diff --git a/CHANGELOG.md b/CHANGELOG.md
index f22ba1edfd3b..01271438bc73 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -137,6 +137,7 @@
 - [Implemented `Table.order_by` for the in-memory table.][3515]
 - [Renamed `File_Format.Text` to `Plain_Text`, updated `File_Format.Delimited`
   API and added builders for customizing less common settings.][3516]
+- [Allow control of sort direction in `First` and `Last` aggregations.][3517]
 
 [debug-shortcuts]:
   https://github.com/enso-org/enso/blob/develop/app/gui/docs/product/shortcuts.md#debug
@@ -215,6 +216,7 @@
 [3514]: https://github.com/enso-org/enso/pull/3514
 [3515]: https://github.com/enso-org/enso/pull/3515
 [3516]: https://github.com/enso-org/enso/pull/3516
+[3517]: https://github.com/enso-org/enso/pull/3517
 
 #### Enso Compiler
 
diff --git a/distribution/lib/Standard/Base/0.0.0-dev/src/Data/Ordering/Sort_Direction.enso b/distribution/lib/Standard/Base/0.0.0-dev/src/Data/Ordering/Sort_Direction.enso
index 9618ffa47862..eb5172680bea 100644
--- a/distribution/lib/Standard/Base/0.0.0-dev/src/Data/Ordering/Sort_Direction.enso
+++ b/distribution/lib/Standard/Base/0.0.0-dev/src/Data/Ordering/Sort_Direction.enso
@@ -19,3 +19,8 @@ type Sort_Direction
              Sort_Direction.Descending
     type Descending
 
+    ## Convert into the sign of the direction
+    to_sign : Integer
+    to_sign = case this of
+        Ascending -> 1
+        Descending -> -1
diff --git a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Dialect/Postgres.enso b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Dialect/Postgres.enso
index 0401d6258a6b..ac60a463c379 100644
--- a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Dialect/Postgres.enso
+++ b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Dialect/Postgres.enso
@@ -151,14 +151,12 @@ first_last_aggregators =
 make_first_aggregator reverse ignore_null args =
     if args.length < 2 then Error.throw (Illegal_State_Error "Insufficient number of arguments for the operation.") else
         result_expr = args.head
-        order_exprs = args.tail
+        order_bys = args.tail
 
         filter_clause = if ignore_null.not then Sql.code "" else
             Sql.code " FILTER (WHERE " ++ result_expr.paren ++ Sql.code " IS NOT NULL)"
-        modified_order_exprs =
-            order_exprs.map expr-> expr ++ Sql.code " ASC NULLS FIRST"
         order_clause =
-            Sql.code " ORDER BY " ++ Sql.join "," modified_order_exprs
+            Sql.code " ORDER BY " ++ Sql.join "," order_bys
         index_expr = case reverse of
             True -> if ignore_null.not then Sql.code "COUNT(*)" else
                 Sql.code "COUNT(" ++ result_expr ++ Sql.code ")"
diff --git a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Aggregate_Helper.enso b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Aggregate_Helper.enso
index df63005f2e71..3e1f2b809e82 100644
--- a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Aggregate_Helper.enso
+++ b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Aggregate_Helper.enso
@@ -1,4 +1,5 @@
 from Standard.Base import all
+from Standard.Base.Data.Text.Text_Ordering as Text_Ordering_Module import Text_Ordering
 
 from Standard.Table.Data.Aggregate_Column import all
 import Standard.Database.Data.Internal.IR
@@ -14,15 +15,15 @@ from Standard.Database.Error as Database_Errors import Unsupported_Database_Oper
 make_aggregate_column : Table -> Aggregate_Column -> Text -> IR.Internal_Column
 make_aggregate_column table aggregate new_name =
     sql_type = table.connection.dialect.resolve_target_sql_type aggregate
-    expression = here.make_expression aggregate
+    expression = here.make_expression aggregate table.connection.dialect
     IR.Internal_Column new_name sql_type expression
 
 ## PRIVATE
    Creates an Internal Representation of the expression that computes a
    requested statistic.
-make_expression : Aggregate_Column -> IR.Expression
-make_expression aggregate =
-    is_non_empty_vector v = if v.is_nothing then False else v.not_empty
+make_expression : Aggregate_Column -> Dialect -> IR.Expression
+make_expression aggregate dialect =
+    is_non_empty_selector v = if v.is_nothing then False else v.columns.not_empty
     case aggregate of
         Group_By c _ -> c.expression
         Count _ -> IR.Operation "COUNT_ROWS" []
@@ -36,20 +37,20 @@ make_expression aggregate =
         Count_Empty c _ -> IR.Operation "COUNT_EMPTY" [c.expression]
         Percentile p c _ -> IR.Operation "PERCENTILE" [IR.Constant Sql_Type.double p, c.expression]
         Mode c _ -> IR.Operation "MODE" [c.expression]
-        First c _ ignore_nothing order_by -> case is_non_empty_vector order_by of
+        First c _ ignore_nothing order_by -> case is_non_empty_selector order_by of
             False -> Error.throw (Unsupported_Database_Operation_Error "`First` aggregation requires at least one `order_by` column.")
             True ->
-                order_exprs = order_by.map .expression
+                order_bys = order_by.columns.map c-> dialect.prepare_order_descriptor c.column.as_internal c.direction Text_Ordering
                 case ignore_nothing of
-                    False -> IR.Operation "FIRST" [c.expression]+order_exprs
-                    True -> IR.Operation "FIRST_NOT_NULL" [c.expression]+order_exprs
-        Last c _ ignore_nothing order_by -> case is_non_empty_vector order_by of
+                    False -> IR.Operation "FIRST" [c.expression]+order_bys
+                    True -> IR.Operation "FIRST_NOT_NULL" [c.expression]+order_bys
+        Last c _ ignore_nothing order_by -> case is_non_empty_selector order_by of
             False -> Error.throw (Unsupported_Database_Operation_Error "`Last` aggregation requires at least one `order_by` column.")
             True ->
-                order_exprs = order_by.map .expression
+                order_bys = order_by.columns.map c-> dialect.prepare_order_descriptor c.column.as_internal c.direction Text_Ordering
                 case ignore_nothing of
-                    False -> IR.Operation "LAST" [c.expression]+order_exprs
-                    True -> IR.Operation "LAST_NOT_NULL" [c.expression]+order_exprs
+                    False -> IR.Operation "LAST" [c.expression]+order_bys
+                    True -> IR.Operation "LAST_NOT_NULL" [c.expression]+order_bys
         Maximum c _ -> IR.Operation "MAX" [c.expression]
         Minimum c _ -> IR.Operation "MIN" [c.expression]
         Shortest c _ -> IR.Operation "SHORTEST" [c.expression]
diff --git a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Base_Generator.enso b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Base_Generator.enso
index 2c8fa0678f3d..b6fef9c79287 100644
--- a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Base_Generator.enso
+++ b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/Base_Generator.enso
@@ -190,6 +190,7 @@ generate_expression dialect expr = case expr of
         op = dialect.operation_map.get_or_else kind (Error.throw <| Unsupported_Database_Operation_Error kind)
         parsed_args = arguments.map (here.generate_expression dialect)
         op parsed_args
+    IR.Order_Descriptor _ _ _ _ -> here.generate_order dialect expr
 
 ## PRIVATE
 
diff --git a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/IR.enso b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/IR.enso
index e6e4137c5910..abea5017d1e3 100644
--- a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/IR.enso
+++ b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Internal/IR.enso
@@ -40,7 +40,7 @@ type Expression
        - kind: the name of the operation, these can be both functions or infix
          operators, the actual implementation is determined by a specific
          dialect.
-       - expression: a list of expressions which are arguments to the operation;
+       - expressions: a list of expressions which are arguments to the operation
          different operations support different amounts of arguments.
     type Operation (kind : Text) (expressions : Vector Expression)
 
@@ -277,7 +277,6 @@ type Join_Kind
        in the query can be used to filter the results.
     type Join_Cross
 
-
 ## PRIVATE
 type Order_Descriptor (expression : Expression) (direction : Sort_Direction) (nulls_order : Nothing | Nulls_Order = Nothing) (collation : Nothing | Text = Nothing)
 
diff --git a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Table.enso b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Table.enso
index 8329515bc009..fdb28b90f287 100644
--- a/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Table.enso
+++ b/distribution/lib/Standard/Database/0.0.0-dev/src/Data/Table.enso
@@ -8,6 +8,7 @@ import Standard.Table.Data.Column as Materialized_Column
 import Standard.Table.Data.Table as Materialized_Table
 import Standard.Table.Internal.Java_Exports
 import Standard.Table.Internal.Table_Helpers
+import Standard.Table.Internal.Problem_Builder
 
 import Standard.Table.Data.Aggregate_Column
 import Standard.Table.Internal.Aggregate_Column_Helper
@@ -480,17 +481,19 @@ type Table
              table.order_by (Sort_Column_Selector.By_Index [Sort_Column.Index 1, Sort_Column.Index -7 Sort_Direction.Descending])
     order_by : Sort_Column_Selector -> Text_Ordering -> Problem_Behavior -> Table
     order_by (columns = (Sort_Column_Selector.By_Name [(Sort_Column.Name (this.columns.at 0 . name))])) text_ordering=Text_Ordering on_problems=Report_Warning = Panic.handle_wrapped_dataflow_error <|
-        columns_for_ordering = Table_Helpers.prepare_order_by this.internal_columns columns on_problems
-        new_order_descriptors = columns_for_ordering.map selected_column->
-            internal_column = selected_column.column
-            associated_selector = selected_column.associated_selector
-            ## TODO [RW] this is only needed because `Vector.map` does not
-               propagate dataflow errors correctly. See:
-               https://www.pivotaltracker.com/story/show/181057718
-            Panic.throw_wrapped_if_error <|
-                this.connection.dialect.prepare_order_descriptor internal_column associated_selector.direction text_ordering
-        new_ctx = this.context.add_orders new_order_descriptors
-        this.updated_context new_ctx
+        problem_builder = Problem_Builder.new
+        columns_for_ordering = Table_Helpers.prepare_order_by this.columns columns problem_builder
+        problem_builder.attach_problems_before on_problems <|
+            new_order_descriptors = columns_for_ordering.map selected_column->
+                internal_column = selected_column.column
+                associated_selector = selected_column.associated_selector
+                ## TODO [RW] this is only needed because `Vector.map` does not
+                   propagate dataflow errors correctly. See:
+                   https://www.pivotaltracker.com/story/show/181057718
+                Panic.throw_wrapped_if_error <|
+                    this.connection.dialect.prepare_order_descriptor internal_column associated_selector.direction text_ordering
+            new_ctx = this.context.add_orders new_order_descriptors
+            this.updated_context new_ctx
 
     ## UNSTABLE
 
diff --git a/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso b/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
index 0e3e3a8e2760..09ffe4d069ec 100644
--- a/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
+++ b/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
@@ -1,6 +1,8 @@
 from Standard.Base import all
 
 from Standard.Table.Data.Column as Column_Module import Column
+import Standard.Table.Data.Column_Selector
+import Standard.Table.Data.Sort_Column_Selector
 
 ## Defines an Aggregate Column
 type Aggregate_Column
@@ -140,7 +142,7 @@ type Aggregate_Column
          not missing value returned.
        - order_by: required for database tables. Specifies how to order the
          results within the group.
-    type First (column:Column|Text|Integer) (new_name:Text|Nothing=Nothing) (ignore_nothing:Boolean=True) (order_by:Column_Selector|Nothing=Nothing)
+    type First (column:Column|Text|Integer) (new_name:Text|Nothing=Nothing) (ignore_nothing:Boolean=True) (order_by:Sort_Column_Selector|Nothing=Nothing)
 
     ## Creates a new column with the last value in each group. If no rows,
        evaluates to `Nothing`.
@@ -153,7 +155,7 @@ type Aggregate_Column
          not missing value returned.
        - order_by: required for database tables. Specifies how to order the
          results within the group.
-    type Last (column:Column|Text|Integer) (new_name:Text|Nothing=Nothing) (ignore_nothing:Boolean=True) (order_by:Column_Selector|Nothing=Nothing)
+    type Last (column:Column|Text|Integer) (new_name:Text|Nothing=Nothing) (ignore_nothing:Boolean=True) (order_by:Sort_Column_Selector|Nothing=Nothing)
 
     ## Creates a new column with the maximum value in each group. If no rows,
        evaluates to `Nothing`.
diff --git a/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Table.enso b/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Table.enso
index 1a138a67b622..ddbe1527aab3 100644
--- a/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Table.enso
+++ b/distribution/lib/Standard/Table/0.0.0-dev/src/Data/Table.enso
@@ -11,6 +11,7 @@ import Standard.Table.Internal.Table_Helpers
 import Standard.Table.Internal.Aggregate_Column_Helper
 import Standard.Table.Internal.Parse_Values_Helper
 import Standard.Table.Internal.Delimited_Reader
+import Standard.Table.Internal.Problem_Builder
 
 from Standard.Table.Data.Order_Rule as Order_Rule_Module import Order_Rule
 from Standard.Table.Data.Column_Selector as Column_Selector_Module import Column_Selector, By_Index
@@ -572,16 +573,14 @@ type Table
              table.order_by (Sort_Column_Selector.By_Index [Sort_Column.Index 1, Sort_Column.Index -7 Sort_Direction.Descending])
     order_by : Sort_Column_Selector -> Text_Ordering -> Problem_Behavior -> Table
     order_by (columns = (Sort_Column_Selector.By_Name [(Sort_Column.Name (this.columns.at 0 . name))])) text_ordering=Text_Ordering on_problems=Report_Warning =
-        columns_for_ordering = Table_Helpers.prepare_order_by this.columns columns on_problems
-        selected_columns = columns_for_ordering.map c->c.column.java_column
-        ordering = columns_for_ordering.map c->
-            case c.associated_selector.direction of
-                Sort_Direction.Ascending -> 1
-                Sort_Direction.Descending -> -1
-        comparator = Comparator.for_text_ordering text_ordering
-        Table <|
-            this.java_table.orderBy selected_columns.to_array ordering.to_array comparator
-
+        problem_builder = Problem_Builder.new
+        columns_for_ordering = Table_Helpers.prepare_order_by this.columns columns problem_builder
+        problem_builder.attach_problems_before on_problems <|
+            selected_columns = columns_for_ordering.map c->c.column.java_column
+            ordering = columns_for_ordering.map c->c.associated_selector.direction.to_sign
+            comparator = Comparator.for_text_ordering text_ordering
+            java_table = this.java_table.orderBy selected_columns.to_array ordering.to_array comparator
+            Table java_table
 
     ## Parses columns within a Table to a specific value type.
        By default, it looks at all `Text` columns and attempts to deduce the
diff --git a/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Aggregate_Column_Helper.enso b/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Aggregate_Column_Helper.enso
index 3b6d84cec9d9..ea4e59eee68e 100644
--- a/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Aggregate_Column_Helper.enso
+++ b/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Aggregate_Column_Helper.enso
@@ -8,6 +8,9 @@ import Standard.Table.Internal.Problem_Builder
 import Standard.Table.Internal.Unique_Name_Strategy
 import Standard.Table.Internal.Table_Helpers
 
+import Standard.Table.Data.Sort_Column_Selector
+import Standard.Table.Data.Sort_Column
+
 import Standard.Base.Data.Ordering.Comparator
 
 from Standard.Table.Error as Error_Module import Missing_Input_Columns, Column_Indexes_Out_Of_Range, No_Output_Columns, Duplicate_Output_Column_Names, Invalid_Output_Column_Names, Invalid_Aggregation, Floating_Point_Grouping, Unquoted_Delimiter, Additional_Warnings
@@ -103,6 +106,9 @@ default_aggregate_column_name aggregate_column =
    indices or column references potentially from a different table) are
    replaced with column references from the provided table.
 
+   Sort_Column_Selectors are resolved to Sort_Column_Select.By_Column with the
+   matched columns coming from the provided table.
+
    This preprocess step is required by some helper function, to avoid having
    to pass the table reference and resolve the column descriptors all the
    time.
@@ -123,9 +129,13 @@ resolve_aggregate table problem_builder aggregate_column =
         resolved = Table_Helpers.select_columns_helper table_columns selector reorder=True problem_builder
         if resolved.is_empty then Error.throw Internal_Missing_Column_Error else resolved
 
-    resolve_selector_or_nothing selector = case selector of
+    resolve_order_by selector = case selector of
         Nothing -> Nothing
-        _ -> resolve_selector_to_vector selector
+        _ ->
+            columns_for_ordering = Table_Helpers.prepare_order_by table_columns selector problem_builder
+            sort_columns = columns_for_ordering.map c->
+                Sort_Column.Column c.column c.associated_selector.direction
+            Sort_Column_Selector.By_Column sort_columns
 
     result = case aggregate_column of
         Group_By c new_name -> Group_By (resolve c) new_name
@@ -149,8 +159,8 @@ resolve_aggregate table problem_builder aggregate_column =
         Mode c new_name -> Mode (resolve c) new_name
         Standard_Deviation c new_name population -> Standard_Deviation (resolve c) new_name population
         Concatenate c new_name separator prefix suffix quote_char -> Concatenate (resolve c) new_name separator prefix suffix quote_char
-        First c new_name ignore_nothing order_by -> First (resolve c) new_name ignore_nothing (resolve_selector_or_nothing order_by)
-        Last c new_name ignore_nothing order_by -> Last (resolve c) new_name ignore_nothing (resolve_selector_or_nothing order_by)
+        First c new_name ignore_nothing order_by -> First (resolve c) new_name ignore_nothing (resolve_order_by order_by)
+        Last c new_name ignore_nothing order_by -> Last (resolve c) new_name ignore_nothing (resolve_order_by order_by)
         Maximum c new_name -> Maximum (resolve c) new_name
         Minimum c new_name -> Minimum (resolve c) new_name
         Shortest c new_name -> Shortest (resolve c) new_name
@@ -188,12 +198,14 @@ java_aggregator name column =
         Mode c _ -> ModeAggregator.new name c.java_column
         First c _ ignore_nothing ordering ->
             if ordering.is_nothing then FirstAggregator.new name c.java_column ignore_nothing else
-                ordering_array = ordering.map .java_column
-                FirstAggregator.new name c.java_column ignore_nothing ordering_array.to_array Comparator.new
+                order_columns = ordering.columns.map c->c.column.java_column
+                order_directions = ordering.columns.map c->c.direction.to_sign
+                FirstAggregator.new name c.java_column ignore_nothing order_columns.to_array order_directions.to_array Comparator.new
         Last c _ ignore_nothing ordering ->
             if ordering.is_nothing then LastAggregator.new name c.java_column ignore_nothing else
-                ordering_array = ordering.map .java_column
-                LastAggregator.new name c.java_column ignore_nothing ordering_array.to_array Comparator.new
+                order_columns = ordering.columns.map c->c.column.java_column
+                order_direction = ordering.columns.map c->c.direction.to_sign
+                LastAggregator.new name c.java_column ignore_nothing order_columns.to_array order_direction.to_array Comparator.new
         Maximum c _ -> MinOrMaxAggregator.new name c.java_column 1 Comparator.new
         Minimum c _ -> MinOrMaxAggregator.new name c.java_column -1 Comparator.new
         Shortest c _ -> ShortestOrLongestAggregator.new name c.java_column -1
diff --git a/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Table_Helpers.enso b/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Table_Helpers.enso
index 44134e741ded..ed20dfb25f1a 100644
--- a/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Table_Helpers.enso
+++ b/distribution/lib/Standard/Table/0.0.0-dev/src/Internal/Table_Helpers.enso
@@ -361,9 +361,8 @@ validate_unique vector problem_callback on=(x->x) =
 type Column_Transform_Element column associated_selector
 
 ## PRIVATE
-prepare_order_by : Vector -> Problem_Behavior -> Vector Column_Transform_Element
-prepare_order_by internal_columns column_selectors on_problems =
-    problem_builder = Problem_Builder.new
+prepare_order_by : Vector -> Problem_Builder -> Vector Column_Transform_Element
+prepare_order_by internal_columns column_selectors problem_builder =
     selected_elements = case column_selectors of
         Sort_Column_Selector.By_Name name_selectors matcher ->
             here.select_columns_by_name internal_columns name_selectors matcher problem_builder name_extractor=(_.name)
@@ -373,7 +372,7 @@ prepare_order_by internal_columns column_selectors on_problems =
             here.select_columns_by_column_reference internal_columns column_selectors problem_builder column_extractor=(_.column)
     if selected_elements.is_empty then
         problem_builder.report_other_warning No_Input_Columns_Selected
-    problem_builder.attach_problems_after on_problems selected_elements
+    selected_elements
 
 ## PRIVATE
    A helper function which can be used by methods that transform a subset of
diff --git a/std-bits/table/src/main/java/org/enso/table/aggregations/First.java b/std-bits/table/src/main/java/org/enso/table/aggregations/First.java
index 1a9f96d04704..8f027b22a5c1 100644
--- a/std-bits/table/src/main/java/org/enso/table/aggregations/First.java
+++ b/std-bits/table/src/main/java/org/enso/table/aggregations/First.java
@@ -11,33 +11,39 @@
 /** Aggregate Column finding the first value in a group. */
 public class First extends Aggregator {
   private final Storage storage;
-  private final Storage[] ordering;
+  private final Storage[] orderByColumns;
+  private final int[] orderByDirections;
   private final Comparator<Object> objectComparator;
   private final boolean ignoreNothing;
 
   public First(String name, Column column, boolean ignoreNothing) {
-    this(name, column, ignoreNothing, null, null);
+    this(name, column, ignoreNothing, null, null, null);
   }
 
   public First(
       String name,
       Column column,
       boolean ignoreNothing,
-      Column[] ordering,
+      Column[] orderByColumns,
+      Long[] orderByDirections,
       Comparator<Object> objectComparator) {
     super(name, Storage.Type.OBJECT);
     this.storage = column.getStorage();
-    this.ordering =
-        ordering == null
+    this.orderByColumns =
+        orderByColumns == null
             ? new Storage[0]
-            : Arrays.stream(ordering).map(Column::getStorage).toArray(Storage[]::new);
+            : Arrays.stream(orderByColumns).map(Column::getStorage).toArray(Storage[]::new);
+    this.orderByDirections =
+        orderByDirections == null
+            ? new int[0]
+            : Arrays.stream(orderByDirections).mapToInt(Long::intValue).toArray();
     this.objectComparator = objectComparator;
     this.ignoreNothing = ignoreNothing;
   }
 
   @Override
   public Object aggregate(List<Integer> indexes) {
-    if (ordering.length == 0) {
+    if (orderByColumns.length == 0) {
       return firstByRowOrder(indexes);
     } else {
       return firstBySpecifiedOrder(indexes);
@@ -54,7 +60,8 @@ private Object firstBySpecifiedOrder(List<Integer> indexes) {
         continue;
       }
 
-      MultiValueKey newKey = new MultiValueKey(this.ordering, row, objectComparator);
+      MultiValueKey newKey =
+          new MultiValueKey(this.orderByColumns, row, this.orderByDirections, objectComparator);
       if (key == null || key.compareTo(newKey) > 0) {
         key = newKey;
         current = storage.getItemBoxed(row);
diff --git a/std-bits/table/src/main/java/org/enso/table/aggregations/Last.java b/std-bits/table/src/main/java/org/enso/table/aggregations/Last.java
index 2b922a59fe5f..a5ecbc836b08 100644
--- a/std-bits/table/src/main/java/org/enso/table/aggregations/Last.java
+++ b/std-bits/table/src/main/java/org/enso/table/aggregations/Last.java
@@ -10,33 +10,39 @@
 
 public class Last extends Aggregator {
   private final Storage storage;
-  private final Storage[] ordering;
+  private final Storage[] orderByColumns;
+  private final int[] orderByDirections;
   private final Comparator<Object> objectComparator;
   private final boolean ignoreNothing;
 
   public Last(String name, Column column, boolean ignoreNothing) {
-    this(name, column, ignoreNothing, null, null);
+    this(name, column, ignoreNothing, null, null, null);
   }
 
   public Last(
       String name,
       Column column,
       boolean ignoreNothing,
-      Column[] ordering,
+      Column[] orderByColumns,
+      Long[] orderByDirections,
       Comparator<Object> objectComparator) {
     super(name, Storage.Type.OBJECT);
     this.storage = column.getStorage();
-    this.ordering =
-        ordering == null
+    this.orderByColumns =
+        orderByColumns == null
             ? new Storage[0]
-            : Arrays.stream(ordering).map(Column::getStorage).toArray(Storage[]::new);
+            : Arrays.stream(orderByColumns).map(Column::getStorage).toArray(Storage[]::new);
+    this.orderByDirections =
+        orderByDirections == null
+            ? new int[0]
+            : Arrays.stream(orderByDirections).mapToInt(Long::intValue).toArray();
     this.objectComparator = objectComparator;
     this.ignoreNothing = ignoreNothing;
   }
 
   @Override
   public Object aggregate(List<Integer> indexes) {
-    if (ordering.length == 0) {
+    if (orderByColumns.length == 0) {
       return lastByRowOrder(indexes);
     } else {
       return lastBySpecifiedOrder(indexes);
@@ -54,7 +60,8 @@ private Object lastBySpecifiedOrder(List<Integer> indexes) {
         continue;
       }
 
-      MultiValueKey newKey = new MultiValueKey(this.ordering, row, objectComparator);
+      MultiValueKey newKey =
+          new MultiValueKey(this.orderByColumns, row, this.orderByDirections, objectComparator);
       if (key == null || key.compareTo(newKey) < 0) {
         key = newKey;
         current = storage.getItemBoxed(row);
diff --git a/test/Table_Tests/src/Aggregate_Spec.enso b/test/Table_Tests/src/Aggregate_Spec.enso
index 33e7fe70bc16..44af889f006e 100644
--- a/test/Table_Tests/src/Aggregate_Spec.enso
+++ b/test/Table_Tests/src/Aggregate_Spec.enso
@@ -2,6 +2,8 @@ from Standard.Base import all
 
 import Standard.Table
 from Standard.Table.Data.Column_Selector import By_Name, By_Index
+import Standard.Table.Data.Sort_Column
+import Standard.Table.Data.Sort_Column_Selector
 from Standard.Table.Data.Aggregate_Column import all
 from Standard.Table.Error as Error_Module import Missing_Input_Columns, Column_Indexes_Out_Of_Range, No_Output_Columns, Duplicate_Output_Column_Names, Invalid_Output_Column_Names, Invalid_Aggregation, Floating_Point_Grouping, Unquoted_Delimiter, Additional_Warnings
 from Standard.Database.Error as Database_Errors import Unsupported_Database_Operation_Error
@@ -142,7 +144,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 5 . at 0 . should_equal -17.960000 epsilon=0.000001
 
         Test.specify "should be able to get first and last values" (pending = resolve_pending test_selection.first_last) <|
-            grouped = table.aggregate [First "Index" (order_by = By_Name ["Hexadecimal", "TextWithNothing"]), Last "ValueWithNothing" (order_by = By_Name ["Value"])]
+            grouped = table.aggregate [First "Index" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Hexadecimal", Sort_Column.Name "TextWithNothing"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value"])]
             materialized = materialize grouped
             grouped.row_count . should_equal 1
             materialized.columns.length . should_equal 2
@@ -151,6 +153,18 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 1 . name . should_equal "Last ValueWithNothing"
             materialized.columns.at 1 . at 0 . should_equal -89.78 epsilon=0.000001
 
+        Test.specify "should be able to get first and last values with mixed ordering" (pending = resolve_pending test_selection.first_last) <|
+            grouped = table.aggregate [First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending, Sort_Column.Name "Code"]), First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Code", Sort_Column.Name "Value" Sort_Direction.Descending]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending])]
+            materialized = materialize grouped
+            grouped.row_count . should_equal 1
+            materialized.columns.length . should_equal 3
+            materialized.columns.at 0 . name . should_equal "First TextWithNothing"
+            materialized.columns.at 0 . at 0 . should_equal "riwaiqq1io"
+            materialized.columns.at 1 . name . should_equal "First TextWithNothing_1"
+            materialized.columns.at 1 . at 0 . should_equal "j4i2ua7uft"
+            materialized.columns.at 2 . name . should_equal "Last ValueWithNothing"
+            materialized.columns.at 2 . at 0 . should_equal -38.56 epsilon=0.000001
+
         Test.specify "should be able to get first and last values with default row order" (pending = resolve_pending test_selection.first_last_row_order) <|
             grouped = table.aggregate [First "Index", Last "Value"]
             materialized = materialize grouped
@@ -259,7 +273,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 2 . at 0 . should_equal Nothing
 
         Test.specify "should be able to get first and last values" (pending = resolve_pending test_selection.first_last) <|
-            grouped = empty_table.aggregate [First "Index" (order_by = By_Name ["Hexadecimal", "TextWithNothing"]), Last "ValueWithNothing" (order_by = By_Name ["Value"])]
+            grouped = empty_table.aggregate [First "Index" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Hexadecimal", Sort_Column.Name "TextWithNothing"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value"])]
             materialized = materialize grouped
             grouped.row_count . should_equal 1
             materialized.columns.length . should_equal 2
@@ -363,7 +377,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 3 . name . should_equal "25%-ile Value"
 
         Test.specify "should be able to get first and last values" (pending = resolve_pending test_selection.first_last) <|
-            grouped = empty_table.aggregate [Group_By 0, First "Index" (order_by = By_Name ["Hexadecimal", "TextWithNothing"]), Last "ValueWithNothing" (order_by = By_Name ["Value"])]
+            grouped = empty_table.aggregate [Group_By 0, First "Index" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Hexadecimal", Sort_Column.Name "TextWithNothing"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value"])]
             materialized = materialize grouped
             grouped.row_count . should_equal 0
             materialized.columns.length . should_equal 3
@@ -517,7 +531,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 6 . at idx . should_equal -18.802000 epsilon=0.000001
 
         Test.specify "should be able to get first and last values" (pending = resolve_pending test_selection.first_last) <|
-            grouped = table.aggregate [Group_By "Index", First "TextWithNothing" (order_by = By_Name ["Value", "Flag"]), Last "ValueWithNothing" (order_by = By_Name ["Value"])]
+            grouped = table.aggregate [Group_By "Index", First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value", Sort_Column.Name "Flag"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value"])]
             materialized = materialize grouped
             grouped.row_count . should_equal 10
             materialized.columns.length . should_equal 3
@@ -529,6 +543,19 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 2 . name . should_equal "Last ValueWithNothing"
             materialized.columns.at 2 . at idx . should_equal -89.78 epsilon=0.000001
 
+        Test.specify "should be able to get first and last values with mixed ordering" (pending = resolve_pending test_selection.first_last) <|
+            grouped = table.aggregate [Group_By "Index", First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending, Sort_Column.Name "Flag"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending])]
+            materialized = materialize grouped
+            grouped.row_count . should_equal 10
+            materialized.columns.length . should_equal 3
+            materialized.columns.at 0 . name . should_equal "Index"
+            idx = find_row [7] materialized
+            idx.is_nothing . should_be_false
+            materialized.columns.at 1 . name . should_equal "First TextWithNothing"
+            materialized.columns.at 1 . at idx . should_equal "riwaiqq1io"
+            materialized.columns.at 2 . name . should_equal "Last ValueWithNothing"
+            materialized.columns.at 2 . at idx . should_equal -63.75 epsilon=0.000001
+
         Test.specify "should be able to get first and last values with default row order" (pending = resolve_pending test_selection.first_last_row_order) <|
             grouped = table.aggregate [Group_By "Index", First "TextWithNothing", Last "Value"]
             materialized = materialize grouped
@@ -701,7 +728,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 7 . at idx . should_equal -17.174000 epsilon=0.000001
 
         Test.specify "should be able to get first and last values" (pending = resolve_pending test_selection.first_last) <|
-            grouped = table.aggregate [Group_By "Flag", First "TextWithNothing" (order_by = By_Name ["Value", "Flag"]), Last "ValueWithNothing" (order_by = By_Name ["Value"]), Group_By "Index"]
+            grouped = table.aggregate [Group_By "Flag", First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value", Sort_Column.Name "Flag"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value"]), Group_By "Index"]
             materialized = materialize grouped
             grouped.row_count . should_equal 20
             materialized.columns.length . should_equal 4
@@ -714,6 +741,20 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             materialized.columns.at 2 . name . should_equal "Last ValueWithNothing"
             materialized.columns.at 2 . at idx . should_equal -89.78 epsilon=0.000001
 
+        Test.specify "should be able to get first and last values with mixed ordering" (pending = resolve_pending test_selection.first_last) <|
+            grouped = table.aggregate [Group_By "Flag", First "TextWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending, Sort_Column.Name "Flag"]), Last "ValueWithNothing" (order_by = Sort_Column_Selector.By_Name [Sort_Column.Name "Value" Sort_Direction.Descending]), Group_By "Index"]
+            materialized = materialize grouped
+            grouped.row_count . should_equal 20
+            materialized.columns.length . should_equal 4
+            materialized.columns.at 0 . name . should_equal "Flag"
+            materialized.columns.at 3 . name . should_equal "Index"
+            idx = find_row [True, 7] materialized [0, 3]
+            idx.is_nothing . should_be_false
+            materialized.columns.at 1 . name . should_equal "First TextWithNothing"
+            materialized.columns.at 1 . at idx . should_equal "13dir782ah"
+            materialized.columns.at 2 . name . should_equal "Last ValueWithNothing"
+            materialized.columns.at 2 . at idx . should_equal 54.48 epsilon=0.000001
+
         Test.specify "should be able to get first and last values with default row order" (pending = resolve_pending test_selection.first_last_row_order) <|
             grouped = table.aggregate [Group_By "Flag", First "TextWithNothing", Last "Value", Group_By "Index"]
             materialized = materialize grouped
@@ -1090,7 +1131,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
     Test.group prefix+"Table.aggregate First and Last" pending=pending <|
         Test.specify "should not return the same value for groups with different values but equal ordering keys" (pending = resolve_pending test_selection.first_last) <|
             t1 = table_builder [["G", ["a", "a"]], ["X", [1, 2]]]
-            order = By_Name ["G"]
+            order = Sort_Column_Selector.By_Name [Sort_Column.Name "G"]
             r1 = t1.aggregate [First "X" (order_by=order), Last "X" (order_by=order)]
             r1.row_count.should_equal 1
             m1 = materialize r1
@@ -1294,7 +1335,7 @@ aggregate_spec prefix table empty_table table_builder materialize is_database te
             if test_selection.first_last.not then
                 Test.specify "with First and Last with ordering" <|
                     table = table_builder [["A", [3,2,1]], ["X", [1,2,3]]]
-                    order = By_Name ["A"]
+                    order = Sort_Column_Selector.By_Name [Sort_Column.Name "A"]
                     expect_sum_and_unsupported_errors 2 <|
                         table.aggregate [Sum "X", First "X" (order_by=order), Last "X" (order_by=order)]