[SPARK-54843][SQL] Try_to_number expression not working for empty string input

### What changes were proposed in this pull request?
Catching the case in `ToNumberParser` when the input string only consists of whitespace, preventing a failure with an internal error later on when trying to create `BigDecimal`.

### Why are the changes needed?
Without this change passing an empty string (`select try_to_number('', '99')`) would fail with the following exception:
```
JVM stacktrace:
java.lang.NumberFormatException
	at java.base/java.math.BigDecimal.(BigDecimal.java:692)
	at java.base/java.math.BigDecimal.(BigDecimal.java:471)
	at java.base/java.math.BigDecimal.(BigDecimal.java:900)
	at org.apache.spark.sql.catalyst.util.ToNumberParser.parseResultToDecimalValue(ToNumberParser.scala:627)
	at org.apache.spark.sql.catalyst.util.ToNumberParser.parse(ToNumberParser.scala:499)
```

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
New unit tests.

### Was this patch authored or co-authored using generative AI tooling?
No.

Closes #53609 from stefankandic/numFormatFix.

Authored-by: Stefan Kandic <[email protected]>
Signed-off-by: Wenchen Fan <[email protected]>
(cherry picked from commit df720c1)
Signed-off-by: Wenchen Fan <[email protected]>

Author: stefankandic

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/ToNumberParser.scala

@@ -495,6 +495,9 @@ class ToNumberParser(numberFormat: String, errorOnFail: Boolean) extends Seriali
       // If we have consumed all the tokens in the format string, but characters remain unconsumed
       // in the input string, then the input string does not match the format string.
       formatMatchFailure(input, numberFormat)
+    } else if (parsedBeforeDecimalPoint.isEmpty && parsedAfterDecimalPoint.isEmpty) {
+      // If no digits were collected (e.g. input was all whitespace), treat as format match failure.
+      formatMatchFailure(input, numberFormat)
     } else {
       parseResultToDecimalValue(negateResult)
     }

sql/core/src/test/scala/org/apache/spark/sql/StringFunctionsSuite.scala

@@ -1342,6 +1342,26 @@ class StringFunctionsSuite extends QueryTest with SharedSparkSession {
     checkAnswer(df.select(try_to_number($"value", lit("$99.99"))), Seq(Row(null)))
   }
 
+  test("try_to_number with whitespace-only input should return NULL") {
+    // Empty string
+    checkAnswer(sql("select try_to_number('', '99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number('', '999')"), Seq(Row(null)))
+
+    // Spaces only
+    checkAnswer(sql("select try_to_number('   ', '99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number(' ', '9')"), Seq(Row(null)))
+
+    // Different whitespace characters (tabs, newlines)
+    checkAnswer(sql("select try_to_number('\t\t', '99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number('\n\n', '99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number(' \t\n ', '99')"), Seq(Row(null)))
+
+    // With format strings containing decimal points, dollar signs, etc.
+    checkAnswer(sql("select try_to_number('   ', '$99.99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number('', '999.99')"), Seq(Row(null)))
+    checkAnswer(sql("select try_to_number('\t', '9,999')"), Seq(Row(null)))
+  }
+
   test("SPARK-44905: stateful lastRegex causes NullPointerException on eval for regexp_replace") {
     val df = sql("select regexp_replace('', '[a\\\\d]{0, 2}', 'x')")
     intercept[SparkRuntimeException](df.queryExecution.optimizedPlan)