Changed parsing to use createColumnGuessingType() to fix issue #593 for CSV and other parsing usages

Author: Jolanrensen

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

@@ -32,6 +32,7 @@ import org.jetbrains.kotlinx.dataframe.exceptions.TypeConversionException
 import org.jetbrains.kotlinx.dataframe.hasNulls
 import org.jetbrains.kotlinx.dataframe.impl.canParse
 import org.jetbrains.kotlinx.dataframe.impl.catchSilent
+import org.jetbrains.kotlinx.dataframe.impl.columns.createColumnGuessingType
 import org.jetbrains.kotlinx.dataframe.impl.createStarProjectedType
 import org.jetbrains.kotlinx.dataframe.impl.javaDurationCanParse
 import org.jetbrains.kotlinx.dataframe.io.isURL
@@ -525,7 +526,10 @@ internal fun DataColumn<String?>.tryParseImpl(options: ParserOptions?): DataColu
     if (type.jvmErasure == String::class && !nullStringParsed) {
         return this // nothing parsed
     }
-    return DataColumn.createUnsafe(name(), parsedValues, type)
+
+    // Create a new column with the parsed values,
+    // createColumnGuessingType is used to handle unifying values if needed
+    return createColumnGuessingType(name(), parsedValues, type)
 }
 
 internal fun <T> DataColumn<String?>.parse(parser: StringParser<T>, options: ParserOptions?): DataColumn<T?> {
@@ -535,7 +539,7 @@ internal fun <T> DataColumn<String?>.parse(parser: StringParser<T>, options: Par
             handler(it.trim()) ?: throw IllegalStateException("Couldn't parse '$it' into type ${parser.type}")
         }
     }
-    return DataColumn.createValueColumn(name(), parsedValues, parser.type.withNullability(hasNulls)) as DataColumn<T?>
+    return createColumnGuessingType(name(), parsedValues, parser.type.withNullability(hasNulls))
 }
 
 internal fun <T> DataFrame<T>.parseImpl(options: ParserOptions?, columns: ColumnsSelector<T, Any?>): DataFrame<T> {

core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/impl/columns/constructors.kt

@@ -249,7 +249,9 @@ internal fun <T> createColumnGuessingType(
                 val df = dataFrameOf(values as Iterable<AnyCol>)
                 DataColumn.createColumnGroup(name, df)
             } else {
-                val df = values.map { (it as AnyRow?)?.toDataFrame() ?: DataFrame.empty(1) }.concat()
+                val df = values.map {
+                    (it as AnyRow?)?.toDataFrame() ?: DataFrame.empty(1)
+                }.concat()
                 DataColumn.createColumnGroup(name, df)
             }.asDataColumn().cast()
         }

core/src/test/kotlin/org/jetbrains/kotlinx/dataframe/io/ParserTests.kt

@@ -7,6 +7,7 @@ import kotlinx.datetime.LocalTime
 import kotlinx.datetime.TimeZone
 import kotlinx.datetime.toKotlinLocalDate
 import kotlinx.datetime.toKotlinLocalDateTime
+import org.jetbrains.kotlinx.dataframe.AnyFrame
 import org.jetbrains.kotlinx.dataframe.DataColumn
 import org.jetbrains.kotlinx.dataframe.DataFrame
 import org.jetbrains.kotlinx.dataframe.api.cast
@@ -16,11 +17,15 @@ import org.jetbrains.kotlinx.dataframe.api.convertToDouble
 import org.jetbrains.kotlinx.dataframe.api.convertToLocalDate
 import org.jetbrains.kotlinx.dataframe.api.convertToLocalDateTime
 import org.jetbrains.kotlinx.dataframe.api.convertToLocalTime
+import org.jetbrains.kotlinx.dataframe.api.first
+import org.jetbrains.kotlinx.dataframe.api.isEmpty
+import org.jetbrains.kotlinx.dataframe.api.isFrameColumn
 import org.jetbrains.kotlinx.dataframe.api.parse
 import org.jetbrains.kotlinx.dataframe.api.parser
 import org.jetbrains.kotlinx.dataframe.api.plus
 import org.jetbrains.kotlinx.dataframe.api.times
 import org.jetbrains.kotlinx.dataframe.api.tryParse
+import org.jetbrains.kotlinx.dataframe.columns.ColumnKind
 import org.jetbrains.kotlinx.dataframe.exceptions.TypeConversionException
 import org.junit.Test
 import java.math.BigDecimal
@@ -215,4 +220,17 @@ class ParserTests {
             Locale.setDefault(currentLocale)
         }
     }
+
+    @Test
+    fun `Issue #593, mixing null and json`() {
+        val col by columnOf("[\"str\"]", "[]", "null")
+        val parsed = col.parse()
+        parsed.type() shouldBe typeOf<AnyFrame>()
+        parsed.kind() shouldBe ColumnKind.Frame
+        require(parsed.isFrameColumn())
+
+        parsed[0]["value"].first() shouldBe "str"
+        parsed[1].isEmpty() shouldBe true
+        parsed[2].isEmpty() shouldBe true
+    }
 }