Dataproc：使用 PySpark 从 BigQuery 读取和写入数据时出现错误

我正在尝试读取一些 BigQuery 数据（ID：my-project.mydatabase.mytable[原始名称受保护]）来自用户管理的 Jupyter Notebook 实例，内部Dataproc https://cloud.google.com/dataproc?hl=es工作台。我正在尝试的灵感来自于this https://cloud.google.com/dataproc-serverless/docs/guides/bigquery-connector-spark-example?hl=en#submit_a_pyspark_wordcount_batch_workload，更具体地说，代码是（请阅读关于代码本身的一些附加注释）：

from pyspark.sql import SparkSession
from pyspark.sql.functions import udf, col
from pyspark.sql.types import IntegerType, ArrayType, StringType
from google.cloud import bigquery

# UPDATE (2022-08-10): BQ conector added
spark = SparkSession.builder.appName('SpacyOverPySpark') \
                    .config('spark.jars.packages', 'com.google.cloud.spark:spark-bigquery-with-dependencies_2.12:0.24.2') \
                    .getOrCreate()

# ------------------ IMPORTING DATA FROM BIG QUERY --------------------------

# UPDATE (2022-08-10): This line now runs...
df = spark.read.format('bigquery').option('table', 'my-project.mydatabase.mytable').load()

# But imports the whole table, which could become expensive and not optimal
print("DataFrame shape: ", (df.count(), len(df.columns)) # 109M records & 9 columns; just need 1M records and one column: "posting"

# I tried the following, BUT with NO success:
# sql = """
# SELECT `posting`
# FROM `mentor-pilot-project.indeed.indeed-data-clean`
# LIMIT 1000000
# """
# df = spark.read.format("bigquery").load(sql)
# print("DataFrame shape: ", (df.count(), len(df.columns)))

# ------- CONTINGENCY PLAN: IMPORTING DATA FROM CLOUD STORAGE ---------------

# This section WORKS (just to enable the following sections)
# HINT: This dataframe contains 1M rows of text, under a single column: "posting"
df = spark.read.csv("gs://hidden_bucket/1M_samples.csv", header=True)

# ---------------------- EXAMPLE CUSTOM PROCESSING --------------------------

# Example Python UDF Python
def split_text(text:str) -> list:
    return text.split()

# Turning Python UDF into Spark UDF
textsplitUDF = udf(lambda z: split_text(z), ArrayType(StringType()))

# "Applying" a UDF on a Spark Dataframe (THIS WORKS OK)
df.withColumn("posting_split", textsplitUDF(col("posting")))

# ------------------ EXPORTING DATA TO BIG QUERY ----------------------------

# UPDATE (2022-08-10) The code causing the error:

# df.write.format('bigquery') \
#   .option('table', 'wordcount_dataset.wordcount_output') \
#   .save()

# has been replace by a code that successfully stores data in BQ:

df.write \
  .format('bigquery') \
  .option("temporaryGcsBucket", "my_temp_bucket_name") \
  .mode("overwrite") \
  .save("my-project.mynewdatabase.mytable")

使用 SQL 查询从 BigQuery 读取数据时，触发的错误为：

Py4JJavaError: An error occurred while calling o195.load.
: com.google.cloud.spark.bigquery.repackaged.com.google.inject.ProvisionException: Unable to provision, see the following errors:

1) Error in custom provider, java.lang.IllegalArgumentException: 'dataset' not parsed or provided.
  at com.google.cloud.spark.bigquery.SparkBigQueryConnectorModule.provideSparkBigQueryConfig(SparkBigQueryConnectorModule.java:65)
  while locating com.google.cloud.spark.bigquery.SparkBigQueryConfig

1 error
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InternalProvisionException.toProvisionException(InternalProvisionException.java:226)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InjectorImpl$1.get(InjectorImpl.java:1097)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InjectorImpl.getInstance(InjectorImpl.java:1131)
    at com.google.cloud.spark.bigquery.BigQueryRelationProvider.createRelationInternal(BigQueryRelationProvider.scala:75)
    at com.google.cloud.spark.bigquery.BigQueryRelationProvider.createRelation(BigQueryRelationProvider.scala:46)
    at org.apache.spark.sql.execution.datasources.DataSource.resolveRelation(DataSource.scala:332)
    at org.apache.spark.sql.DataFrameReader.loadV1Source(DataFrameReader.scala:242)
    at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:230)
    at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:197)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)
    at py4j.Gateway.invoke(Gateway.java:282)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:238)
    at java.lang.Thread.run(Thread.java:750)
Caused by: java.lang.IllegalArgumentException: 'dataset' not parsed or provided.
    at com.google.cloud.bigquery.connector.common.BigQueryUtil.lambda$parseTableId$2(BigQueryUtil.java:153)
    at java.util.Optional.orElseThrow(Optional.java:290)
    at com.google.cloud.bigquery.connector.common.BigQueryUtil.parseTableId(BigQueryUtil.java:153)
    at com.google.cloud.spark.bigquery.SparkBigQueryConfig.from(SparkBigQueryConfig.java:237)
    at com.google.cloud.spark.bigquery.SparkBigQueryConnectorModule.provideSparkBigQueryConfig(SparkBigQueryConnectorModule.java:67)
    at com.google.cloud.spark.bigquery.SparkBigQueryConnectorModule$$FastClassByGuice$$db983008.invoke(<generated>)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.ProviderMethod$FastClassProviderMethod.doProvision(ProviderMethod.java:264)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.ProviderMethod.doProvision(ProviderMethod.java:173)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InternalProviderInstanceBindingImpl$CyclicFactory.provision(InternalProviderInstanceBindingImpl.java:185)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InternalProviderInstanceBindingImpl$CyclicFactory.get(InternalProviderInstanceBindingImpl.java:162)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.ProviderToInternalFactoryAdapter.get(ProviderToInternalFactoryAdapter.java:40)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.SingletonScope$1.get(SingletonScope.java:168)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InternalFactoryToProviderAdapter.get(InternalFactoryToProviderAdapter.java:39)
    at com.google.cloud.spark.bigquery.repackaged.com.google.inject.internal.InjectorImpl$1.get(InjectorImpl.java:1094)
    ... 18 more

向BigQuery写入数据时，出现错误：

Py4JJavaError: An error occurred while calling o167.save.
: java.lang.ClassNotFoundException: Failed to find data source: bigquery. Please find packages at http://spark.apache.org/third-party-projects.html

UPDATE:(2022-09-10) 向BigQuery写入数据时出错的问题已经解决，请参考上面的代码以及下面的评论部分。

我究竟做错了什么？

讨论中发现的要点：

1. 通过以下方式将 BigQuery 连接器添加为依赖项spark.jars=<gcs-uri> or spark.jars.packages=com.google.cloud.spark:spark-bigquery-with-dependencies_<scala-version>:<version>.

2. 指定正确的表名<project>.<dataset>.<table> format.

3. 数据帧写入器的默认模式是errorifexists。当写入不存在的表时，数据集必须存在，该表将自动创建。写入现有表时，模式需要设置为"append" or "overwrite" in df.write.mode(<mode>)...save().

4. 写入 BQ 表时，执行以下任一操作

   a) 直接写入（自支持）0.26.0 https://mvnrepository.com/artifact/com.google.cloud.spark/spark-bigquery-with-dependencies_2.12/0.26.0)

   df.write \
     .format("bigquery") \
     .option("writeMethod", "direct") \
     .save("dataset.table")
   

   b) 或间接写

   df.write \
     .format("bigquery") \
     .option("temporaryGcsBucket","some-bucket") \
     .save("dataset.table")
   

   看到这个doc https://github.com/GoogleCloudDataproc/spark-bigquery-connector#writing-data-to-bigquery.

5. 通过 SQL 查询从 BigQuery 读取数据时，添加强制属性viewsEnabled=true and materializationDataset=<dataset>:

   spark.conf.set("viewsEnabled","true")
   spark.conf.set("materializationDataset","<dataset>")
   
   sql = """
     SELECT tag, COUNT(*) c
     FROM (
       SELECT SPLIT(tags, '|') tags
       FROM `bigquery-public-data.stackoverflow.posts_questions` a
       WHERE EXTRACT(YEAR FROM creation_date)>=2014
     ), UNNEST(tags) tag
     GROUP BY 1
     ORDER BY 2 DESC
     LIMIT 10
     """
   df = spark.read.format("bigquery").load(sql)
   df.show()
   

   看到这个doc https://github.com/GoogleCloudDataproc/spark-bigquery-connector#reading-data-from-a-bigquery-query.