DLHELP-10200 Use upper median in case of even number of values (#466)

* DLHELP-10200 Use upper median in case of even number of values * mypy
datalens-tech · May 30, 2024 · 60a8c27 · 60a8c27
1 parent 7f871d3
commit 60a8c27
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 13 deletions.
diff --git a/lib/dl_connector_oracle/dl_connector_oracle/formula/definitions/functions_aggregation.py b/lib/dl_connector_oracle/dl_connector_oracle/formula/definitions/functions_aggregation.py
@@ -42,7 +42,7 @@
     # median
     base.AggMedian(
         variants=[
-            V(D.ORACLE, lambda expr: within_group(sa.func.PERCENTILE_DISC(0.5), expr)),
+            V(D.ORACLE, lambda expr: within_group(sa.func.PERCENTILE_DISC(0.5), expr.desc())),
         ]
     ),
     # min

diff --git a/...connector_postgresql/dl_connector_postgresql/formula/definitions/functions_aggregation.py b/...connector_postgresql/dl_connector_postgresql/formula/definitions/functions_aggregation.py
@@ -70,7 +70,7 @@ def _all_concat_pg(expr: ClauseElement, sep: str = ", ") -> ClauseElement:
         variants=[
             V(
                 D.and_above(D.POSTGRESQL_9_4) | D.COMPENG,
-                lambda expr: (within_group(sa.func.PERCENTILE_DISC(0.5), expr)),
+                lambda expr: (within_group(sa.func.PERCENTILE_DISC(0.5), expr.desc())),
             ),
         ]
     ),
@@ -82,7 +82,6 @@ def _all_concat_pg(expr: ClauseElement, sep: str = ", ") -> ClauseElement:
             V(
                 D.and_above(D.POSTGRESQL_9_4) | D.COMPENG,
                 lambda expr, quant: (
-                    # In MSSQLSRV window functions cannot be used as aggregations
                     within_group(
                         sa.func.percentile_disc(quantile_value(un_literal(quant))),
                         expr,

diff --git a/...l_connector_snowflake/dl_connector_snowflake/formula/definitions/functions_aggregation.py b/...l_connector_snowflake/dl_connector_snowflake/formula/definitions/functions_aggregation.py
@@ -23,7 +23,7 @@ class AggMedianIntSF(base.AggMedian):
         V(
             D.SNOWFLAKE,
             lambda expr: sa.func.TO_NUMBER(
-                within_group(sa.func.PERCENTILE_DISC(0.5), expr),
+                within_group(sa.func.PERCENTILE_DISC(0.5), expr.desc()),
             ),
         ),
     ]
@@ -37,7 +37,7 @@ class AggMedianFloatSF(base.AggMedian):
     variants = [
         V(
             D.SNOWFLAKE,
-            lambda expr: within_group(sa.func.PERCENTILE_DISC(0.5), expr),
+            lambda expr: within_group(sa.func.PERCENTILE_DISC(0.5), expr.desc()),
         ),
     ]
 
@@ -52,7 +52,7 @@ class AggMedianDateSF(base.AggMedian):
                 sa.func.TO_TIMESTAMP(
                     within_group(
                         sa.func.PERCENTILE_DISC(0.5),
-                        sa.func.DATE_PART("EPOCH_SECOND", expr),
+                        sa.func.DATE_PART("EPOCH_SECOND", expr.desc()),
                     )
                 )
             ),
@@ -69,7 +69,7 @@ class AggMedianDateTimeSF(base.AggMedian):
             lambda expr: sa.func.TO_TIMESTAMP(
                 within_group(
                     sa.func.PERCENTILE_DISC(0.5),
-                    sa.func.DATE_PART("EPOCH_SECOND", expr),
+                    sa.func.DATE_PART("EPOCH_SECOND", expr.desc()),
                 )
             ),
         ),

diff --git a/lib/dl_formula_testing/dl_formula_testing/testcases/functions_aggregation.py b/lib/dl_formula_testing/dl_formula_testing/testcases/functions_aggregation.py
@@ -132,19 +132,20 @@ def test_median(self, dbe: DbEvaluator, data_table: sa.Table) -> None:
         if not self.supports_median:
             pytest.skip()
 
+        int_values = data_table.int_values  # type: ignore  # 2024-05-29 # TODO: "Table" has no attribute "int_values"  [attr-defined]
         date_values = data_table.date_values  # type: ignore  # 2024-01-29 # TODO: "Table" has no attribute "date_values"  [attr-defined]
         datetime_values = data_table.datetime_values  # type: ignore  # 2024-01-29 # TODO: "Table" has no attribute "datetime_values"  [attr-defined]
 
         VALUE_TV = TypeVar("VALUE_TV")
 
-        def medians(values: Collection[Union[VALUE_TV]]) -> tuple[VALUE_TV, VALUE_TV]:
+        def median(values: Collection[Union[VALUE_TV]]) -> VALUE_TV:
             values = sorted(values)  # type: ignore  # 2024-01-30 # TODO: Value of type variable "SupportsRichComparisonT" of "sorted" cannot be "VALUE_TV"  [type-var]
-            middle = len(values) // 2 - 1
-            return (values[middle], values[middle + 1])
+            upper_middle = len(values) // 2 + 1
+            return values[upper_middle]
 
-        assert dbe.eval("MEDIAN([int_value])", from_=data_table) in (40, 50)
-        assert dbe.eval("MEDIAN([date_value])", from_=data_table) in medians(date_values)
-        assert dbe.eval("MEDIAN([datetime_value])", from_=data_table) in medians(datetime_values)
+        assert dbe.eval("MEDIAN([int_value])", from_=data_table) == median(int_values)
+        assert dbe.eval("MEDIAN([date_value])", from_=data_table) == median(date_values)
+        assert dbe.eval("MEDIAN([datetime_value])", from_=data_table) == median(datetime_values)
 
     def test_any(self, dbe: DbEvaluator, data_table: sa.Table) -> None:
         if not self.supports_any: