Decimal precision exceeds max precision despite decimal having the correct size and precision

I have this code

spark = (
    SparkSession.builder
    .appName("pyspark-sandbox")
    .getOrCreate()
)
spark.conf.set("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MICROS")  # type: ignore

faker = Faker()
value = faker.pydecimal(left_digits=28, right_digits=10)
print(value) # 8824750032877062776842530687.8719544506

df = spark.createDataFrame([[value]], schema=['DecimalItem'])
df = df.withColumn('DecimalItem', col('DecimalItem').cast(DecimalType(38, 10)))
df.show()

But on show I get this error

org.apache.spark.SparkArithmeticException: [DECIMAL_PRECISION_EXCEEDS_MAX_PRECISION] Decimal precision 46 exceeds max precision 38.

The value 8824750032877062776842530687.8719544506 seems to fit into DecimalType, yet it fails. What is the problem?

After some investigation, I found out that if you pass StructType schema, when creating a dataframe - it works properly without issues.

spark = (
    SparkSession.builder
    .appName("pyspark-sandbox")
    .getOrCreate()
)
spark.conf.set("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MICROS")  # type: ignore

faker = Faker()
value = faker.pydecimal(left_digits=28, right_digits=10)
print(value) # 8824750032877062776842530687.8719544506

df = spark.createDataFrame([[value]], schema=StructType([StructField('DecimalItem', DecimalType(38, 10), True)]))
df.show() 

Not sure why there is an issue with cast though. I guess you can't cast to a different DecimalType if Pyspark came up with its own?

Because if I check the schema using df.printSchema in my original approach

DecimalItem: decimal(38,18) (nullable = true)

it went with 38,18 by default.