@finos/legend-data-cube 0.0.43 → 0.0.44

package/src/stores/core/DataCubeQuerySnapshotBuilder.ts

@@ -25,44 +25,81 @@
 import {
   V1_AppliedFunction,
   V1_CInteger,
-  V1_Collection,
   extractElementNameFromPath as _name,
   matchFunctionName,
   type V1_ValueSpecification,
   V1_Lambda,
 } from '@finos/legend-graph';
 import type { DataCubeQuery } from './model/DataCubeQuery.js';
-import { DataCubeQuerySnapshot } from './DataCubeQuerySnapshot.js';
-import { _toCol, type DataCubeColumn } from './model/DataCubeColumn.js';
-import { assertTrue, at, guaranteeNonNullable } from '@finos/legend-shared';
 import {
-  DataCubeQuerySortDirection,
+  DataCubeQuerySnapshot,
+  type DataCubeQuerySnapshotAggregateColumn,
+  type DataCubeQuerySnapshotProcessingContext,
+  type DataCubeQuerySnapshotSortColumn,
+} from './DataCubeQuerySnapshot.js';
+import {
+  _findCol,
+  _toCol,
+  type DataCubeColumn,
+} from './model/DataCubeColumn.js';
+import {
+  assertTrue,
+  at,
+  guaranteeNonNullable,
+  isNonNullable,
+} from '@finos/legend-shared';
+import {
   DataCubeFunction,
   type DataCubeQueryFunctionMap,
 } from './DataCubeQueryEngine.js';
-import { buildDefaultConfiguration } from './DataCubeConfigurationBuilder.js';
+import { newConfiguration } from './DataCubeConfigurationBuilder.js';
 import {
   _colSpecArrayParam,
-  _colSpecParam,
-  _funcMatch,
   _param,
   _extractExtendedColumns,
   _filter,
   _relationType,
   _genericTypeParam,
   _packageableType,
+  _aggCol,
+  _sort,
+  _unwrapLambda,
+  _pivotSort,
+  _groupBySort,
+  _validatePivot,
+  _checkDuplicateColumns,
+  _validateGroupBy,
 } from './DataCubeQuerySnapshotBuilderUtils.js';
 import type { DataCubeSource } from './model/DataCubeSource.js';
-import type { DataCubeQueryFilterOperation } from './filter/DataCubeQueryFilterOperation.js';
-import type { DataCubeQueryAggregateOperation } from './aggregation/DataCubeQueryAggregateOperation.js';
+import type { DataCubeEngine } from './DataCubeEngine.js';
+import {
+  _deserializeValueSpecification,
+  _serializeValueSpecification,
+} from './DataCubeQueryBuilderUtils.js';
 
 // --------------------------------- BUILDING BLOCKS ---------------------------------
 
 const _SUPPORTED_TOP_LEVEL_FUNCTIONS: {
   func: string;
+  /**
+   * If there are multiple signature to a function, such as extend(), this indicates
+   * the minimum number of parameters. And a stopping condition needs to be provided.
+   *
+   * Note that this is a naive mechanism to process simple functions, by no means, it's
+   * meant to mimic generic function matcher.
+   */
   parameters: number;
+  stopCondition?: ((_func: V1_AppliedFunction) => boolean) | undefined;
 }[] = [
-  { func: DataCubeFunction.EXTEND, parameters: 2 }, // TODO: support both signatures of extend()
+  {
+    func: DataCubeFunction.EXTEND,
+    parameters: 1,
+    // handle OLAP form where first parameter is over() expression used to construct the window
+    stopCondition: (_func) =>
+      matchFunctionName(_func.function, DataCubeFunction.EXTEND) &&
+      _func.parameters[0] instanceof V1_AppliedFunction &&
+      matchFunctionName(_func.parameters[0].function, DataCubeFunction.OVER),
+  },
   { func: DataCubeFunction.FILTER, parameters: 1 },
   { func: DataCubeFunction.SELECT, parameters: 1 },
   { func: DataCubeFunction.GROUP_BY, parameters: 2 },
@@ -102,14 +139,6 @@ enum _FUNCTION_SEQUENCE_COMPOSITION_PART {
   LIMIT = 'limit',
 }
 
-// function isFunctionInValidSequence(
-//   value: string,
-// ): value is _FUNCTION_SEQUENCE_COMPOSITION_PART {
-//   return Object.values(_FUNCTION_SEQUENCE_COMPOSITION_PART).includes(
-//     value as _FUNCTION_SEQUENCE_COMPOSITION_PART,
-//   );
-// }
-
 // This corresponds to the function sequence that we currently support:
 //
 // ->extend()*
@@ -199,10 +228,10 @@ const _FUNCTION_SEQUENCE_COMPOSITION_PATTERN_REGEXP = new RegExp(
         ? `(${node.funcs
             .map(
               (childNode) =>
-                `(?<${childNode.name}><${_name(childNode.func)}>____\\d+)${childNode.repeat ? '*' : !childNode.required ? '?' : ''}`,
+                `(?<${childNode.name}>${childNode.repeat ? `(?:<${_name(childNode.func)}>____\\d+)*` : `<${_name(childNode.func)}>____\\d+`})${childNode.repeat ? '' : !childNode.required ? '?' : ''}`,
             )
             .join('')})${node.repeat ? '*' : !node.required ? '?' : ''}`
-        : `(?<${node.name}><${_name(node.func)}>____\\d+)${node.repeat ? '*' : !node.required ? '?' : ''}`,
+        : `(?<${node.name}>${node.repeat ? `(?:<${_name(node.func)}>____\\d+)*` : `<${_name(node.func)}>____\\d+`})${node.repeat ? '' : !node.required ? '?' : ''}`,
     )
     .join('')}$`,
 );
@@ -216,7 +245,7 @@ function extractFunctionMap(
 ): DataCubeQueryFunctionMap {
   // Make sure this is a sequence of function calls
   if (!(query instanceof V1_AppliedFunction)) {
-    throw new Error(`Can't process expression: Expected a function expression`);
+    throw new Error(`Can't process expression: expected a function expression`);
   }
   const sequence: V1_AppliedFunction[] = [];
   let currentFunc = query;
@@ -229,7 +258,7 @@ function extractFunctionMap(
     // Check that all functions in sequence are supported (matching name and number of parameters)
     if (!supportedFunc) {
       throw new Error(
-        `Can't process expression: Found unsupported function ${currentFunc.function}()`,
+        `Can't process expression: found unsupported function ${currentFunc.function}()`,
       );
     }
 
@@ -238,16 +267,22 @@ function extractFunctionMap(
     // t(...)->z()->y()->x() and simultaneously, remove the first parameter from each function for
     // simplicity, except for the innermost function
     if (currentFunc.parameters.length > supportedFunc.parameters) {
+      // if stop condition is fulfilled, no more function sequence drilling is needed
+      if (supportedFunc.stopCondition?.(currentFunc)) {
+        sequence.unshift(currentFunc);
+        break;
+      }
+
       // assert that the supported function has the expected number of parameters
       assertTrue(
         currentFunc.parameters.length === supportedFunc.parameters + 1,
-        `Can't process ${_name(currentFunc.function)}() expression: Expected at most ${supportedFunc.parameters + 1} parameters provided, got ${currentFunc.parameters.length}`,
+        `Can't process ${_name(currentFunc.function)}() expression: expected at most ${supportedFunc.parameters + 1} parameters provided, got ${currentFunc.parameters.length}`,
       );
       const func = _param(
         currentFunc,
         0,
         V1_AppliedFunction,
-        `Can't process expression: Expected a sequence of function calls (e.g. x()->y()->z())`,
+        `Can't process expression: expected a sequence of function calls (e.g. x()->y()->z())`,
       );
       currentFunc.parameters = currentFunc.parameters.slice(1);
       sequence.unshift(currentFunc);
@@ -267,35 +302,46 @@ function extractFunctionMap(
   );
   if (!matchResult) {
     throw new Error(
-      `Can't process expression: Unsupported function composition ${sequence.map((fn) => `${_name(fn.function)}()`).join('->')} (supported composition: ${_FUNCTION_SEQUENCE_COMPOSITION_PATTERN.map((node) => `${'funcs' in node ? `[${node.funcs.map((childNode) => `${_name(childNode.func)}()`).join('->')}]` : `${_name(node.func)}()`}`).join('->')})`,
+      `Can't process expression: unsupported function composition ${sequence.map((fn) => `${_name(fn.function)}()`).join('->')} (supported composition: ${_FUNCTION_SEQUENCE_COMPOSITION_PATTERN.map((node) => `${'funcs' in node ? `[${node.funcs.map((childNode) => `${_name(childNode.func)}()`).join('->')}]` : `${_name(node.func)}()`}`).join('->')})`,
     );
   }
 
-  const _process = (key: string): V1_AppliedFunction | undefined => {
+  const _process = (key: string): V1_AppliedFunction[] | undefined => {
     const match = matchResult.groups?.[key];
-    if (!match || match.indexOf('____') === -1) {
+    if (!match) {
       return undefined;
     }
-    const idx = Number(match.split('____')[1]);
-    if (isNaN(idx) || idx >= sequence.length) {
+    const funcMatches = match.match(/\<.*?\>____\d+/g);
+    if (!funcMatches?.length) {
       return undefined;
     }
-    const func = at(sequence, idx);
-    return func;
+
+    return funcMatches
+      .map((funcMatch) => {
+        const idx = Number(funcMatch.split('____')[1]);
+        if (isNaN(idx) || idx >= sequence.length) {
+          return undefined;
+        }
+        const func = at(sequence, idx);
+        return func;
+      })
+      .filter(isNonNullable);
   };
 
   return {
     leafExtend: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.LEAF_EXTEND),
-    select: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.SELECT),
-    filter: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.FILTER),
-    pivotSort: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT_SORT),
-    pivot: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT),
-    pivotCast: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT_CAST),
-    groupBy: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.GROUP_BY),
-    groupBySort: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.GROUP_BY_SORT),
+    select: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.SELECT)?.[0],
+    filter: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.FILTER)?.[0],
+    pivotSort: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT_SORT)?.[0],
+    pivot: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT)?.[0],
+    pivotCast: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.PIVOT_CAST)?.[0],
+    groupBy: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.GROUP_BY)?.[0],
+    groupBySort: _process(
+      _FUNCTION_SEQUENCE_COMPOSITION_PART.GROUP_BY_SORT,
+    )?.[0],
     groupExtend: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.GROUP_EXTEND),
-    sort: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.SORT),
-    limit: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.LIMIT),
+    sort: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.SORT)?.[0],
+    limit: _process(_FUNCTION_SEQUENCE_COMPOSITION_PART.LIMIT)?.[0],
   };
 }
 
@@ -307,31 +353,34 @@ function extractFunctionMap(
  * Implementation-wise, this extracts the function call sequence, then walk the
  * sequence in order to fill in the information for the snapshot.
  */
-export function validateAndBuildQuerySnapshot(
+export async function validateAndBuildQuerySnapshot(
   partialQuery: V1_ValueSpecification,
   source: DataCubeSource,
   baseQuery: DataCubeQuery,
-  filterOperations: DataCubeQueryFilterOperation[],
-  aggregateOperations: DataCubeQueryAggregateOperation[],
+  engine: DataCubeEngine,
 ) {
   // --------------------------------- BASE ---------------------------------
   // Build the function call sequence and the function map to make the
   // analysis more ergonomic
 
-  const funcMap = extractFunctionMap(partialQuery);
+  // Clone the query since we will mutate it during the process
+  const query = _deserializeValueSpecification(
+    _serializeValueSpecification(partialQuery),
+  );
+  const funcMap = extractFunctionMap(query);
   const snapshot = DataCubeQuerySnapshot.create({});
   const data = snapshot.data;
-  const columnNames = new Set<string>();
+  const registeredColumns = new Map<string, DataCubeColumn>();
   /**
    * We want to make sure all columns, either from source or created, e.g. extended columns,
    * have unique names. This is to simplify the logic within DataCube so different components
    * can easily refer to columns by name without having to worry about conflicts.
    */
   const _checkColName = (col: DataCubeColumn, message: string) => {
-    if (columnNames.has(col.name)) {
+    if (registeredColumns.has(col.name)) {
       throw new Error(message);
     }
-    columnNames.add(col.name);
+    registeredColumns.set(col.name, col);
   };
   const colsMap = new Map<string, DataCubeColumn>();
   const _getCol = (colName: string) => {
@@ -342,53 +391,66 @@ export function validateAndBuildQuerySnapshot(
     return _toCol(column);
   };
   const _setCol = (col: DataCubeColumn) => colsMap.set(col.name, col);
-  // const _unsetCol = (col: DataCubeColumn) => colsMap.delete(col.name);
 
   // -------------------------------- SOURCE --------------------------------
 
   data.sourceColumns = source.columns;
-  data.sourceColumns.forEach((col) => _setCol(col));
+
+  // validate
+  _checkDuplicateColumns(
+    data.sourceColumns,
+    (colName) =>
+      `Can't process source: found duplicate source columns '${colName}'`,
+  );
   data.sourceColumns.forEach((col) =>
     _checkColName(
       col,
-      `Can't process source column '${col.name}': another column with the same name is already registered`,
+      `Can't process source: another column with name '${col.name}' is already registered`,
     ),
   );
 
+  data.sourceColumns.forEach((col) => _setCol(col));
+
   // --------------------------- LEAF-LEVEL EXTEND ---------------------------
 
-  if (funcMap.leafExtend) {
-    // TODO: get column types (call engine to get the type)
-    data.leafExtendedColumns = _extractExtendedColumns(funcMap.leafExtend);
-    // TODO: populate the column types
-    data.leafExtendedColumns.forEach((col) => _setCol(col));
+  if (funcMap.leafExtend?.length) {
+    data.leafExtendedColumns = await _extractExtendedColumns(
+      funcMap.leafExtend,
+      Array.from(colsMap.values()),
+      engine,
+    );
+
+    // validate
+    // NOTE: these duplication checks might not be necessary since compiler would catch these
+    // issues anyway, but we leave them here to be defensive
+    _checkDuplicateColumns(
+      data.leafExtendedColumns,
+      (colName) =>
+        `Can't process extend() expression: found duplicate extended columns '${colName}'`,
+    );
     data.leafExtendedColumns.forEach((col) =>
       _checkColName(
         col,
-        `Can't process leaf-level extended column '${col.name}': another column with the same name is already registered`,
+        `Can't process extend() expression: another column with name '${col.name}' is already registered`,
       ),
     );
+
+    data.leafExtendedColumns.forEach((col) => _setCol(col));
   }
 
   // --------------------------------- FILTER ---------------------------------
 
   if (funcMap.filter) {
-    // TODO: verify column presence
-    // TODO: verify column types
     const lambda = _param(
       funcMap.filter,
       0,
       V1_Lambda,
-      `Can't process filter() expression: Expected parameter at index 0 to be a lambda expression`,
-    );
-    assertTrue(
-      lambda.body.length === 1,
-      `Can't process filter() expression: Expected lambda body to have exactly 1 expression`,
+      `Can't process filter() expression: expected parameter at index 0 to be a lambda expression`,
     );
     data.filter = _filter(
-      guaranteeNonNullable(lambda.body[0]),
+      _unwrapLambda(lambda, `Can't process filter() expression`),
       _getCol,
-      filterOperations,
+      engine.filterOperations,
     );
   }
 
@@ -398,119 +460,156 @@ export function validateAndBuildQuerySnapshot(
     data.selectColumns = _colSpecArrayParam(funcMap.select, 0).colSpecs.map(
       (colSpec) => _getCol(colSpec.name),
     );
-    // TODO: remove columns that are not selected from colsMap
+
+    // validate
+    _checkDuplicateColumns(
+      data.selectColumns,
+      (colName) =>
+        `Can't process select() expression: found duplicate select columns '${colName}'`,
+    );
+
+    // restrict the set of available columns to only selected columns
+    colsMap.clear();
+    data.selectColumns.forEach((col) => _setCol(col));
+  } else {
+    // mandate that if select() expression is not present, we consider this
+    // as no-column is selected
+    colsMap.clear();
   }
 
   // --------------------------------- PIVOT ---------------------------------
 
-  if (funcMap.pivot && funcMap.pivotCast) {
-    // TODO: verify column presence
-    // TODO: verify column types
+  let pivotAggColumns: DataCubeQuerySnapshotAggregateColumn[] = [];
+  let pivotSortColumns: DataCubeQuerySnapshotSortColumn[] = [];
+  if (funcMap.pivot && funcMap.pivotCast && funcMap.pivotSort) {
+    const pivotColumns = _colSpecArrayParam(funcMap.pivot, 0).colSpecs.map(
+      (colSpec) => _getCol(colSpec.name),
+    );
+    const castColumns = _relationType(
+      _genericTypeParam(funcMap.pivotCast, 0).genericType,
+    ).columns.map((column) => ({
+      name: column.name,
+      type: _packageableType(column.genericType).fullPath,
+    }));
     data.pivot = {
-      columns: _colSpecArrayParam(funcMap.pivot, 0).colSpecs.map((colSpec) =>
-        _getCol(colSpec.name),
-      ),
-      castColumns: _relationType(
-        _genericTypeParam(funcMap.pivotCast, 0).genericType,
-      ).columns.map((column) => ({
-        name: column.name,
-        type: _packageableType(column.genericType).fullPath,
-      })),
+      columns: pivotColumns,
+      castColumns: castColumns,
     };
+
+    // process aggregate columns
+    pivotAggColumns = _colSpecArrayParam(funcMap.pivot, 1).colSpecs.map(
+      (colSpec) => _aggCol(colSpec, _getCol, engine.aggregateOperations),
+    );
+    // process sort columns
+    pivotSortColumns = _pivotSort(funcMap.pivotSort, pivotColumns, _getCol);
+
+    // validate
+    _validatePivot(data.pivot, pivotAggColumns, Array.from(colsMap.values()));
+
+    // restrict the set of available columns to only casted columns
+    colsMap.clear();
+    castColumns.forEach((col) => _setCol(col));
   }
-  // TODO: verify groupBy agg columns, pivot agg columns and configuration agree
 
   // --------------------------------- GROUP BY ---------------------------------
 
-  if (funcMap.groupBy) {
-    // TODO: verify column presence
-    // TODO: verify column types
+  let groupByAggColumns: DataCubeQuerySnapshotAggregateColumn[] = [];
+  let groupBySortColumns: DataCubeQuerySnapshotSortColumn[] = [];
+  if (funcMap.groupBy && funcMap.groupBySort) {
+    const groupByColumns = _colSpecArrayParam(funcMap.groupBy, 0).colSpecs.map(
+      (colSpec) => _getCol(colSpec.name),
+    );
     data.groupBy = {
-      columns: _colSpecArrayParam(funcMap.groupBy, 0).colSpecs.map((colSpec) =>
-        _getCol(colSpec.name),
-      ),
+      columns: groupByColumns,
     };
-    // TODO: verify groupBy agg columns, pivot agg columns and configuration agree
-    // TODO: verify sort columns
-    // TODO: use configuration information present in the baseQuery configuration?
+
+    // process aggregate columns
+    groupByAggColumns = _colSpecArrayParam(funcMap.groupBy, 1).colSpecs.map(
+      (colSpec) => _aggCol(colSpec, _getCol, engine.aggregateOperations),
+    );
+    // process sort columns
+    groupBySortColumns = _groupBySort(
+      funcMap.groupBySort,
+      groupByColumns,
+      _getCol,
+    );
+
+    // validate
+    _validateGroupBy(
+      data.groupBy,
+      groupByAggColumns,
+      data.pivot,
+      pivotAggColumns,
+      Array.from(colsMap.values()),
+    );
   }
 
   // --------------------------- GROUP-LEVEL EXTEND ---------------------------
 
-  if (funcMap.groupExtend) {
-    // TODO: get column types (call engine to get the type)
-    data.groupExtendedColumns = _extractExtendedColumns(funcMap.groupExtend);
-    // TODO: populate the column types
-    data.groupExtendedColumns.forEach((col) => _setCol(col));
+  if (funcMap.groupExtend?.length) {
+    data.groupExtendedColumns = await _extractExtendedColumns(
+      funcMap.groupExtend,
+      Array.from(colsMap.values()),
+      engine,
+    );
+
+    // validate
+    // NOTE: these duplication checks might not be necessary since compiler would catch these
+    // issues anyway, but we leave them here to be defensive
+    _checkDuplicateColumns(
+      data.groupExtendedColumns,
+      (colName) =>
+        `Can't process extend() expression: found duplicate extended columns '${colName}'`,
+    );
     data.groupExtendedColumns.forEach((col) =>
       _checkColName(
         col,
-        `Can't process group-level extended column '${col.name}': another column with the same name is already registered`,
+        `Can't process extend() expression: another column with name '${col.name}' is already registered`,
       ),
     );
+
+    data.groupExtendedColumns.forEach((col) => _setCol(col));
   }
 
   // --------------------------------- SORT ---------------------------------
 
   if (funcMap.sort) {
-    data.sortColumns = _param(funcMap.sort, 0, V1_Collection).values.map(
-      (value) => {
-        const sortColFunc = _funcMatch(value, [
-          DataCubeFunction.ASCENDING,
-          DataCubeFunction.DESCENDING,
-        ]);
-        return {
-          ..._getCol(_colSpecParam(sortColFunc, 0).name),
-          direction: matchFunctionName(
-            sortColFunc.function,
-            DataCubeFunction.ASCENDING,
-          )
-            ? DataCubeQuerySortDirection.ASCENDING
-            : DataCubeQuerySortDirection.DESCENDING,
-        };
-      },
+    data.sortColumns = _sort(funcMap.sort, _getCol);
+
+    // validate
+    _checkDuplicateColumns(
+      data.sortColumns,
+      (colName) =>
+        `Can't process sort() expression: found duplicate sort columns '${colName}'`,
     );
   }
 
   // --------------------------------- LIMIT ---------------------------------
 
   if (funcMap.limit) {
+    // NOTE: negative number -10 is parsed as minus(10) so this check will also
+    // reject negative number
     const value = _param(
       funcMap.limit,
       0,
       V1_CInteger,
-      `Can't process limit() expression: Expected parameter at index 0 to be an integer instance value`,
+      `Can't process limit() expression: expected limit to be a non-negative integer value`,
     );
     data.limit = value.value;
   }
 
   // --------------------------------- CONFIGURATION ---------------------------------
-  //
-  // A data-cube conceptually consists of a data query, in form of a Pure query, instead
-  // of a particular specification object format, and this configuration, that holds mostly
-  // layout and styling customization. But there are overlaps, i.e. certain "meta" query
-  // configuration are stored in this configuration, e.g. column aggregation type, because
-  // a column aggregation's preference needs to be specified even when there's no aggregation
-  // specified over that column in the data query.
-  //
-  // But in the example above, if the column is part of an aggregation, we have to ensure
-  // the configuration is consistent with the query. Conflicts can happen, for example:
-  // - column kind and type conflict with aggregation
-  // - column kind and type conflict with the column configuration
-  //
-  // In those cases, we need to reconcile to make sure the query and the configuration to agree.
-  // The query will take precedence when conflicts happen, and if the conflict cannot be resolved
-  // somehow, we will throw a validation error. We decide so because in certain cases, configuration
-  // needs to be generated from default presets, such as for use cases where the query comes from a
-  // different source, such as Studio or Query, or another part of Engine, where the layout
-  // configuration is not specified.
-  //
-  // ----------------------------------------------------------------------------------
 
   const configuration = validateAndBuildConfiguration(
-    snapshot,
-    funcMap,
+    {
+      snapshot,
+      groupByAggColumns,
+      groupBySortColumns,
+      pivotAggColumns,
+      pivotSortColumns,
+    },
     baseQuery,
+    engine,
   );
   data.configuration = configuration.serialize();
 
@@ -518,32 +617,155 @@ export function validateAndBuildQuerySnapshot(
 }
 
 /**
- * TODO: @datacube roundtrip - implement the logic to reconcile the configuration with the query
- * - [ ] columns (missing/extra columns - remove or generate default column configuration)
- * - [ ] base off the type and kind, check the settings to see if it's compatible
- * - [ ] verify groupBy agg columns, pivot agg columns and configuration agree
- * - [ ] verify groupBy sort columns and tree column sort direction configuration agree
+ * Builds and/or validates the configuration.
+ *
+ * TL;DR;
+ * If not provided, generate a default configuration based off the metadata extracted
+ * when processing the query in previous steps.
+ * If provided, check if the configuration aggree with the query processing metadata.
+ *
+ * CONTEXT:
+ * A data-cube conceptually consists of a data query, in form of a Pure query, instead
+ * of a particular specification object format, and this configuration, that holds mostly
+ * layout and styling customization. But there are overlaps, i.e. certain _meta_ query
+ * configuration are stored in this configuration, e.g. column aggregation type, because
+ * a column aggregation's preference needs to be specified even when there's no aggregation
+ * specified over that column in the data query.
+ *
+ * But in the example above, if the column is part of an aggregation, we have to ensure
+ * the configuration is consistent with the query. Conflicts can happen, for example:
+ * - column kind and type conflict with aggregation
+ * - column kind and type conflict with the column configuration
+ *
+ * In those cases, we need to make sure the query and the configuration to agree.
+ * If a config is provided, we will need to validate that config. If none is provided,
+ * we will generate a config from the query processing metadata, in which case, no
+ * validation is needed. The latter case comes up quite often where the query comes from a
+ * different source, such as Studio or Query, or another part of Engine, and the layout
+ * configuration is not specified.
  */
 function validateAndBuildConfiguration(
-  snapshot: DataCubeQuerySnapshot,
-  funcMap: DataCubeQueryFunctionMap,
+  context: DataCubeQuerySnapshotProcessingContext,
   baseQuery: DataCubeQuery,
+  engine: DataCubeEngine,
 ) {
-  const data = snapshot.data;
-  const configuration =
-    baseQuery.configuration?.clone() ??
-    buildDefaultConfiguration([
-      ...data.sourceColumns,
-      ...data.leafExtendedColumns,
-      ...data.groupExtendedColumns,
-    ]);
-
-  // column selection
-  configuration.columns.forEach((column) => {
-    column.isSelected = Boolean(
-      data.groupExtendedColumns.find((col) => col.name === column.name) ??
-        data.selectColumns.find((col) => col.name === column.name),
+  const data = context.snapshot.data;
+  const config = baseQuery.configuration;
+  // generate a default configuration anyway to be used to compare with the
+  // provided configuration for validation purpose
+  const _config = newConfiguration(context);
+
+  if (!config) {
+    return _config;
+  }
+
+  // check tree column sort direction (only relevant if groupBy is present)
+  if (data.groupBy) {
+    assertTrue(
+      config.treeColumnSortDirection === _config.treeColumnSortDirection,
+      `Can't process configuration: tree column sort direction mismatch (expected: '${_config.treeColumnSortDirection.toLowerCase()}', found: '${config.treeColumnSortDirection.toLowerCase()}')`,
+    );
+  }
+
+  // check columns
+  const columns = config.columns;
+  const _columns = _config.columns;
+  const columnNames = new Set<string>();
+
+  // check for duplicate columns
+  columns.forEach((col) => {
+    if (columnNames.has(col.name)) {
+      throw new Error(
+        `Can't process configuration: found duplicate columns '${col.name}'`,
+      );
+    } else {
+      columnNames.add(col.name);
+    }
+  });
+
+  // check for extra columns
+  columns.forEach((col) => {
+    if (!_findCol(_columns, col.name)) {
+      throw new Error(
+        `Can't process configuration: found extra column '${col.name}'`,
+      );
+    }
+  });
+
+  // check for missing columns
+  _columns.forEach((col) => {
+    if (!_findCol(columns, col.name)) {
+      throw new Error(
+        `Can't process configuration: missing column '${col.name}'`,
+      );
+    }
+  });
+
+  // check for columns ordering
+  const columnsOrdering = [
+    ...data.selectColumns,
+    ...data.groupExtendedColumns,
+    ...[...data.sourceColumns, ...data.leafExtendedColumns].filter(
+      (col) => !_findCol(data.selectColumns, col.name),
+    ),
+  ];
+  columnsOrdering.forEach((_col, idx) => {
+    const col = at(columns, idx);
+    assertTrue(
+      _col.name === col.name,
+      `Can't process configuration: column ordering mismatch at index ${idx} (expected: '${_col.name}', found: '${col.name})', expected ordering: ${columnsOrdering.map((c) => c.name).join(', ')}`,
+    );
+  });
+
+  columns.forEach((column) => {
+    const _column = guaranteeNonNullable(_findCol(_columns, column.name));
+
+    // check type
+    assertTrue(
+      column.type === _column.type,
+      `Can't process configuration: type mismatch for column '${column.name}' (expected: '${_column.type}', found: '${column.type}')`,
+    );
+
+    // check selection
+    assertTrue(
+      column.isSelected === _column.isSelected,
+      `Can't process configuration: selection mismatch for column '${column.name}' (expected: '${_column.isSelected}', found: '${column.isSelected}')`,
     );
+
+    // check kind (only relevant if aggregation is present)
+    if (data.pivot ?? data.groupBy) {
+      assertTrue(
+        column.kind === _column.kind,
+        `Can't process configuration: kind mismatch for column '${column.name}' (expected: '${_column.kind.toLowerCase()}', found: '${column.kind.toLowerCase()}')`,
+      );
+    }
+
+    // check aggregation (only relevant if aggregation is present)
+    if (data.pivot ?? data.groupBy) {
+      assertTrue(
+        column.aggregateOperator === _column.aggregateOperator,
+        `Can't process configuration: aggregation operator mismatch for column '${column.name}' (expected: '${_column.aggregateOperator}', found: '${column.aggregateOperator}')`,
+      );
+      assertTrue(
+        engine
+          .getAggregateOperation(column.aggregateOperator)
+          .isCompatibleWithParameterValues(column.aggregationParameters),
+        `Can't process configuration: incompatible aggregation parameter values for column '${column.name}' (operator: '${column.aggregateOperator}')`,
+      );
+    }
+
+    // check pivot sort direction and exclusion (only relevant if pivot is present)
+    if (data.pivot) {
+      assertTrue(
+        column.excludedFromPivot === _column.excludedFromPivot,
+        `Can't process configuration: pivot exclusion mismatch for column '${column.name}' (expected: '${_column.excludedFromPivot}', found: '${column.excludedFromPivot}')`,
+      );
+      assertTrue(
+        column.pivotSortDirection === _column.pivotSortDirection,
+        `Can't process configuration: pivot sort direction mismatch for column '${column.name}' (expected: '${_column.pivotSortDirection?.toLowerCase() ?? 'none'}', found: '${column.pivotSortDirection?.toLowerCase() ?? 'none'}')`,
+      );
+    }
   });
-  return configuration;
+
+  return config;
 }