expanded tests, rewrote fix a tiny bit and removed unnecessary kdocs. The new behavior is expected, no need to mention it

Author: Jolanrensen

core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/api/valueCounts.kt

@@ -49,7 +49,6 @@ private typealias ValueCountsParams = Nothing
  * - the distinct combinations of these values,
  * - a new [Int] column with the number of occurrences of each combination.
  * This column is called `"count"` by default unless it is overridden by [resultColumn\].
- * If no rows are counted, the result still contains the selected top-level columns and the count column.
  *
  * By default, the result is sorted by count in descending order (the most frequent combination first),
  * and a row is not counted at all if any of its selected values is {@include [NALink]}.
@@ -100,7 +99,6 @@ internal val defaultCountColumnName: String = ValueCount::count.name
  * - the column with the distinct values of the original [DataColumn]
  * - a new [Int] column with the number of occurrences of each value.
  * This column is called `"count"` by default unless it is overridden by [resultColumn].
- * If no values are counted, the result still contains the observed-value column and the count column.
  *
  * By default, the result is sorted by count in descending order (the most frequent value first),
  * and {@include [NALink]} values (`null`s and `NaN`s) are not counted.
@@ -150,11 +148,14 @@ public fun <T> DataColumn<T>.valueCounts(
     }
     if (dropNA) grouped = grouped.filter { !it.first.isNA }
     val nulls = if (dropNA) false else hasNulls()
-    val values = if (grouped.isEmpty()) {
-        get(emptyList<Int>())
-    } else {
-        DataColumn.createByType(name(), grouped.map { it.first }, type().withNullability(nulls))
-    }
+    val values =
+        if (grouped.isEmpty()) {
+            // #2034, keep the original name and type of the column even when the input is empty
+            // nullability is irrelevant, but let's keep it for consistency
+            take(0).setNullable(nulls)
+        } else {
+            DataColumn.createByType(name(), grouped.map { it.first }, type().withNullability(nulls))
+        }
     val countName = if (resultColumn == name()) resultColumn + "1" else resultColumn
     val counts = DataColumn.createByType(countName, grouped.map { it.second }, typeOf<Int>())
     return dataFrameOf(values, counts).cast()

core/src/test/kotlin/org/jetbrains/kotlinx/dataframe/api/valueCounts.kt

@@ -266,7 +266,7 @@ class ValueCountsTests {
     }
 
     @Test
-    fun `value counts of an empty DataFrame retains all observed columns`() {
+    fun `value counts of an empty DataFrame retains all columns`() {
         val result = df.take(0).valueCounts()
 
         result.columnNames() shouldBe listOf("name", "city", "age", "count")
@@ -283,6 +283,7 @@ class ValueCountsTests {
         result.columnNames() shouldBe listOf("city", "count")
         result.nrow shouldBe 0
         result["city"].type() shouldBe typeOf<String>()
+        result["count"].type() shouldBe typeOf<Int>()
     }
 
     @Test
@@ -292,6 +293,8 @@ class ValueCountsTests {
 
         result.columnNames() shouldBe listOf("city", "count")
         result.nrow shouldBe 0
+        result["city"].type() shouldBe typeOf<String?>() // TODO #2034
+        result["count"].type() shouldBe typeOf<Int>()
     }
 
     @Test
@@ -302,6 +305,9 @@ class ValueCountsTests {
         result.columnNames() shouldBe listOf("details", "count")
         result["details"].asColumnGroup().columnNames() shouldBe listOf("name", "city")
         result.nrow shouldBe 0
+        result["details"].asColumnGroup()["name"].type() shouldBe typeOf<String?>() // TODO #2034
+        result["details"].asColumnGroup()["city"].type() shouldBe typeOf<String?>() // TODO #2034
+        result["count"].type() shouldBe typeOf<Int>()
     }
 
     // endregion