json - Spark SQL DataFrame pretty-print

Question

我对 Scala 不是很好(我更喜欢 R 语言)我希望使用 Scala Spark 外壳。我已经尝试了 explode() 函数，但无法进一步...

scala> val sqlDf = spark.sql("select t.articles.donneesComptablesArticle.taxes from  dau_temp t")
sqlDf: org.apache.spark.sql.DataFrame = [taxes: array<array<struct<baseImposition:bigint,codeCommunautaire:string,codeNatureTaxe:string,codeTaxe:string,droitCautionnable:boolean,droitPercu:boolean,imputationCreditCautionne:boolean,montantLiquidation:bigint,quotite:double,statutAi2:boolean,statutDeLiquidation:string,statutRessourcesPropres:boolean,typeTaxe:string>>>]

scala> sqlDf.show
16/12/21 15:13:21 WARN util.Utils: Truncated the string representation of a plan since it was too large. This behavior can be adjusted by setting 'spark.debug.maxToStringFields' in SparkEnv.conf.
+--------------------+
|               taxes|
+--------------------+
|[WrappedArray([12...|
+--------------------+


scala> sqlDf.printSchema
root
 |-- taxes: array (nullable = true)
 |    |-- element: array (containsNull = true)
 |    |    |-- element: struct (containsNull = true)
 |    |    |    |-- baseImposition: long (nullable = true)
 |    |    |    |-- codeCommunautaire: string (nullable = true)
 |    |    |    |-- codeNatureTaxe: string (nullable = true)
 |    |    |    |-- codeTaxe: string (nullable = true)
 |    |    |    |-- droitCautionnable: boolean (nullable = true)
 |    |    |    |-- droitPercu: boolean (nullable = true)
 |    |    |    |-- imputationCreditCautionne: boolean (nullable = true)
 |    |    |    |-- montantLiquidation: long (nullable = true)
 |    |    |    |-- quotite: double (nullable = true)
 |    |    |    |-- statutAi2: boolean (nullable = true)
 |    |    |    |-- statutDeLiquidation: string (nullable = true)
 |    |    |    |-- statutRessourcesPropres: boolean (nullable = true)
 |    |    |    |-- typeTaxe: string (nullable = true)

scala> val sqlDfTaxes = sqlDf.select(explode(sqlDf("taxes")))
sqlDfTaxes: org.apache.spark.sql.DataFrame = [col: array<struct<baseImposition:bigint,codeCommunautaire:string,codeNatureTaxe:string,codeTaxe:string,droitCautionnable:boolean,droitPercu:boolean,imputationCreditCautionne:boolean,montantLiquidation:bigint,quotite:double,statutAi2:boolean,statutDeLiquidation:string,statutRessourcesPropres:boolean,typeTaxe:string>>]

scala> sqlDfTaxes.show()
16/12/21 15:22:28 WARN util.Utils: Truncated the string representation of a plan since it was too large. This behavior can be adjusted by setting 'spark.debug.maxToStringFields' in SparkEnv.conf.
+--------------------+
|                 col|
+--------------------+
|[[12564,B00,TVA,A...|
+--------------------+

“可读”内容如下所示(这是我的目标L:带有标题的经典行 x 列结构显示):

codeTaxe codeCommunautaire baseImposition quotite montantLiquidation statutDeLiquidation
A445               B00          12564    20.0               2513                   C
U165               A00          12000     4.7                564                   C
codeNatureTaxe typeTaxe statutRessourcesPropres statutAi2 imputationCreditCautionne
TVA    ADVAL                   FALSE      TRUE                     FALSE
DD    ADVAL                    TRUE     FALSE                      TRUE
droitCautionnable droitPercu
FALSE       TRUE
FALSE       TRUE

每一行的类是(使用 R 包 sparklyr 找到的):

<jobj[100]>
  class org.apache.spark.sql.catalyst.expressions.GenericRowWithSchema
  [12564,B00,TVA,A445,false,true,false,2513,20.0,true,C,false,ADVAL]

[[1]][[1]][[2]]
<jobj[101]>
  class org.apache.spark.sql.catalyst.expressions.GenericRowWithSchema
  [12000,A00,DD,U165,false,true,true,564,4.7,false,C,true,ADVAL]

Answer 1

你可以在每一列上展开:

val flattenedtaxes = sqlDf.withColumn("codeCommunautaire",  org.apache.spark.sql.functions.explode($"taxes. codeCommunautaire"))

在此之后，您的 flattenedtaxes 将有 2 列税(所有列都是原样)新列 codeCommunautaire