如何解决 com.mongodb.spark.exceptions.MongoTypeConversionException：无法转换... Java Spark

Question

您好，我是 Java Spark 的新手，这几天一直在寻找解决方案。

我正在将 MongoDB 数据加载到 hive 表中，但是，我在 saveAsTable 时发现了一些错误，出现此错误

com.mongodb.spark.exceptions.MongoTypeConversionException: Cannot cast STRING into a StructType(StructField(oid,StringType,true)) (value: BsonString{value='54d3e8aeda556106feba7fa2'})

我尝试增加 sampleSize，不同的 mongo-spark-connector 版本，...但没有工作解决方案。

我无法弄清楚根本原因是什么，以及需要做的差距是什么？

最令人困惑的部分是我有类似的数据集，使用相同的流程没有问题。

mongodb 数据模式就像嵌套的结构和数组

root
 |-- sample: struct (nullable = true)
 |    |-- parent: struct (nullable = true)
 |    |    |-- expanded: array (nullable = true)
 |    |    |    |-- element: struct (containsNull = true)
 |    |    |    |    |-- distance: integer (nullable = true)
 |    |    |    |    |-- id: struct (nullable = true)
 |    |    |    |    |    |-- oid: string (nullable = true)
 |    |    |    |    |-- keys: array (nullable = true)
 |    |    |    |    |    |-- element: string (containsNull = true)
 |    |    |    |    |-- name: string (nullable = true)
 |    |    |    |    |-- parent_id: array (nullable = true)
 |    |    |    |    |    |-- element: struct (containsNull = true)
 |    |    |    |    |    |    |-- oid: string (nullable = true)
 |    |    |    |    |-- type: string (nullable = true)
 |    |    |-- id: array (nullable = true)
 |    |    |    |-- element: struct (containsNull = true)
 |    |    |    |    |-- oid: string (nullable = true)

样本数据

    "sample": {
      "expanded": [
        {
          "distance": 0,
          "type": "domain",
          "id": "54d3e17b5cf737074d4065b0",
          "parent_id": [
            "54d3e1775cf737074d406599"
          ],
          "name": "level2"
        },
        {
          "distance": 1,
          "type": "domain",
          "id": "54d3e1775cf737074d406599",
          "name": "level1"
        }
      ],
      "id": [
        "54d3e17b5cf737074d4065b0"
      ]
    }

示例代码

public static void main(final String[] args) throws InterruptedException {
    // spark session read mongodb
    SparkSession mongo_spark = SparkSession.builder()
            .master("local")
            .appName("MongoSparkConnectorIntro")
            .config("mongo_spark.master", "local")
            .config("spark.mongodb.input.uri", "mongodb://localhost:27017/test_db.test_collection")
            .enableHiveSupport()
            .getOrCreate();

    // Create a JavaSparkContext using the SparkSession's SparkContext object
    JavaSparkContext jsc = new JavaSparkContext(mongo_spark.sparkContext());

    // Load data and infer schema, disregard toDF() name as it returns Dataset
    Dataset<Row> implicitDS = MongoSpark.load(jsc).toDF();
    implicitDS.printSchema();
    implicitDS.show();

    // createOrReplaceTempView to see if the data being read
    // implicitDS.createOrReplaceTempView("my_table");
    // implicitDS.printSchema();
    // implicitDS.show();

    // saveAsTable
    implicitDS.write().saveAsTable("my_table");
    mongo_spark.sql("SELECT * FROM my_table limit 1").show();

    mongo_spark.stop();
}

如果有人有一些想法，我将非常感激。 谢谢

Answer 1

随着我适当增加样本量，这个问题不再存在了。

How to config Java Spark sparksession samplesize

Answer 2

我遇到了同样的问题，sampleSize 部分解决了这个问题，但如果您有大量数据，则无法解决。

这是解决此问题的方法。将此方法与增加的 sampleSize（在我的情况下为 100000）一起使用：

def fix_schema(schema: StructType) -> StructType:
    """Fix spark schema due to inconsistent MongoDB schema collection.

    It fixes such issues like:
        Cannot cast STRING into a NullType
        Cannot cast STRING into a StructType

    :param schema: a source schema taken from a Spark DataFrame to be fixed
    """
    if isinstance(schema, StructType):
        return StructType([fix_schema(field) for field in schema.fields])
    if isinstance(schema, ArrayType):
        return ArrayType(fix_schema(schema.elementType))
    if isinstance(schema, StructField) and is_struct_oid_obj(schema):
        return StructField(name=schema.name, dataType=StringType(), nullable=schema.nullable)
    elif isinstance(schema, StructField):
        return StructField(schema.name, fix_schema(schema.dataType), schema.nullable)
    if isinstance(schema, NullType):
        return StringType()
    return schema


def is_struct_oid_obj(struct_field: StructField) -> bool:
    """
    Checks that our schema has StructType field with single oid name inside

    :param struct_field: a StructField from Spark schema
    :return bool
    """
    return (isinstance(struct_field.dataType, StructType)
            and len(struct_field.dataType.fields) == 1
            and struct_field.dataType.fields[0].name == "oid")