n    }\n  }\n\n  if (query.distinct && !query.fnSelect && !query.select) {\n    throw new Error(`DISTINCT requires a SELECT clause.`)\n  }\n\n  // Process the SELECT clause early - always create __select_results\n  // This eliminates duplication and allows for DISTINCT implementation\n  if (query.fnSelect) {\n    // Handle functional select - apply the function to transform the row\n    pipeline = pipeline.pipe(\n      map(([key, namespacedRow]) => {\n        const selectResults = query.fnSelect!(namespacedRow)\n        return [\n          key,\n          {\n            ...namespacedRow,\n            __select_results: selectResults,\n          },\n        ] as [string, typeof namespacedRow & { __select_results: any }]\n      })\n    )\n  } else if (query.select) {\n    pipeline = processSelectToResults(pipeline, query.select, allInputs)\n  } else {\n    // If no SELECT clause, create __select_results with the main table data\n    pipeline = pipeline.pipe(\n      map(([key, namespacedRow]) => {\n        const selectResults =\n          !query.join && !query.groupBy\n            ? namespacedRow[mainTableAlias]\n            : namespacedRow\n\n        return [\n          key,\n          {\n            ...namespacedRow,\n            __select_results: selectResults,\n          },\n        ] as [string, typeof namespacedRow & { __select_results: any }]\n      })\n    )\n  }\n\n  // Process the GROUP BY clause if it exists\n  if (query.groupBy && query.groupBy.length > 0) {\n    pipeline = processGroupBy(\n      pipeline,\n      query.groupBy,\n      query.having,\n      query.select,\n      query.fnHaving\n    )\n  } else if (query.select) {\n    // Check if SELECT contains aggregates but no GROUP BY (implicit single-group aggregation)\n    const hasAggregates = Object.values(query.select).some(\n      (expr) => expr.type === `agg`\n    )\n    if (hasAggregates) {\n      // Handle implicit single-group aggregation\n      pipeline = processGroupBy(\n        pipeline,\n        [], // Empty group by means single group\n        query.having,\n        query.select,\n        query.fnHaving\n      )\n    }\n  }\n\n  // Process the HAVING clause if it exists (only applies after GROUP BY)\n  if (query.having && (!query.groupBy || query.groupBy.length === 0)) {\n    // Check if we have aggregates in SELECT that would trigger implicit grouping\n    const hasAggregates = query.select\n      ? Object.values(query.select).some((expr) => expr.type === `agg`)\n      : false\n\n    if (!hasAggregates) {\n      throw new Error(`HAVING clause requires GROUP BY clause`)\n    }\n  }\n\n  // Process functional HAVING clauses outside of GROUP BY (treat as additional WHERE filters)\n  if (\n    query.fnHaving &&\n    query.fnHaving.length > 0 &&\n    (!query.groupBy || query.groupBy.length === 0)\n  ) {\n    // If there's no GROUP BY but there are fnHaving clauses, apply them as filters\n    for (const fnHaving of query.fnHaving) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return fnHaving(namespacedRow)\n        })\n      )\n    }\n  }\n\n  // Process the DISTINCT clause if it exists\n  if (query.distinct) {\n    pipeline = pipeline.pipe(distinct(([_key, row]) => row.__select_results))\n  }\n\n  // Process orderBy parameter if it exists\n  if (query.orderBy && query.orderBy.length > 0) {\n    const orderedPipeline = processOrderBy(\n      pipeline,\n      query.orderBy,\n      query.limit,\n      query.offset\n    )\n\n    // Final step: extract the __select_results and include orderBy index\n    const resultPipeline = orderedPipeline.pipe(\n      map(([key, [row, orderByIndex]]) => {\n        // Extract the final results from __select_results and include orderBy index\n        const finalResults = (row as any).__select_results\n        return [key, [finalResults, orderByIndex]] as [unknown, [any, string]]\n      })\n    )\n\n    const result = resultPipeline\n    // Cache the result before returning\n    cache.set(query, result)\n    return result\n  } else if (query.limit !== undefined || query.offset !== undefined) {\n    // If there's a limit or offset without orderBy, throw an error\n    throw new Error(\n      `LIMIT and OFFSET require an ORDER BY clause to ensure deterministic results`\n    )\n  }\n\n  // Final step: extract the __select_results and return tuple format (no orderBy)\n  const resultPipeline: ResultStream = pipeline.pipe(\n    map(([key, row]) => {\n      // Extract the final results from __select_results and return [key, [results, undefined]]\n      const finalResults = (row as any).__select_results\n      return [key, [finalResults, undefined]] as [\n        unknown,\n        [any, string | undefined],\n      ]\n    })\n  )\n\n  const result = resultPipeline\n  // Cache the result before returning\n  cache.set(query, result)\n  return result\n}\n\n/**\n * Processes the FROM clause to extract the main table alias and input stream\n */\nfunction processFrom(\n  from: CollectionRef | QueryRef,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache\n): { alias: string; input: KeyedStream } {\n  switch (from.type) {\n    case `collectionRef`: {\n      const input = allInputs[from.collection.id]\n      if (!input) {\n        throw new Error(\n          `Input for collection \"${from.collection.id}\" not found in inputs map`\n        )\n      }\n      return { alias: from.alias, input }\n    }\n    case `queryRef`: {\n      // Recursively compile the sub-query with cache\n      const subQueryInput = compileQuery(from.query, allInputs, cache)\n\n      // Subqueries may return [key, [value, orderByIndex]] (with ORDER BY) or [key, value] (without ORDER BY)\n      // We need to extract just the value for use in parent queries\n      const extractedInput = subQueryInput.pipe(\n        map((data: any) => {\n          const [key, [value, _orderByIndex]] = data\n          return [key, value] as [unknown, any]\n        })\n      )\n\n      return { alias: from.alias, input: extractedInput }\n    }\n    default:\n      throw new Error(`Unsupported FROM type: ${(from as any).type}`)\n  }\n}\n"],"names":["resultPipeline","result"],"mappings":";;;;;;AAyBO,SAAS,aACd,OACA,QACA,QAAoB,oBAAI,WACV;AAEd,QAAM,eAAe,MAAM,IAAI,KAAK;AACpC,MAAI,cAAc;AAChB,WAAO;AAAA,EACT;AAGA,QAAM,YAAY,EAAE,GAAG,OAAA;AAGvB,QAAM,SAAsC,CAAA;AAG5C,QAAM,EAAE,OAAO,gBAAgB,OAAO,cAAc;AAAA,IAClD,MAAM;AAAA,IACN;AAAA,IACA;AAAA,EAAA;AAEF,SAAO,cAAc,IAAI;AAGzB,MAAI,WAAqC,UAAU;AAAA,IACjD,IAAI,CAAC,CAAC,KAAK,GAAG,MAAM;AAElB,YAAM,MAAM,CAAC,KAAK,EAAE,CAAC,cAAc,GAAG,KAAK;AAI3C,aAAO;AAAA,IACT,CAAC;AAAA,EAAA;AAIH,MAAI,MAAM,QAAQ,MAAM,KAAK,SAAS,GAAG;AACvC,eAAW;AAAA,MACT;AAAA,MACA,MAAM;AAAA,MACN;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AAGA,MAAI,MAAM,SAAS,MAAM,MAAM,SAAS,GAAG;AAEzC,UAAM,iBAAiB,MAAM,MAAM,IAAI,CAAC,UAAU,kBAAkB,KAAK,CAAC;AAG1E,eAAW,iBAAiB,gBAAgB;AAC1C,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,cAAc,aAAa;AAAA,QACpC,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,eAAW,WAAW,MAAM,SAAS;AACnC,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,QAAQ,aAAa;AAAA,QAC9B,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAEA,MAAI,MAAM,YAAY,CAAC,MAAM,YAAY,CAAC,MAAM,QAAQ;AACtD,UAAM,IAAI,MAAM,oCAAoC;AAAA,EACtD;AAIA,MAAI,MAAM,UAAU;AAElB,eAAW,SAAS;AAAA,MAClB,IAAI,CAAC,CAAC,KAAK,aAAa,MAAM;AAC5B,cAAM,gBAAgB,MAAM,SAAU,aAAa;AACnD,eAAO;AAAA,UACL;AAAA,UACA;AAAA,YACE,GAAG;AAAA,YACH,kBAAkB;AAAA,UAAA;AAAA,QACpB;AAAA,MAEJ,CAAC;AAAA,IAAA;AAAA,EAEL,WAAW,MAAM,QAAQ;AACvB,eAAW,uBAAuB,UAAU,MAAM,MAAiB;AAAA,EACrE,OAAO;AAEL,eAAW,SAAS;AAAA,MAClB,IAAI,CAAC,CAAC,KAAK,aAAa,MAAM;AAC5B,cAAM,gBACJ,CAAC,MAAM,QAAQ,CAAC,MAAM,UAClB,cAAc,cAAc,IAC5B;AAEN,eAAO;AAAA,UACL;AAAA,UACA;AAAA,YACE,GAAG;AAAA,YACH,kBAAkB;AAAA,UAAA;AAAA,QACpB;AAAA,MAEJ,CAAC;AAAA,IAAA;AAAA,EAEL;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,eAAW;AAAA,MACT;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,IAAA;AAAA,EAEV,WAAW,MAAM,QAAQ;AAEvB,UAAM,gBAAgB,OAAO,OAAO,MAAM,MAAM,EAAE;AAAA,MAChD,CAAC,SAAS,KAAK,SAAS;AAAA,IAAA;AAE1B,QAAI,eAAe;AAEjB,iBAAW;AAAA,QACT;AAAA,QACA,CAAA;AAAA;AAAA,QACA,MAAM;AAAA,QACN,MAAM;AAAA,QACN,MAAM;AAAA,MAAA;AAAA,IAEV;AAAA,EACF;AAGA,MAAI,MAAM,WAAW,CAAC,MAAM,WAAW,MAAM,QAAQ,WAAW,IAAI;AAElE,UAAM,gBAAgB,MAAM,SACxB,OAAO,OAAO,MAAM,MAAM,EAAE,KAAK,CAAC,SAAS,KAAK,SAAS,KAAK,IAC9D;AAEJ,QAAI,CAAC,eAAe;AAClB,YAAM,IAAI,MAAM,wCAAwC;AAAA,IAC1D;AAAA,EACF;AAGA,MACE,MAAM,YACN,MAAM,SAAS,SAAS,MACvB,CAAC,MAAM,WAAW,MAAM,QAAQ,WAAW,IAC5C;AAEA,eAAW,YAAY,MAAM,UAAU;AACrC,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,SAAS,aAAa;AAAA,QAC/B,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAGA,MAAI,MAAM,UAAU;AAClB,eAAW,SAAS,KAAK,SAAS,CAAC,CAAC,MAAM,GAAG,MAAM,IAAI,gBAAgB,CAAC;AAAA,EAC1E;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,IAAA;AAIR,UAAMA,kBAAiB,gBAAgB;AAAA,MACrC,IAAI,CAAC,CAAC,KAAK,CAAC,KAAK,YAAY,CAAC,MAAM;AAElC,cAAM,eAAgB,IAAY;AAClC,eAAO,CAAC,KAAK,CAAC,cAAc,YAAY,CAAC;AAAA,MAC3C,CAAC;AAAA,IAAA;AAGH,UAAMC,UAASD;AAEf,UAAM,IAAI,OAAOC,OAAM;AACvB,WAAOA;AAAAA,EACT,WAAW,MAAM,UAAU,UAAa,MAAM,WAAW,QAAW;AAElE,UAAM,IAAI;AAAA,MACR;AAAA,IAAA;AAAA,EAEJ;AAGA,QAAM,iBAA+B,SAAS;AAAA,IAC5C,IAAI,CAAC,CAAC,KAAK,GAAG,MAAM;AAElB,YAAM,eAAgB,IAAY;AAClC,aAAO,CAAC,KAAK,CAAC,cAAc,MAAS,CAAC;AAAA,IAIxC,CAAC;AAAA,EAAA;AAGH,QAAM,SAAS;AAEf,QAAM,IAAI,OAAO,MAAM;AACvB,SAAO;AACT;AAKA,SAAS,YACP,MACA,WACA,OACuC;AACvC,UAAQ,KAAK,MAAA;AAAA,IACX,KAAK,iBAAiB;AACpB,YAAM,QAAQ,UAAU,KAAK,WAAW,EAAE;AAC1C,UAAI,CAAC,OAAO;AACV,cAAM,IAAI;AAAA,UACR,yBAAyB,KAAK,WAAW,EAAE;AAAA,QAAA;AAAA,MAE/C;AACA,aAAO,EAAE,OAAO,KAAK,OAAO,MAAA;AAAA,IAC9B;AAAA,IACA,KAAK,YAAY;AAEf,YAAM,gBAAgB,aAAa,KAAK,OAAO,WAAW,KAAK;AAI/D,YAAM,iBAAiB,cAAc;AAAA,QACnC,IAAI,CAAC,SAAc;AACjB,gBAAM,CAAC,KAAK,CAAC,OAAO,aAAa,CAAC,IAAI;AACtC,iBAAO,CAAC,KAAK,KAAK;AAAA,QACpB,CAAC;AAAA,MAAA;AAGH,aAAO,EAAE,OAAO,KAAK,OAAO,OAAO,eAAA;AAAA,IACrC;AAAA,IACA;AACE,YAAM,IAAI,MAAM,0BAA2B,KAAa,IAAI,EAAE;AAAA,EAAA;AAEpE;"}
+{"version":3,"file":"index.js","sources":["../../../../src/query/compiler/index.ts"],"sourcesContent":["import { distinct, filter, map } from \"@electric-sql/d2mini\"\nimport { optimizeQuery } from \"../optimizer.js\"\nimport { compileExpression } from \"./evaluators.js\"\nimport { processJoins } from \"./joins.js\"\nimport { processGroupBy } from \"./group-by.js\"\nimport { processOrderBy } from \"./order-by.js\"\nimport { processSelectToResults } from \"./select.js\"\nimport type { CollectionRef, QueryIR, QueryRef } from \"../ir.js\"\nimport type {\n  KeyedStream,\n  NamespacedAndKeyedStream,\n  ResultStream,\n} from \"../../types.js\"\nimport type { QueryCache, QueryMapping } from \"./types.js\"\n\n/**\n * Compiles a query2 IR into a D2 pipeline\n * @param rawQuery The query IR to compile\n * @param inputs Mapping of collection names to input streams\n * @param cache Optional cache for compiled subqueries (used internally for recursion)\n * @param queryMapping Optional mapping from optimized queries to original queries\n * @returns A stream builder representing the compiled query\n */\nexport function compileQuery(\n  rawQuery: QueryIR,\n  inputs: Record<string, KeyedStream>,\n  cache: QueryCache = new WeakMap(),\n  queryMapping: QueryMapping = new WeakMap()\n): ResultStream {\n  // Check if the original raw query has already been compiled\n  const cachedResult = cache.get(rawQuery)\n  if (cachedResult) {\n    return cachedResult\n  }\n\n  // Optimize the query before compilation\n  const query = optimizeQuery(rawQuery)\n\n  // Create mapping from optimized query to original for caching\n  queryMapping.set(query, rawQuery)\n  mapNestedQueries(query, rawQuery, queryMapping)\n\n  // Create a copy of the inputs map to avoid modifying the original\n  const allInputs = { ...inputs }\n\n  // Create a map of table aliases to inputs\n  const tables: Record<string, KeyedStream> = {}\n\n  // Process the FROM clause to get the main table\n  const { alias: mainTableAlias, input: mainInput } = processFrom(\n    query.from,\n    allInputs,\n    cache,\n    queryMapping\n  )\n  tables[mainTableAlias] = mainInput\n\n  // Prepare the initial pipeline with the main table wrapped in its alias\n  let pipeline: NamespacedAndKeyedStream = mainInput.pipe(\n    map(([key, row]) => {\n      // Initialize the record with a nested structure\n      const ret = [key, { [mainTableAlias]: row }] as [\n        string,\n        Record<string, typeof row>,\n      ]\n      return ret\n    })\n  )\n\n  // Process JOIN clauses if they exist\n  if (query.join && query.join.length > 0) {\n    pipeline = processJoins(\n      pipeline,\n      query.join,\n      tables,\n      mainTableAlias,\n      allInputs,\n      cache,\n      queryMapping\n    )\n  }\n\n  // Process the WHERE clause if it exists\n  if (query.where && query.where.length > 0) {\n    // Compile all WHERE expressions\n    const compiledWheres = query.where.map((where) => compileExpression(where))\n\n    // Apply each WHERE condition as a filter (they are ANDed together)\n    for (const compiledWhere of compiledWheres) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return compiledWhere(namespacedRow)\n        })\n      )\n    }\n  }\n\n  // Process functional WHERE clauses if they exist\n  if (query.fnWhere && query.fnWhere.length > 0) {\n    for (const fnWhere of query.fnWhere) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return fnWhere(namespacedRow)\n        })\n      )\n    }\n  }\n\n  if (query.distinct && !query.fnSelect && !query.select) {\n    throw new Error(`DISTINCT requires a SELECT clause.`)\n  }\n\n  // Process the SELECT clause early - always create __select_results\n  // This eliminates duplication and allows for DISTINCT implementation\n  if (query.fnSelect) {\n    // Handle functional select - apply the function to transform the row\n    pipeline = pipeline.pipe(\n      map(([key, namespacedRow]) => {\n        const selectResults = query.fnSelect!(namespacedRow)\n        return [\n          key,\n          {\n            ...namespacedRow,\n            __select_results: selectResults,\n          },\n        ] as [string, typeof namespacedRow & { __select_results: any }]\n      })\n    )\n  } else if (query.select) {\n    pipeline = processSelectToResults(pipeline, query.select, allInputs)\n  } else {\n    // If no SELECT clause, create __select_results with the main table data\n    pipeline = pipeline.pipe(\n      map(([key, namespacedRow]) => {\n        const selectResults =\n          !query.join && !query.groupBy\n            ? namespacedRow[mainTableAlias]\n            : namespacedRow\n\n        return [\n          key,\n          {\n            ...namespacedRow,\n            __select_results: selectResults,\n          },\n        ] as [string, typeof namespacedRow & { __select_results: any }]\n      })\n    )\n  }\n\n  // Process the GROUP BY clause if it exists\n  if (query.groupBy && query.groupBy.length > 0) {\n    pipeline = processGroupBy(\n      pipeline,\n      query.groupBy,\n      query.having,\n      query.select,\n      query.fnHaving\n    )\n  } else if (query.select) {\n    // Check if SELECT contains aggregates but no GROUP BY (implicit single-group aggregation)\n    const hasAggregates = Object.values(query.select).some(\n      (expr) => expr.type === `agg`\n    )\n    if (hasAggregates) {\n      // Handle implicit single-group aggregation\n      pipeline = processGroupBy(\n        pipeline,\n        [], // Empty group by means single group\n        query.having,\n        query.select,\n        query.fnHaving\n      )\n    }\n  }\n\n  // Process the HAVING clause if it exists (only applies after GROUP BY)\n  if (query.having && (!query.groupBy || query.groupBy.length === 0)) {\n    // Check if we have aggregates in SELECT that would trigger implicit grouping\n    const hasAggregates = query.select\n      ? Object.values(query.select).some((expr) => expr.type === `agg`)\n      : false\n\n    if (!hasAggregates) {\n      throw new Error(`HAVING clause requires GROUP BY clause`)\n    }\n  }\n\n  // Process functional HAVING clauses outside of GROUP BY (treat as additional WHERE filters)\n  if (\n    query.fnHaving &&\n    query.fnHaving.length > 0 &&\n    (!query.groupBy || query.groupBy.length === 0)\n  ) {\n    // If there's no GROUP BY but there are fnHaving clauses, apply them as filters\n    for (const fnHaving of query.fnHaving) {\n      pipeline = pipeline.pipe(\n        filter(([_key, namespacedRow]) => {\n          return fnHaving(namespacedRow)\n        })\n      )\n    }\n  }\n\n  // Process the DISTINCT clause if it exists\n  if (query.distinct) {\n    pipeline = pipeline.pipe(distinct(([_key, row]) => row.__select_results))\n  }\n\n  // Process orderBy parameter if it exists\n  if (query.orderBy && query.orderBy.length > 0) {\n    const orderedPipeline = processOrderBy(\n      pipeline,\n      query.orderBy,\n      query.limit,\n      query.offset\n    )\n\n    // Final step: extract the __select_results and include orderBy index\n    const resultPipeline = orderedPipeline.pipe(\n      map(([key, [row, orderByIndex]]) => {\n        // Extract the final results from __select_results and include orderBy index\n        const finalResults = (row as any).__select_results\n        return [key, [finalResults, orderByIndex]] as [unknown, [any, string]]\n      })\n    )\n\n    const result = resultPipeline\n    // Cache the result before returning (use original query as key)\n    cache.set(rawQuery, result)\n    return result\n  } else if (query.limit !== undefined || query.offset !== undefined) {\n    // If there's a limit or offset without orderBy, throw an error\n    throw new Error(\n      `LIMIT and OFFSET require an ORDER BY clause to ensure deterministic results`\n    )\n  }\n\n  // Final step: extract the __select_results and return tuple format (no orderBy)\n  const resultPipeline: ResultStream = pipeline.pipe(\n    map(([key, row]) => {\n      // Extract the final results from __select_results and return [key, [results, undefined]]\n      const finalResults = (row as any).__select_results\n      return [key, [finalResults, undefined]] as [\n        unknown,\n        [any, string | undefined],\n      ]\n    })\n  )\n\n  const result = resultPipeline\n  // Cache the result before returning (use original query as key)\n  cache.set(rawQuery, result)\n  return result\n}\n\n/**\n * Processes the FROM clause to extract the main table alias and input stream\n */\nfunction processFrom(\n  from: CollectionRef | QueryRef,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache,\n  queryMapping: QueryMapping\n): { alias: string; input: KeyedStream } {\n  switch (from.type) {\n    case `collectionRef`: {\n      const input = allInputs[from.collection.id]\n      if (!input) {\n        throw new Error(\n          `Input for collection \"${from.collection.id}\" not found in inputs map`\n        )\n      }\n      return { alias: from.alias, input }\n    }\n    case `queryRef`: {\n      // Find the original query for caching purposes\n      const originalQuery = queryMapping.get(from.query) || from.query\n\n      // Recursively compile the sub-query with cache\n      const subQueryInput = compileQuery(\n        originalQuery,\n        allInputs,\n        cache,\n        queryMapping\n      )\n\n      // Subqueries may return [key, [value, orderByIndex]] (with ORDER BY) or [key, value] (without ORDER BY)\n      // We need to extract just the value for use in parent queries\n      const extractedInput = subQueryInput.pipe(\n        map((data: any) => {\n          const [key, [value, _orderByIndex]] = data\n          return [key, value] as [unknown, any]\n        })\n      )\n\n      return { alias: from.alias, input: extractedInput }\n    }\n    default:\n      throw new Error(`Unsupported FROM type: ${(from as any).type}`)\n  }\n}\n\n/**\n * Recursively maps optimized subqueries to their original queries for proper caching.\n * This ensures that when we encounter the same QueryRef object in different contexts,\n * we can find the original query to check the cache.\n */\nfunction mapNestedQueries(\n  optimizedQuery: QueryIR,\n  originalQuery: QueryIR,\n  queryMapping: QueryMapping\n): void {\n  // Map the FROM clause if it's a QueryRef\n  if (\n    optimizedQuery.from.type === `queryRef` &&\n    originalQuery.from.type === `queryRef`\n  ) {\n    queryMapping.set(optimizedQuery.from.query, originalQuery.from.query)\n    // Recursively map nested queries\n    mapNestedQueries(\n      optimizedQuery.from.query,\n      originalQuery.from.query,\n      queryMapping\n    )\n  }\n\n  // Map JOIN clauses if they exist\n  if (optimizedQuery.join && originalQuery.join) {\n    for (\n      let i = 0;\n      i < optimizedQuery.join.length && i < originalQuery.join.length;\n      i++\n    ) {\n      const optimizedJoin = optimizedQuery.join[i]!\n      const originalJoin = originalQuery.join[i]!\n\n      if (\n        optimizedJoin.from.type === `queryRef` &&\n        originalJoin.from.type === `queryRef`\n      ) {\n        queryMapping.set(optimizedJoin.from.query, originalJoin.from.query)\n        // Recursively map nested queries in joins\n        mapNestedQueries(\n          optimizedJoin.from.query,\n          originalJoin.from.query,\n          queryMapping\n        )\n      }\n    }\n  }\n}\n"],"names":["resultPipeline","result"],"mappings":";;;;;;;AAuBO,SAAS,aACd,UACA,QACA,QAAoB,oBAAI,WACxB,eAA6B,oBAAI,WACnB;AAEd,QAAM,eAAe,MAAM,IAAI,QAAQ;AACvC,MAAI,cAAc;AAChB,WAAO;AAAA,EACT;AAGA,QAAM,QAAQ,cAAc,QAAQ;AAGpC,eAAa,IAAI,OAAO,QAAQ;AAChC,mBAAiB,OAAO,UAAU,YAAY;AAG9C,QAAM,YAAY,EAAE,GAAG,OAAA;AAGvB,QAAM,SAAsC,CAAA;AAG5C,QAAM,EAAE,OAAO,gBAAgB,OAAO,cAAc;AAAA,IAClD,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EAAA;AAEF,SAAO,cAAc,IAAI;AAGzB,MAAI,WAAqC,UAAU;AAAA,IACjD,IAAI,CAAC,CAAC,KAAK,GAAG,MAAM;AAElB,YAAM,MAAM,CAAC,KAAK,EAAE,CAAC,cAAc,GAAG,KAAK;AAI3C,aAAO;AAAA,IACT,CAAC;AAAA,EAAA;AAIH,MAAI,MAAM,QAAQ,MAAM,KAAK,SAAS,GAAG;AACvC,eAAW;AAAA,MACT;AAAA,MACA,MAAM;AAAA,MACN;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AAGA,MAAI,MAAM,SAAS,MAAM,MAAM,SAAS,GAAG;AAEzC,UAAM,iBAAiB,MAAM,MAAM,IAAI,CAAC,UAAU,kBAAkB,KAAK,CAAC;AAG1E,eAAW,iBAAiB,gBAAgB;AAC1C,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,cAAc,aAAa;AAAA,QACpC,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,eAAW,WAAW,MAAM,SAAS;AACnC,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,QAAQ,aAAa;AAAA,QAC9B,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAEA,MAAI,MAAM,YAAY,CAAC,MAAM,YAAY,CAAC,MAAM,QAAQ;AACtD,UAAM,IAAI,MAAM,oCAAoC;AAAA,EACtD;AAIA,MAAI,MAAM,UAAU;AAElB,eAAW,SAAS;AAAA,MAClB,IAAI,CAAC,CAAC,KAAK,aAAa,MAAM;AAC5B,cAAM,gBAAgB,MAAM,SAAU,aAAa;AACnD,eAAO;AAAA,UACL;AAAA,UACA;AAAA,YACE,GAAG;AAAA,YACH,kBAAkB;AAAA,UAAA;AAAA,QACpB;AAAA,MAEJ,CAAC;AAAA,IAAA;AAAA,EAEL,WAAW,MAAM,QAAQ;AACvB,eAAW,uBAAuB,UAAU,MAAM,MAAiB;AAAA,EACrE,OAAO;AAEL,eAAW,SAAS;AAAA,MAClB,IAAI,CAAC,CAAC,KAAK,aAAa,MAAM;AAC5B,cAAM,gBACJ,CAAC,MAAM,QAAQ,CAAC,MAAM,UAClB,cAAc,cAAc,IAC5B;AAEN,eAAO;AAAA,UACL;AAAA,UACA;AAAA,YACE,GAAG;AAAA,YACH,kBAAkB;AAAA,UAAA;AAAA,QACpB;AAAA,MAEJ,CAAC;AAAA,IAAA;AAAA,EAEL;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,eAAW;AAAA,MACT;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,IAAA;AAAA,EAEV,WAAW,MAAM,QAAQ;AAEvB,UAAM,gBAAgB,OAAO,OAAO,MAAM,MAAM,EAAE;AAAA,MAChD,CAAC,SAAS,KAAK,SAAS;AAAA,IAAA;AAE1B,QAAI,eAAe;AAEjB,iBAAW;AAAA,QACT;AAAA,QACA,CAAA;AAAA;AAAA,QACA,MAAM;AAAA,QACN,MAAM;AAAA,QACN,MAAM;AAAA,MAAA;AAAA,IAEV;AAAA,EACF;AAGA,MAAI,MAAM,WAAW,CAAC,MAAM,WAAW,MAAM,QAAQ,WAAW,IAAI;AAElE,UAAM,gBAAgB,MAAM,SACxB,OAAO,OAAO,MAAM,MAAM,EAAE,KAAK,CAAC,SAAS,KAAK,SAAS,KAAK,IAC9D;AAEJ,QAAI,CAAC,eAAe;AAClB,YAAM,IAAI,MAAM,wCAAwC;AAAA,IAC1D;AAAA,EACF;AAGA,MACE,MAAM,YACN,MAAM,SAAS,SAAS,MACvB,CAAC,MAAM,WAAW,MAAM,QAAQ,WAAW,IAC5C;AAEA,eAAW,YAAY,MAAM,UAAU;AACrC,iBAAW,SAAS;AAAA,QAClB,OAAO,CAAC,CAAC,MAAM,aAAa,MAAM;AAChC,iBAAO,SAAS,aAAa;AAAA,QAC/B,CAAC;AAAA,MAAA;AAAA,IAEL;AAAA,EACF;AAGA,MAAI,MAAM,UAAU;AAClB,eAAW,SAAS,KAAK,SAAS,CAAC,CAAC,MAAM,GAAG,MAAM,IAAI,gBAAgB,CAAC;AAAA,EAC1E;AAGA,MAAI,MAAM,WAAW,MAAM,QAAQ,SAAS,GAAG;AAC7C,UAAM,kBAAkB;AAAA,MACtB;AAAA,MACA,MAAM;AAAA,MACN,MAAM;AAAA,MACN,MAAM;AAAA,IAAA;AAIR,UAAMA,kBAAiB,gBAAgB;AAAA,MACrC,IAAI,CAAC,CAAC,KAAK,CAAC,KAAK,YAAY,CAAC,MAAM;AAElC,cAAM,eAAgB,IAAY;AAClC,eAAO,CAAC,KAAK,CAAC,cAAc,YAAY,CAAC;AAAA,MAC3C,CAAC;AAAA,IAAA;AAGH,UAAMC,UAASD;AAEf,UAAM,IAAI,UAAUC,OAAM;AAC1B,WAAOA;AAAAA,EACT,WAAW,MAAM,UAAU,UAAa,MAAM,WAAW,QAAW;AAElE,UAAM,IAAI;AAAA,MACR;AAAA,IAAA;AAAA,EAEJ;AAGA,QAAM,iBAA+B,SAAS;AAAA,IAC5C,IAAI,CAAC,CAAC,KAAK,GAAG,MAAM;AAElB,YAAM,eAAgB,IAAY;AAClC,aAAO,CAAC,KAAK,CAAC,cAAc,MAAS,CAAC;AAAA,IAIxC,CAAC;AAAA,EAAA;AAGH,QAAM,SAAS;AAEf,QAAM,IAAI,UAAU,MAAM;AAC1B,SAAO;AACT;AAKA,SAAS,YACP,MACA,WACA,OACA,cACuC;AACvC,UAAQ,KAAK,MAAA;AAAA,IACX,KAAK,iBAAiB;AACpB,YAAM,QAAQ,UAAU,KAAK,WAAW,EAAE;AAC1C,UAAI,CAAC,OAAO;AACV,cAAM,IAAI;AAAA,UACR,yBAAyB,KAAK,WAAW,EAAE;AAAA,QAAA;AAAA,MAE/C;AACA,aAAO,EAAE,OAAO,KAAK,OAAO,MAAA;AAAA,IAC9B;AAAA,IACA,KAAK,YAAY;AAEf,YAAM,gBAAgB,aAAa,IAAI,KAAK,KAAK,KAAK,KAAK;AAG3D,YAAM,gBAAgB;AAAA,QACpB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MAAA;AAKF,YAAM,iBAAiB,cAAc;AAAA,QACnC,IAAI,CAAC,SAAc;AACjB,gBAAM,CAAC,KAAK,CAAC,OAAO,aAAa,CAAC,IAAI;AACtC,iBAAO,CAAC,KAAK,KAAK;AAAA,QACpB,CAAC;AAAA,MAAA;AAGH,aAAO,EAAE,OAAO,KAAK,OAAO,OAAO,eAAA;AAAA,IACrC;AAAA,IACA;AACE,YAAM,IAAI,MAAM,0BAA2B,KAAa,IAAI,EAAE;AAAA,EAAA;AAEpE;AAOA,SAAS,iBACP,gBACA,eACA,cACM;AAEN,MACE,eAAe,KAAK,SAAS,cAC7B,cAAc,KAAK,SAAS,YAC5B;AACA,iBAAa,IAAI,eAAe,KAAK,OAAO,cAAc,KAAK,KAAK;AAEpE;AAAA,MACE,eAAe,KAAK;AAAA,MACpB,cAAc,KAAK;AAAA,MACnB;AAAA,IAAA;AAAA,EAEJ;AAGA,MAAI,eAAe,QAAQ,cAAc,MAAM;AAC7C,aACM,IAAI,GACR,IAAI,eAAe,KAAK,UAAU,IAAI,cAAc,KAAK,QACzD,KACA;AACA,YAAM,gBAAgB,eAAe,KAAK,CAAC;AAC3C,YAAM,eAAe,cAAc,KAAK,CAAC;AAEzC,UACE,cAAc,KAAK,SAAS,cAC5B,aAAa,KAAK,SAAS,YAC3B;AACA,qBAAa,IAAI,cAAc,KAAK,OAAO,aAAa,KAAK,KAAK;AAElE;AAAA,UACE,cAAc,KAAK;AAAA,UACnB,aAAa,KAAK;AAAA,UAClB;AAAA,QAAA;AAAA,MAEJ;AAAA,IACF;AAAA,EACF;AACF;"}

package/dist/esm/query/compiler/joins.d.ts

@@ -1,11 +1,7 @@
-import { JoinClause, QueryIR } from '../ir.js';
-import { KeyedStream, NamespacedAndKeyedStream, ResultStream } from '../../types.js';
-/**
- * Cache for compiled subqueries to avoid duplicate compilation
- */
-type QueryCache = WeakMap<QueryIR, ResultStream>;
+import { JoinClause } from '../ir.js';
+import { KeyedStream, NamespacedAndKeyedStream } from '../../types.js';
+import { QueryCache, QueryMapping } from './types.js';
 /**
  * Processes all join clauses in a query
  */
-export declare function processJoins(pipeline: NamespacedAndKeyedStream, joinClauses: Array<JoinClause>, tables: Record<string, KeyedStream>, mainTableAlias: string, allInputs: Record<string, KeyedStream>, cache: QueryCache): NamespacedAndKeyedStream;
-export {};
+export declare function processJoins(pipeline: NamespacedAndKeyedStream, joinClauses: Array<JoinClause>, tables: Record<string, KeyedStream>, mainTableAlias: string, allInputs: Record<string, KeyedStream>, cache: QueryCache, queryMapping: QueryMapping): NamespacedAndKeyedStream;

package/dist/esm/query/compiler/joins.js

@@ -1,7 +1,7 @@
 import { map, join, consolidate, filter } from "@electric-sql/d2mini";
 import { compileExpression } from "./evaluators.js";
 import { compileQuery } from "./index.js";
-function processJoins(pipeline, joinClauses, tables, mainTableAlias, allInputs, cache) {
+function processJoins(pipeline, joinClauses, tables, mainTableAlias, allInputs, cache, queryMapping) {
   let resultPipeline = pipeline;
   for (const joinClause of joinClauses) {
     resultPipeline = processJoin(
@@ -10,16 +10,18 @@ function processJoins(pipeline, joinClauses, tables, mainTableAlias, allInputs,
       tables,
       mainTableAlias,
       allInputs,
-      cache
+      cache,
+      queryMapping
     );
   }
   return resultPipeline;
 }
-function processJoin(pipeline, joinClause, tables, mainTableAlias, allInputs, cache) {
+function processJoin(pipeline, joinClause, tables, mainTableAlias, allInputs, cache, queryMapping) {
   const { alias: joinedTableAlias, input: joinedInput } = processJoinSource(
     joinClause.from,
     allInputs,
-    cache
+    cache,
+    queryMapping
   );
   tables[joinedTableAlias] = joinedInput;
   const joinType = joinClause.type === `cross` ? `inner` : joinClause.type === `outer` ? `full` : joinClause.type;
@@ -47,7 +49,7 @@ function processJoin(pipeline, joinClause, tables, mainTableAlias, allInputs, ca
     processJoinResults(joinClause.type)
   );
 }
-function processJoinSource(from, allInputs, cache) {
+function processJoinSource(from, allInputs, cache, queryMapping) {
   switch (from.type) {
     case `collectionRef`: {
       const input = allInputs[from.collection.id];
@@ -59,7 +61,13 @@ function processJoinSource(from, allInputs, cache) {
       return { alias: from.alias, input };
     }
     case `queryRef`: {
-      const subQueryInput = compileQuery(from.query, allInputs, cache);
+      const originalQuery = queryMapping.get(from.query) || from.query;
+      const subQueryInput = compileQuery(
+        originalQuery,
+        allInputs,
+        cache,
+        queryMapping
+      );
       const extractedInput = subQueryInput.pipe(
         map((data) => {
           const [key, [value, _orderByIndex]] = data;

package/dist/esm/query/compiler/joins.js.map

@@ -1 +1 @@
-{"version":3,"file":"joins.js","sources":["../../../../src/query/compiler/joins.ts"],"sourcesContent":["import {\n  consolidate,\n  filter,\n  join as joinOperator,\n  map,\n} from \"@electric-sql/d2mini\"\nimport { compileExpression } from \"./evaluators.js\"\nimport { compileQuery } from \"./index.js\"\nimport type { IStreamBuilder, JoinType } from \"@electric-sql/d2mini\"\nimport type { CollectionRef, JoinClause, QueryIR, QueryRef } from \"../ir.js\"\nimport type {\n  KeyedStream,\n  NamespacedAndKeyedStream,\n  NamespacedRow,\n  ResultStream,\n} from \"../../types.js\"\n\n/**\n * Cache for compiled subqueries to avoid duplicate compilation\n */\ntype QueryCache = WeakMap<QueryIR, ResultStream>\n\n/**\n * Processes all join clauses in a query\n */\nexport function processJoins(\n  pipeline: NamespacedAndKeyedStream,\n  joinClauses: Array<JoinClause>,\n  tables: Record<string, KeyedStream>,\n  mainTableAlias: string,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache\n): NamespacedAndKeyedStream {\n  let resultPipeline = pipeline\n\n  for (const joinClause of joinClauses) {\n    resultPipeline = processJoin(\n      resultPipeline,\n      joinClause,\n      tables,\n      mainTableAlias,\n      allInputs,\n      cache\n    )\n  }\n\n  return resultPipeline\n}\n\n/**\n * Processes a single join clause\n */\nfunction processJoin(\n  pipeline: NamespacedAndKeyedStream,\n  joinClause: JoinClause,\n  tables: Record<string, KeyedStream>,\n  mainTableAlias: string,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache\n): NamespacedAndKeyedStream {\n  // Get the joined table alias and input stream\n  const { alias: joinedTableAlias, input: joinedInput } = processJoinSource(\n    joinClause.from,\n    allInputs,\n    cache\n  )\n\n  // Add the joined table to the tables map\n  tables[joinedTableAlias] = joinedInput\n\n  // Convert join type to D2 join type\n  const joinType: JoinType =\n    joinClause.type === `cross`\n      ? `inner`\n      : joinClause.type === `outer`\n        ? `full`\n        : (joinClause.type as JoinType)\n\n  // Pre-compile the join expressions\n  const compiledLeftExpr = compileExpression(joinClause.left)\n  const compiledRightExpr = compileExpression(joinClause.right)\n\n  // Prepare the main pipeline for joining\n  const mainPipeline = pipeline.pipe(\n    map(([currentKey, namespacedRow]) => {\n      // Extract the join key from the left side of the join condition\n      const leftKey = compiledLeftExpr(namespacedRow)\n\n      // Return [joinKey, [originalKey, namespacedRow]]\n      return [leftKey, [currentKey, namespacedRow]] as [\n        unknown,\n        [string, typeof namespacedRow],\n      ]\n    })\n  )\n\n  // Prepare the joined pipeline\n  const joinedPipeline = joinedInput.pipe(\n    map(([currentKey, row]) => {\n      // Wrap the row in a namespaced structure\n      const namespacedRow: NamespacedRow = { [joinedTableAlias]: row }\n\n      // Extract the join key from the right side of the join condition\n      const rightKey = compiledRightExpr(namespacedRow)\n\n      // Return [joinKey, [originalKey, namespacedRow]]\n      return [rightKey, [currentKey, namespacedRow]] as [\n        unknown,\n        [string, typeof namespacedRow],\n      ]\n    })\n  )\n\n  // Apply the join operation\n  if (![`inner`, `left`, `right`, `full`].includes(joinType)) {\n    throw new Error(`Unsupported join type: ${joinClause.type}`)\n  }\n  return mainPipeline.pipe(\n    joinOperator(joinedPipeline, joinType),\n    consolidate(),\n    processJoinResults(joinClause.type)\n  )\n}\n\n/**\n * Processes the join source (collection or sub-query)\n */\nfunction processJoinSource(\n  from: CollectionRef | QueryRef,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache\n): { alias: string; input: KeyedStream } {\n  switch (from.type) {\n    case `collectionRef`: {\n      const input = allInputs[from.collection.id]\n      if (!input) {\n        throw new Error(\n          `Input for collection \"${from.collection.id}\" not found in inputs map`\n        )\n      }\n      return { alias: from.alias, input }\n    }\n    case `queryRef`: {\n      // Recursively compile the sub-query with cache\n      const subQueryInput = compileQuery(from.query, allInputs, cache)\n\n      // Subqueries may return [key, [value, orderByIndex]] (with ORDER BY) or [key, value] (without ORDER BY)\n      // We need to extract just the value for use in parent queries\n      const extractedInput = subQueryInput.pipe(\n        map((data: any) => {\n          const [key, [value, _orderByIndex]] = data\n          return [key, value] as [unknown, any]\n        })\n      )\n\n      return { alias: from.alias, input: extractedInput as KeyedStream }\n    }\n    default:\n      throw new Error(`Unsupported join source type: ${(from as any).type}`)\n  }\n}\n\n/**\n * Processes the results of a join operation\n */\nfunction processJoinResults(joinType: string) {\n  return function (\n    pipeline: IStreamBuilder<\n      [\n        key: string,\n        [\n          [string, NamespacedRow] | undefined,\n          [string, NamespacedRow] | undefined,\n        ],\n      ]\n    >\n  ): NamespacedAndKeyedStream {\n    return pipeline.pipe(\n      // Process the join result and handle nulls\n      filter((result) => {\n        const [_key, [main, joined]] = result\n        const mainNamespacedRow = main?.[1]\n        const joinedNamespacedRow = joined?.[1]\n\n        // Handle different join types\n        if (joinType === `inner`) {\n          return !!(mainNamespacedRow && joinedNamespacedRow)\n        }\n\n        if (joinType === `left`) {\n          return !!mainNamespacedRow\n        }\n\n        if (joinType === `right`) {\n          return !!joinedNamespacedRow\n        }\n\n        // For full joins, always include\n        return true\n      }),\n      map((result) => {\n        const [_key, [main, joined]] = result\n        const mainKey = main?.[0]\n        const mainNamespacedRow = main?.[1]\n        const joinedKey = joined?.[0]\n        const joinedNamespacedRow = joined?.[1]\n\n        // Merge the namespaced rows\n        const mergedNamespacedRow: NamespacedRow = {}\n\n        // Add main row data if it exists\n        if (mainNamespacedRow) {\n          Object.assign(mergedNamespacedRow, mainNamespacedRow)\n        }\n\n        // Add joined row data if it exists\n        if (joinedNamespacedRow) {\n          Object.assign(mergedNamespacedRow, joinedNamespacedRow)\n        }\n\n        // We create a composite key that combines the main and joined keys\n        const resultKey = `[${mainKey},${joinedKey}]`\n\n        return [resultKey, mergedNamespacedRow] as [string, NamespacedRow]\n      })\n    )\n  }\n}\n"],"names":["joinOperator"],"mappings":";;;AAyBO,SAAS,aACd,UACA,aACA,QACA,gBACA,WACA,OAC0B;AAC1B,MAAI,iBAAiB;AAErB,aAAW,cAAc,aAAa;AACpC,qBAAiB;AAAA,MACf;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AAEA,SAAO;AACT;AAKA,SAAS,YACP,UACA,YACA,QACA,gBACA,WACA,OAC0B;AAE1B,QAAM,EAAE,OAAO,kBAAkB,OAAO,gBAAgB;AAAA,IACtD,WAAW;AAAA,IACX;AAAA,IACA;AAAA,EAAA;AAIF,SAAO,gBAAgB,IAAI;AAG3B,QAAM,WACJ,WAAW,SAAS,UAChB,UACA,WAAW,SAAS,UAClB,SACC,WAAW;AAGpB,QAAM,mBAAmB,kBAAkB,WAAW,IAAI;AAC1D,QAAM,oBAAoB,kBAAkB,WAAW,KAAK;AAG5D,QAAM,eAAe,SAAS;AAAA,IAC5B,IAAI,CAAC,CAAC,YAAY,aAAa,MAAM;AAEnC,YAAM,UAAU,iBAAiB,aAAa;AAG9C,aAAO,CAAC,SAAS,CAAC,YAAY,aAAa,CAAC;AAAA,IAI9C,CAAC;AAAA,EAAA;AAIH,QAAM,iBAAiB,YAAY;AAAA,IACjC,IAAI,CAAC,CAAC,YAAY,GAAG,MAAM;AAEzB,YAAM,gBAA+B,EAAE,CAAC,gBAAgB,GAAG,IAAA;AAG3D,YAAM,WAAW,kBAAkB,aAAa;AAGhD,aAAO,CAAC,UAAU,CAAC,YAAY,aAAa,CAAC;AAAA,IAI/C,CAAC;AAAA,EAAA;AAIH,MAAI,CAAC,CAAC,SAAS,QAAQ,SAAS,MAAM,EAAE,SAAS,QAAQ,GAAG;AAC1D,UAAM,IAAI,MAAM,0BAA0B,WAAW,IAAI,EAAE;AAAA,EAC7D;AACA,SAAO,aAAa;AAAA,IAClBA,KAAa,gBAAgB,QAAQ;AAAA,IACrC,YAAA;AAAA,IACA,mBAAmB,WAAW,IAAI;AAAA,EAAA;AAEtC;AAKA,SAAS,kBACP,MACA,WACA,OACuC;AACvC,UAAQ,KAAK,MAAA;AAAA,IACX,KAAK,iBAAiB;AACpB,YAAM,QAAQ,UAAU,KAAK,WAAW,EAAE;AAC1C,UAAI,CAAC,OAAO;AACV,cAAM,IAAI;AAAA,UACR,yBAAyB,KAAK,WAAW,EAAE;AAAA,QAAA;AAAA,MAE/C;AACA,aAAO,EAAE,OAAO,KAAK,OAAO,MAAA;AAAA,IAC9B;AAAA,IACA,KAAK,YAAY;AAEf,YAAM,gBAAgB,aAAa,KAAK,OAAO,WAAW,KAAK;AAI/D,YAAM,iBAAiB,cAAc;AAAA,QACnC,IAAI,CAAC,SAAc;AACjB,gBAAM,CAAC,KAAK,CAAC,OAAO,aAAa,CAAC,IAAI;AACtC,iBAAO,CAAC,KAAK,KAAK;AAAA,QACpB,CAAC;AAAA,MAAA;AAGH,aAAO,EAAE,OAAO,KAAK,OAAO,OAAO,eAAA;AAAA,IACrC;AAAA,IACA;AACE,YAAM,IAAI,MAAM,iCAAkC,KAAa,IAAI,EAAE;AAAA,EAAA;AAE3E;AAKA,SAAS,mBAAmB,UAAkB;AAC5C,SAAO,SACL,UAS0B;AAC1B,WAAO,SAAS;AAAA;AAAA,MAEd,OAAO,CAAC,WAAW;AACjB,cAAM,CAAC,MAAM,CAAC,MAAM,MAAM,CAAC,IAAI;AAC/B,cAAM,oBAAoB,6BAAO;AACjC,cAAM,sBAAsB,iCAAS;AAGrC,YAAI,aAAa,SAAS;AACxB,iBAAO,CAAC,EAAE,qBAAqB;AAAA,QACjC;AAEA,YAAI,aAAa,QAAQ;AACvB,iBAAO,CAAC,CAAC;AAAA,QACX;AAEA,YAAI,aAAa,SAAS;AACxB,iBAAO,CAAC,CAAC;AAAA,QACX;AAGA,eAAO;AAAA,MACT,CAAC;AAAA,MACD,IAAI,CAAC,WAAW;AACd,cAAM,CAAC,MAAM,CAAC,MAAM,MAAM,CAAC,IAAI;AAC/B,cAAM,UAAU,6BAAO;AACvB,cAAM,oBAAoB,6BAAO;AACjC,cAAM,YAAY,iCAAS;AAC3B,cAAM,sBAAsB,iCAAS;AAGrC,cAAM,sBAAqC,CAAA;AAG3C,YAAI,mBAAmB;AACrB,iBAAO,OAAO,qBAAqB,iBAAiB;AAAA,QACtD;AAGA,YAAI,qBAAqB;AACvB,iBAAO,OAAO,qBAAqB,mBAAmB;AAAA,QACxD;AAGA,cAAM,YAAY,IAAI,OAAO,IAAI,SAAS;AAE1C,eAAO,CAAC,WAAW,mBAAmB;AAAA,MACxC,CAAC;AAAA,IAAA;AAAA,EAEL;AACF;"}
+{"version":3,"file":"joins.js","sources":["../../../../src/query/compiler/joins.ts"],"sourcesContent":["import {\n  consolidate,\n  filter,\n  join as joinOperator,\n  map,\n} from \"@electric-sql/d2mini\"\nimport { compileExpression } from \"./evaluators.js\"\nimport { compileQuery } from \"./index.js\"\nimport type { IStreamBuilder, JoinType } from \"@electric-sql/d2mini\"\nimport type { CollectionRef, JoinClause, QueryRef } from \"../ir.js\"\nimport type {\n  KeyedStream,\n  NamespacedAndKeyedStream,\n  NamespacedRow,\n} from \"../../types.js\"\nimport type { QueryCache, QueryMapping } from \"./types.js\"\n\n/**\n * Processes all join clauses in a query\n */\nexport function processJoins(\n  pipeline: NamespacedAndKeyedStream,\n  joinClauses: Array<JoinClause>,\n  tables: Record<string, KeyedStream>,\n  mainTableAlias: string,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache,\n  queryMapping: QueryMapping\n): NamespacedAndKeyedStream {\n  let resultPipeline = pipeline\n\n  for (const joinClause of joinClauses) {\n    resultPipeline = processJoin(\n      resultPipeline,\n      joinClause,\n      tables,\n      mainTableAlias,\n      allInputs,\n      cache,\n      queryMapping\n    )\n  }\n\n  return resultPipeline\n}\n\n/**\n * Processes a single join clause\n */\nfunction processJoin(\n  pipeline: NamespacedAndKeyedStream,\n  joinClause: JoinClause,\n  tables: Record<string, KeyedStream>,\n  mainTableAlias: string,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache,\n  queryMapping: QueryMapping\n): NamespacedAndKeyedStream {\n  // Get the joined table alias and input stream\n  const { alias: joinedTableAlias, input: joinedInput } = processJoinSource(\n    joinClause.from,\n    allInputs,\n    cache,\n    queryMapping\n  )\n\n  // Add the joined table to the tables map\n  tables[joinedTableAlias] = joinedInput\n\n  // Convert join type to D2 join type\n  const joinType: JoinType =\n    joinClause.type === `cross`\n      ? `inner`\n      : joinClause.type === `outer`\n        ? `full`\n        : (joinClause.type as JoinType)\n\n  // Pre-compile the join expressions\n  const compiledLeftExpr = compileExpression(joinClause.left)\n  const compiledRightExpr = compileExpression(joinClause.right)\n\n  // Prepare the main pipeline for joining\n  const mainPipeline = pipeline.pipe(\n    map(([currentKey, namespacedRow]) => {\n      // Extract the join key from the left side of the join condition\n      const leftKey = compiledLeftExpr(namespacedRow)\n\n      // Return [joinKey, [originalKey, namespacedRow]]\n      return [leftKey, [currentKey, namespacedRow]] as [\n        unknown,\n        [string, typeof namespacedRow],\n      ]\n    })\n  )\n\n  // Prepare the joined pipeline\n  const joinedPipeline = joinedInput.pipe(\n    map(([currentKey, row]) => {\n      // Wrap the row in a namespaced structure\n      const namespacedRow: NamespacedRow = { [joinedTableAlias]: row }\n\n      // Extract the join key from the right side of the join condition\n      const rightKey = compiledRightExpr(namespacedRow)\n\n      // Return [joinKey, [originalKey, namespacedRow]]\n      return [rightKey, [currentKey, namespacedRow]] as [\n        unknown,\n        [string, typeof namespacedRow],\n      ]\n    })\n  )\n\n  // Apply the join operation\n  if (![`inner`, `left`, `right`, `full`].includes(joinType)) {\n    throw new Error(`Unsupported join type: ${joinClause.type}`)\n  }\n  return mainPipeline.pipe(\n    joinOperator(joinedPipeline, joinType),\n    consolidate(),\n    processJoinResults(joinClause.type)\n  )\n}\n\n/**\n * Processes the join source (collection or sub-query)\n */\nfunction processJoinSource(\n  from: CollectionRef | QueryRef,\n  allInputs: Record<string, KeyedStream>,\n  cache: QueryCache,\n  queryMapping: QueryMapping\n): { alias: string; input: KeyedStream } {\n  switch (from.type) {\n    case `collectionRef`: {\n      const input = allInputs[from.collection.id]\n      if (!input) {\n        throw new Error(\n          `Input for collection \"${from.collection.id}\" not found in inputs map`\n        )\n      }\n      return { alias: from.alias, input }\n    }\n    case `queryRef`: {\n      // Find the original query for caching purposes\n      const originalQuery = queryMapping.get(from.query) || from.query\n\n      // Recursively compile the sub-query with cache\n      const subQueryInput = compileQuery(\n        originalQuery,\n        allInputs,\n        cache,\n        queryMapping\n      )\n\n      // Subqueries may return [key, [value, orderByIndex]] (with ORDER BY) or [key, value] (without ORDER BY)\n      // We need to extract just the value for use in parent queries\n      const extractedInput = subQueryInput.pipe(\n        map((data: any) => {\n          const [key, [value, _orderByIndex]] = data\n          return [key, value] as [unknown, any]\n        })\n      )\n\n      return { alias: from.alias, input: extractedInput as KeyedStream }\n    }\n    default:\n      throw new Error(`Unsupported join source type: ${(from as any).type}`)\n  }\n}\n\n/**\n * Processes the results of a join operation\n */\nfunction processJoinResults(joinType: string) {\n  return function (\n    pipeline: IStreamBuilder<\n      [\n        key: string,\n        [\n          [string, NamespacedRow] | undefined,\n          [string, NamespacedRow] | undefined,\n        ],\n      ]\n    >\n  ): NamespacedAndKeyedStream {\n    return pipeline.pipe(\n      // Process the join result and handle nulls\n      filter((result) => {\n        const [_key, [main, joined]] = result\n        const mainNamespacedRow = main?.[1]\n        const joinedNamespacedRow = joined?.[1]\n\n        // Handle different join types\n        if (joinType === `inner`) {\n          return !!(mainNamespacedRow && joinedNamespacedRow)\n        }\n\n        if (joinType === `left`) {\n          return !!mainNamespacedRow\n        }\n\n        if (joinType === `right`) {\n          return !!joinedNamespacedRow\n        }\n\n        // For full joins, always include\n        return true\n      }),\n      map((result) => {\n        const [_key, [main, joined]] = result\n        const mainKey = main?.[0]\n        const mainNamespacedRow = main?.[1]\n        const joinedKey = joined?.[0]\n        const joinedNamespacedRow = joined?.[1]\n\n        // Merge the namespaced rows\n        const mergedNamespacedRow: NamespacedRow = {}\n\n        // Add main row data if it exists\n        if (mainNamespacedRow) {\n          Object.assign(mergedNamespacedRow, mainNamespacedRow)\n        }\n\n        // Add joined row data if it exists\n        if (joinedNamespacedRow) {\n          Object.assign(mergedNamespacedRow, joinedNamespacedRow)\n        }\n\n        // We create a composite key that combines the main and joined keys\n        const resultKey = `[${mainKey},${joinedKey}]`\n\n        return [resultKey, mergedNamespacedRow] as [string, NamespacedRow]\n      })\n    )\n  }\n}\n"],"names":["joinOperator"],"mappings":";;;AAoBO,SAAS,aACd,UACA,aACA,QACA,gBACA,WACA,OACA,cAC0B;AAC1B,MAAI,iBAAiB;AAErB,aAAW,cAAc,aAAa;AACpC,qBAAiB;AAAA,MACf;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IAAA;AAAA,EAEJ;AAEA,SAAO;AACT;AAKA,SAAS,YACP,UACA,YACA,QACA,gBACA,WACA,OACA,cAC0B;AAE1B,QAAM,EAAE,OAAO,kBAAkB,OAAO,gBAAgB;AAAA,IACtD,WAAW;AAAA,IACX;AAAA,IACA;AAAA,IACA;AAAA,EAAA;AAIF,SAAO,gBAAgB,IAAI;AAG3B,QAAM,WACJ,WAAW,SAAS,UAChB,UACA,WAAW,SAAS,UAClB,SACC,WAAW;AAGpB,QAAM,mBAAmB,kBAAkB,WAAW,IAAI;AAC1D,QAAM,oBAAoB,kBAAkB,WAAW,KAAK;AAG5D,QAAM,eAAe,SAAS;AAAA,IAC5B,IAAI,CAAC,CAAC,YAAY,aAAa,MAAM;AAEnC,YAAM,UAAU,iBAAiB,aAAa;AAG9C,aAAO,CAAC,SAAS,CAAC,YAAY,aAAa,CAAC;AAAA,IAI9C,CAAC;AAAA,EAAA;AAIH,QAAM,iBAAiB,YAAY;AAAA,IACjC,IAAI,CAAC,CAAC,YAAY,GAAG,MAAM;AAEzB,YAAM,gBAA+B,EAAE,CAAC,gBAAgB,GAAG,IAAA;AAG3D,YAAM,WAAW,kBAAkB,aAAa;AAGhD,aAAO,CAAC,UAAU,CAAC,YAAY,aAAa,CAAC;AAAA,IAI/C,CAAC;AAAA,EAAA;AAIH,MAAI,CAAC,CAAC,SAAS,QAAQ,SAAS,MAAM,EAAE,SAAS,QAAQ,GAAG;AAC1D,UAAM,IAAI,MAAM,0BAA0B,WAAW,IAAI,EAAE;AAAA,EAC7D;AACA,SAAO,aAAa;AAAA,IAClBA,KAAa,gBAAgB,QAAQ;AAAA,IACrC,YAAA;AAAA,IACA,mBAAmB,WAAW,IAAI;AAAA,EAAA;AAEtC;AAKA,SAAS,kBACP,MACA,WACA,OACA,cACuC;AACvC,UAAQ,KAAK,MAAA;AAAA,IACX,KAAK,iBAAiB;AACpB,YAAM,QAAQ,UAAU,KAAK,WAAW,EAAE;AAC1C,UAAI,CAAC,OAAO;AACV,cAAM,IAAI;AAAA,UACR,yBAAyB,KAAK,WAAW,EAAE;AAAA,QAAA;AAAA,MAE/C;AACA,aAAO,EAAE,OAAO,KAAK,OAAO,MAAA;AAAA,IAC9B;AAAA,IACA,KAAK,YAAY;AAEf,YAAM,gBAAgB,aAAa,IAAI,KAAK,KAAK,KAAK,KAAK;AAG3D,YAAM,gBAAgB;AAAA,QACpB;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MAAA;AAKF,YAAM,iBAAiB,cAAc;AAAA,QACnC,IAAI,CAAC,SAAc;AACjB,gBAAM,CAAC,KAAK,CAAC,OAAO,aAAa,CAAC,IAAI;AACtC,iBAAO,CAAC,KAAK,KAAK;AAAA,QACpB,CAAC;AAAA,MAAA;AAGH,aAAO,EAAE,OAAO,KAAK,OAAO,OAAO,eAAA;AAAA,IACrC;AAAA,IACA;AACE,YAAM,IAAI,MAAM,iCAAkC,KAAa,IAAI,EAAE;AAAA,EAAA;AAE3E;AAKA,SAAS,mBAAmB,UAAkB;AAC5C,SAAO,SACL,UAS0B;AAC1B,WAAO,SAAS;AAAA;AAAA,MAEd,OAAO,CAAC,WAAW;AACjB,cAAM,CAAC,MAAM,CAAC,MAAM,MAAM,CAAC,IAAI;AAC/B,cAAM,oBAAoB,6BAAO;AACjC,cAAM,sBAAsB,iCAAS;AAGrC,YAAI,aAAa,SAAS;AACxB,iBAAO,CAAC,EAAE,qBAAqB;AAAA,QACjC;AAEA,YAAI,aAAa,QAAQ;AACvB,iBAAO,CAAC,CAAC;AAAA,QACX;AAEA,YAAI,aAAa,SAAS;AACxB,iBAAO,CAAC,CAAC;AAAA,QACX;AAGA,eAAO;AAAA,MACT,CAAC;AAAA,MACD,IAAI,CAAC,WAAW;AACd,cAAM,CAAC,MAAM,CAAC,MAAM,MAAM,CAAC,IAAI;AAC/B,cAAM,UAAU,6BAAO;AACvB,cAAM,oBAAoB,6BAAO;AACjC,cAAM,YAAY,iCAAS;AAC3B,cAAM,sBAAsB,iCAAS;AAGrC,cAAM,sBAAqC,CAAA;AAG3C,YAAI,mBAAmB;AACrB,iBAAO,OAAO,qBAAqB,iBAAiB;AAAA,QACtD;AAGA,YAAI,qBAAqB;AACvB,iBAAO,OAAO,qBAAqB,mBAAmB;AAAA,QACxD;AAGA,cAAM,YAAY,IAAI,OAAO,IAAI,SAAS;AAE1C,eAAO,CAAC,WAAW,mBAAmB;AAAA,MACxC,CAAC;AAAA,IAAA;AAAA,EAEL;AACF;"}

package/dist/esm/query/compiler/types.d.ts

@@ -0,0 +1,10 @@
+import { QueryIR } from '../ir.js';
+import { ResultStream } from '../../types.js';
+/**
+ * Cache for compiled subqueries to avoid duplicate compilation
+ */
+export type QueryCache = WeakMap<QueryIR, ResultStream>;
+/**
+ * Mapping from optimized queries back to their original queries for caching
+ */
+export type QueryMapping = WeakMap<QueryIR, QueryIR>;

package/dist/esm/query/optimizer.d.ts

@@ -0,0 +1,42 @@
+import { BasicExpression, QueryIR } from './ir.js';
+/**
+ * Represents a WHERE clause after source analysis
+ */
+export interface AnalyzedWhereClause {
+    /** The WHERE expression */
+    expression: BasicExpression<boolean>;
+    /** Set of table/source aliases that this WHERE clause touches */
+    touchedSources: Set<string>;
+}
+/**
+ * Represents WHERE clauses grouped by the sources they touch
+ */
+export interface GroupedWhereClauses {
+    /** WHERE clauses that touch only a single source, grouped by source alias */
+    singleSource: Map<string, BasicExpression<boolean>>;
+    /** WHERE clauses that touch multiple sources, combined into one expression */
+    multiSource?: BasicExpression<boolean>;
+}
+/**
+ * Main query optimizer entry point that lifts WHERE clauses into subqueries.
+ *
+ * This function implements multi-level predicate pushdown optimization by recursively
+ * moving WHERE clauses through nested subqueries to get them as close to the data
+ * sources as possible, then removing redundant subqueries.
+ *
+ * @param query - The QueryIR to optimize
+ * @returns A new QueryIR with optimizations applied (or original if no optimization possible)
+ *
+ * @example
+ * ```typescript
+ * const originalQuery = {
+ *   from: new CollectionRef(users, 'u'),
+ *   join: [{ from: new CollectionRef(posts, 'p'), ... }],
+ *   where: [eq(u.dept_id, 1), gt(p.views, 100)]
+ * }
+ *
+ * const optimized = optimizeQuery(originalQuery)
+ * // Result: Single-source clauses moved to deepest possible subqueries
+ * ```
+ */
+export declare function optimizeQuery(query: QueryIR): QueryIR;

package/dist/esm/query/optimizer.js

@@ -0,0 +1,283 @@
+import { deepEquals } from "../utils.js";
+import { QueryRef, CollectionRef, Func } from "./ir.js";
+function optimizeQuery(query) {
+  let optimized = query;
+  let previousOptimized;
+  let iterations = 0;
+  const maxIterations = 10;
+  while (iterations < maxIterations && !deepEquals(optimized, previousOptimized)) {
+    previousOptimized = optimized;
+    optimized = applyRecursiveOptimization(optimized);
+    iterations++;
+  }
+  const cleaned = removeRedundantSubqueries(optimized);
+  return cleaned;
+}
+function applyRecursiveOptimization(query) {
+  var _a;
+  const subqueriesOptimized = {
+    ...query,
+    from: query.from.type === `queryRef` ? new QueryRef(
+      applyRecursiveOptimization(query.from.query),
+      query.from.alias
+    ) : query.from,
+    join: (_a = query.join) == null ? void 0 : _a.map((joinClause) => ({
+      ...joinClause,
+      from: joinClause.from.type === `queryRef` ? new QueryRef(
+        applyRecursiveOptimization(joinClause.from.query),
+        joinClause.from.alias
+      ) : joinClause.from
+    }))
+  };
+  return applySingleLevelOptimization(subqueriesOptimized);
+}
+function applySingleLevelOptimization(query) {
+  if (!query.where || query.where.length === 0) {
+    return query;
+  }
+  if (!query.join || query.join.length === 0) {
+    return query;
+  }
+  const splitWhereClauses = splitAndClauses(query.where);
+  const analyzedClauses = splitWhereClauses.map(
+    (clause) => analyzeWhereClause(clause)
+  );
+  const groupedClauses = groupWhereClauses(analyzedClauses);
+  return applyOptimizations(query, groupedClauses);
+}
+function removeRedundantSubqueries(query) {
+  var _a;
+  return {
+    ...query,
+    from: removeRedundantFromClause(query.from),
+    join: (_a = query.join) == null ? void 0 : _a.map((joinClause) => ({
+      ...joinClause,
+      from: removeRedundantFromClause(joinClause.from)
+    }))
+  };
+}
+function removeRedundantFromClause(from) {
+  if (from.type === `collectionRef`) {
+    return from;
+  }
+  const processedQuery = removeRedundantSubqueries(from.query);
+  if (isRedundantSubquery(processedQuery)) {
+    const innerFrom = removeRedundantFromClause(processedQuery.from);
+    if (innerFrom.type === `collectionRef`) {
+      return new CollectionRef(innerFrom.collection, from.alias);
+    } else {
+      return new QueryRef(innerFrom.query, from.alias);
+    }
+  }
+  return new QueryRef(processedQuery, from.alias);
+}
+function isRedundantSubquery(query) {
+  return (!query.where || query.where.length === 0) && !query.select && (!query.groupBy || query.groupBy.length === 0) && (!query.having || query.having.length === 0) && (!query.orderBy || query.orderBy.length === 0) && (!query.join || query.join.length === 0) && query.limit === void 0 && query.offset === void 0 && !query.fnSelect && (!query.fnWhere || query.fnWhere.length === 0) && (!query.fnHaving || query.fnHaving.length === 0);
+}
+function splitAndClauses(whereClauses) {
+  const result = [];
+  for (const clause of whereClauses) {
+    if (clause.type === `func` && clause.name === `and`) {
+      const splitArgs = splitAndClauses(
+        clause.args
+      );
+      result.push(...splitArgs);
+    } else {
+      result.push(clause);
+    }
+  }
+  return result;
+}
+function analyzeWhereClause(clause) {
+  const touchedSources = /* @__PURE__ */ new Set();
+  function collectSources(expr) {
+    switch (expr.type) {
+      case `ref`:
+        if (expr.path && expr.path.length > 0) {
+          const firstElement = expr.path[0];
+          if (firstElement) {
+            touchedSources.add(firstElement);
+          }
+        }
+        break;
+      case `func`:
+        if (expr.args) {
+          expr.args.forEach(collectSources);
+        }
+        break;
+      case `val`:
+        break;
+      case `agg`:
+        if (expr.args) {
+          expr.args.forEach(collectSources);
+        }
+        break;
+    }
+  }
+  collectSources(clause);
+  return {
+    expression: clause,
+    touchedSources
+  };
+}
+function groupWhereClauses(analyzedClauses) {
+  const singleSource = /* @__PURE__ */ new Map();
+  const multiSource = [];
+  for (const clause of analyzedClauses) {
+    if (clause.touchedSources.size === 1) {
+      const source = Array.from(clause.touchedSources)[0];
+      if (!singleSource.has(source)) {
+        singleSource.set(source, []);
+      }
+      singleSource.get(source).push(clause.expression);
+    } else if (clause.touchedSources.size > 1) {
+      multiSource.push(clause.expression);
+    }
+  }
+  const combinedSingleSource = /* @__PURE__ */ new Map();
+  for (const [source, clauses] of singleSource) {
+    combinedSingleSource.set(source, combineWithAnd(clauses));
+  }
+  const combinedMultiSource = multiSource.length > 0 ? combineWithAnd(multiSource) : void 0;
+  return {
+    singleSource: combinedSingleSource,
+    multiSource: combinedMultiSource
+  };
+}
+function applyOptimizations(query, groupedClauses) {
+  const actuallyOptimized = /* @__PURE__ */ new Set();
+  const optimizedFrom = optimizeFromWithTracking(
+    query.from,
+    groupedClauses.singleSource,
+    actuallyOptimized
+  );
+  const optimizedJoins = query.join ? query.join.map((joinClause) => ({
+    ...joinClause,
+    from: optimizeFromWithTracking(
+      joinClause.from,
+      groupedClauses.singleSource,
+      actuallyOptimized
+    )
+  })) : void 0;
+  const remainingWhereClauses = [];
+  if (groupedClauses.multiSource) {
+    remainingWhereClauses.push(groupedClauses.multiSource);
+  }
+  for (const [source, clause] of groupedClauses.singleSource) {
+    if (!actuallyOptimized.has(source)) {
+      remainingWhereClauses.push(clause);
+    }
+  }
+  const optimizedQuery = {
+    // Copy all non-optimized fields as-is
+    select: query.select,
+    groupBy: query.groupBy ? [...query.groupBy] : void 0,
+    having: query.having ? [...query.having] : void 0,
+    orderBy: query.orderBy ? [...query.orderBy] : void 0,
+    limit: query.limit,
+    offset: query.offset,
+    fnSelect: query.fnSelect,
+    fnWhere: query.fnWhere ? [...query.fnWhere] : void 0,
+    fnHaving: query.fnHaving ? [...query.fnHaving] : void 0,
+    // Use the optimized FROM and JOIN clauses
+    from: optimizedFrom,
+    join: optimizedJoins,
+    // Only include WHERE clauses that weren't successfully optimized
+    where: remainingWhereClauses.length > 0 ? remainingWhereClauses : []
+  };
+  return optimizedQuery;
+}
+function deepCopyQuery(query) {
+  return {
+    // Recursively copy the FROM clause
+    from: query.from.type === `collectionRef` ? new CollectionRef(query.from.collection, query.from.alias) : new QueryRef(deepCopyQuery(query.from.query), query.from.alias),
+    // Copy all other fields, creating new arrays where necessary
+    select: query.select,
+    join: query.join ? query.join.map((joinClause) => ({
+      type: joinClause.type,
+      left: joinClause.left,
+      right: joinClause.right,
+      from: joinClause.from.type === `collectionRef` ? new CollectionRef(
+        joinClause.from.collection,
+        joinClause.from.alias
+      ) : new QueryRef(
+        deepCopyQuery(joinClause.from.query),
+        joinClause.from.alias
+      )
+    })) : void 0,
+    where: query.where ? [...query.where] : void 0,
+    groupBy: query.groupBy ? [...query.groupBy] : void 0,
+    having: query.having ? [...query.having] : void 0,
+    orderBy: query.orderBy ? [...query.orderBy] : void 0,
+    limit: query.limit,
+    offset: query.offset,
+    fnSelect: query.fnSelect,
+    fnWhere: query.fnWhere ? [...query.fnWhere] : void 0,
+    fnHaving: query.fnHaving ? [...query.fnHaving] : void 0
+  };
+}
+function optimizeFromWithTracking(from, singleSourceClauses, actuallyOptimized) {
+  const whereClause = singleSourceClauses.get(from.alias);
+  if (!whereClause) {
+    if (from.type === `collectionRef`) {
+      return new CollectionRef(from.collection, from.alias);
+    }
+    return new QueryRef(deepCopyQuery(from.query), from.alias);
+  }
+  if (from.type === `collectionRef`) {
+    const subQuery = {
+      from: new CollectionRef(from.collection, from.alias),
+      where: [whereClause]
+    };
+    actuallyOptimized.add(from.alias);
+    return new QueryRef(subQuery, from.alias);
+  }
+  if (!isSafeToPushIntoExistingSubquery(from.query)) {
+    return new QueryRef(deepCopyQuery(from.query), from.alias);
+  }
+  const existingWhere = from.query.where || [];
+  const optimizedSubQuery = {
+    ...deepCopyQuery(from.query),
+    where: [...existingWhere, whereClause]
+  };
+  actuallyOptimized.add(from.alias);
+  return new QueryRef(optimizedSubQuery, from.alias);
+}
+function isSafeToPushIntoExistingSubquery(query) {
+  if (query.select) {
+    const hasAggregates = Object.values(query.select).some(
+      (expr) => expr.type === `agg`
+    );
+    if (hasAggregates) {
+      return false;
+    }
+  }
+  if (query.groupBy && query.groupBy.length > 0) {
+    return false;
+  }
+  if (query.having && query.having.length > 0) {
+    return false;
+  }
+  if (query.orderBy && query.orderBy.length > 0) {
+    if (query.limit !== void 0 || query.offset !== void 0) {
+      return false;
+    }
+  }
+  if (query.fnSelect || query.fnWhere && query.fnWhere.length > 0 || query.fnHaving && query.fnHaving.length > 0) {
+    return false;
+  }
+  return true;
+}
+function combineWithAnd(expressions) {
+  if (expressions.length === 0) {
+    throw new Error(`Cannot combine empty expression list`);
+  }
+  if (expressions.length === 1) {
+    return expressions[0];
+  }
+  return new Func(`and`, expressions);
+}
+export {
+  optimizeQuery
+};
+//# sourceMappingURL=optimizer.js.map