Include support to limit rows in a group by operation

h2/src/main/org/h2/api/ErrorCode.java

@@ -2240,7 +2240,15 @@ public class ErrorCode {
      */
     public static final int GROUP_BY_NOT_IN_THE_RESULT = 90157;
 
-    // next is 90158
+    /**
+     * The error with code <code>90158</code> is thrown when the in memory
+     * table containing all the entries of a group by statement would exceed
+     * the maximum capacity specified using the system property 
+     * <code>h2.maxGroupByEntries</code>.
+     */
+    public static final int GROUP_BY_TABLE_TOO_LARGE = 90158;
+
+    // next is 90159
 
     private ErrorCode() {
         // utility class

h2/src/main/org/h2/command/query/SelectGroups.java

@@ -13,10 +13,13 @@
 import java.util.Map.Entry;
 import java.util.TreeMap;
 
+import org.h2.api.ErrorCode;
 import org.h2.engine.SessionLocal;
+import org.h2.engine.SysProperties;
 import org.h2.expression.Expression;
 import org.h2.expression.analysis.DataAnalysisOperation;
 import org.h2.expression.analysis.PartitionData;
+import org.h2.message.DbException;
 import org.h2.value.Value;
 import org.h2.value.ValueRow;
 
@@ -46,13 +49,25 @@ public abstract class SelectGroups {
 
     private static final class Grouped extends SelectGroups {
 
+        /**
+         * Memory usage for each treemap entry, in addition to the value,
+         * consisting of 5 pointers + object overhead
+         */
+        private static final int TREEMAP_ENTRY_OVERHEAD = 56;
+
         private final int[] groupIndex;
 
         /**
          * Map of group-by key to group-by expression data e.g. AggregateData
          */
         private TreeMap<ValueRow, Object[]> groupByData;
 
+        /**
+         * Approximation of the memory used by groupByData, based on the size of the
+         * keys that it contains
+         */
+        private int groupByDataSize;
+
         /**
          * Key into groupByData that produces currentGroupByExprData. Not used
          * in lazy mode.
@@ -93,8 +108,18 @@ public void nextSource() {
             }
             Object[] values = groupByData.get(currentGroupsKey);
             if (values == null) {
+                if (SysProperties.maxGroupByMemoryUsage>0 && groupByDataSize>=SysProperties.maxGroupByMemoryUsage) {
+                    throw DbException.get(ErrorCode.GROUP_BY_TABLE_TOO_LARGE);
+                }
                 values = createRow();
                 groupByData.put(currentGroupsKey, values);
+
+                // TODO: how to track the memory of values?
+                // Most values seem to take a minimum of 24 bytes, Boolean and Null take 0 bytes,
+                // Some objects such as strings and arrays take more. However, for most cases, the
+                // agregation values will be numeric, so maybe values.length * 32 + 32 is a 
+                // good enough approximation?
+                groupByDataSize += currentGroupsKey.getMemory() + TREEMAP_ENTRY_OVERHEAD;
             }
             currentGroupByExprData = values;
             currentGroupRowId++;

h2/src/main/org/h2/engine/SysProperties.java

@@ -390,6 +390,22 @@ public class SysProperties {
     public static final String AUTH_CONFIG_FILE =
             Utils.getProperty("h2.authConfigFile", null);
 
+    /**
+     * System property <code>h2.maxGroupByMemoryUsage </code>
+     * (default: -1).<br />
+     * maxGroupByMemoryUsage defines a suggested upper limit to the memory
+     * used in certain group by operations. If this number is exceeded
+     * the query will fail with a memory management error.
+     * The default value of -1 means that the use of memory is not
+     * limited.
+     * 
+     * This setting is experimental and the memory usage calculation is
+     * approximate. In the future it will be removed or replaced when a
+     * better approach for handling group bys is introduced.
+     */
+    public static int maxGroupByMemoryUsage =
+            Utils.getProperty("h2.maxGroupByMemoryUsage", -1);
+
     private static final String H2_BASE_DIR = "h2.baseDir";
 
     private SysProperties() {

h2/src/main/org/h2/message/DbException.java

@@ -551,6 +551,7 @@ public static SQLException getJdbcSQLException(String message, String sql, Strin
         case ACCESS_DENIED_TO_CLASS_1:
         case RESULT_SET_READONLY:
         case CURRENT_SEQUENCE_VALUE_IS_NOT_DEFINED_IN_SESSION_1:
+        case GROUP_BY_TABLE_TOO_LARGE:
             return new JdbcSQLNonTransientException(message, sql, state, errorCode, cause, stackTrace);
         case FEATURE_NOT_SUPPORTED_1:
             return new JdbcSQLFeatureNotSupportedException(message, sql, state, errorCode, cause, stackTrace);

h2/src/main/org/h2/res/_messages_cs.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Obecná chyba: {0}
 HY004=Neznámý datový typ: {0}
 HYC00=Vlastnost není podporována: {0}

h2/src/main/org/h2/res/_messages_de.prop

@@ -195,6 +195,7 @@
 90155=Erzeugte Spalte {0} kann nicht durch eine referentielle Integrität mit dem Ausdruck {1} veränderbar sein
 90156=Spalten-Alias ist nicht für den Audruck {0} angegeben
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Allgemeiner Fehler: {0}
 HY004=Unbekannter Datentyp: {0}
 HYC00=Dieses Feature wird nicht unterstützt: {0}

h2/src/main/org/h2/res/_messages_en.prop

@@ -195,6 +195,7 @@
 90155=Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=Column alias is not specified for expression {0}
 90157=Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=Group by table has too many entries.
 HY000=General error: {0}
 HY004=Unknown data type: {0}
 HYC00=Feature not supported: {0}

h2/src/main/org/h2/res/_messages_es.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Error General : {0}
 HY004=Tipo de dato desconocido : {0}
 HYC00=Caracteristica no soportada: {0}

h2/src/main/org/h2/res/_messages_fr.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Erreur générale: {0}
 HY004=Type de données inconnu: {0}
 HYC00=Fonctionnalité non supportée: {0}

h2/src/main/org/h2/res/_messages_ja.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=一般エラー: {0}
 HY004=不明なデータ型: {0}
 HYC00=機能はサポートされていません: {0}

h2/src/main/org/h2/res/_messages_pl.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Błąd ogólny: {0}
 HY004=Nieznany typ danych: {0}
 HYC00=Cecha nie jest wspierana: {0}

h2/src/main/org/h2/res/_messages_pt_br.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Erro geral: {0}
 HY004=Tipo de dados desconhecido: {0}
 HYC00=Recurso não suportado: {0}

h2/src/main/org/h2/res/_messages_ru.prop

@@ -196,6 +196,7 @@
 90155=Генерируемый столбец {0} не может обновляться ссылочным ограничением с пунктом {1}
 90156=Имя столбца не указано для выражения {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Внутренняя ошибка: {0}
 HY004=Неизвестный тип данных: {0}
 HYC00=Данная функция не поддерживается: {0}

h2/src/main/org/h2/res/_messages_sk.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=Všeobecná chyba: {0}
 HY004=Neznámy dátový typ: {0}
 HYC00=Vlastnosť nie je podporovaná: {0}

h2/src/main/org/h2/res/_messages_zh_cn.prop

@@ -195,6 +195,7 @@
 90155=#Generated column {0} cannot be updatable by a referential constraint with {1} clause
 90156=#Column alias is not specified for expression {0}
 90157=#Column index {0} in GROUP BY clause is outside valid range 1 - {1}
+90158=#Group by table has too many entries.
 HY000=常规错误: {0}
 HY004=位置数据类型: {0}
 HYC00=不支持的特性: {0}

h2/src/test/org/h2/test/scripts/TestScript.java

@@ -34,6 +34,7 @@
 import org.h2.command.Prepared;
 import org.h2.command.dml.ScriptCommand;
 import org.h2.command.query.Query;
+import org.h2.engine.SysProperties;
 import org.h2.engine.Mode.ModeEnum;
 import org.h2.jdbc.JdbcConnection;
 import org.h2.jdbc.JdbcPreparedStatement;
@@ -234,6 +235,11 @@ public void test() throws Exception {
                 "table", "values", "window" }) {
             testScript("queries/" + s + ".sql");
         }
+
+        SysProperties.maxGroupByMemoryUsage = 200;
+        testScript("other/large_group_by.sql");
+        SysProperties.maxGroupByMemoryUsage = -1;
+
         if (config.mvStore) {
             testScript("other/two_phase_commit.sql");
             testScript("other/unique_include.sql");

h2/src/test/org/h2/test/scripts/other/large_group_by.sql

@@ -0,0 +1,32 @@
+-- Copyright 2021-2021 H2 Group. Multiple-Licensed under the MPL 2.0,
+-- and the EPL 1.0 (https://h2database.com/html/license.html).
+-- Initial Developer: H2 Group
+--
+
+-- Test the limits for temporary group by in memory structures
+
+CREATE TABLE TEST (A INT, B VARCHAR) AS (VALUES (10, 'STRING 1'), (11, 'STRING 2'));
+> ok
+
+SELECT A, count(*) from TEST group by A;
+> A  COUNT(*)
+> -- --------
+> 10 1
+> 11 1
+> rows: 2
+
+SELECT B, count(*) from TEST WHERE A<11 group by B;
+> B        COUNT(*)
+> -------- --------
+> STRING 1 1
+> rows: 1
+
+SELECT B, count(*) from TEST group by B;
+> exception GROUP_BY_TABLE_TOO_LARGE
+
+SELECT A, max(B) from TEST group by A;
+> A  MAX(B)
+> -- --------
+> 10 STRING 1
+> 11 STRING 2
+> rows: 2