如何检查列的数值是否包含通过SQL查询的字母

我在AWS S3中有一个CSV文件，正在加载到AWS Glue，即用于对来自S3的源数据文件应用转换。它提供了PySpark脚本环境。数据看起来有点像这样：如何检查列的数值是否包含通过SQL查询的字母

"ID","CNTRY_CD","SUB_ID","PRIME_KEY","DATE"  
"123","IND","25635525","11243749772","2017-10-17"  
"123","IND","25632349","112322abcd","2017-10-17"  
"123","IND","25635234","11243kjsd434","2017-10-17"  
"123","IND","25639822","1124374343","2017-10-17"

预期结果应该是这样的：

"123","IND","25632349","112322abcd","2017-10-17"  
"123","IND","25635234","11243kjsd434","2017-10-17"

在这里，我通过整型的名字“PRIME_KEY”工作的领域，可能包含英文字母，这导致数据格式不正确。

现在的需求是，我需要找出Integer类型的主键列是否包含任何使用SQL查询的数字值的字母数字字符。正则表达式到目前为止，我已经尝试了几个变种，以做到这一点像下面的一个，但没有运气：

args = getResolvedOptions(sys.argv, ['JOB_NAME']) 
glueContext = GlueContext(SparkContext.getOrCreate()) 
spark = glueContext.spark_session 
job = Job(glueContext) 
job.init(args['JOB_NAME'], args) 
# s3 output directory 
output_dir = "s3://aws-glue-scripts../.." 

# Data Catalog: database and table name 
db_name = "sampledb" 
glue_tbl_name = "sampleTable" 

datasource = glueContext.create_dynamic_frame.from_catalog(database = db_name, table_name = glue_tbl_name) 
datasource_df = datasource.toDF() 
datasource_df.registerTempTable("sample_tbl") 
invalid_primarykey_values_df = spark.sql("SELECT * FROM sample_tbl WHERE CAST(PRIME_KEY AS STRING) RLIKE '([a-z]+[0-9]+)|([0-9]+[a-z]+)'") 
invalid_primarykey_values_df.show()

这个脚本的输出如下：

SELECT * 
FROM table_name 
WHERE column_name IS NOT NULL AND 
CAST(column_name AS VARCHAR(100)) LIKE \'%[0-9a-z0-9]%\'

源脚本

+ --- + -------- + -------- + ------------ + ---------- + ----------- + --------------- +

+ --- + -------- + -------- + ------------ + ---------- + ----------- + --------------- +

| 123 | IND | 25635525 | [11243749772，null] | 2017-10-17 |

| 123 | IND | 25632349 | [null，112322ab .. | 2017-10-17 |

| 123 | IND | 25635234 | [null，11243kjsd .. | 2017-10-17 |

| 123 | IND | 25639822 | [1124374343，null] | 2017-10-17 |

+ -------- + -------- + -------------------- + ------ ---- + ----------- + --------------- +

我突出显示了我正在处理的字段的值。它看起来有点不同于源数据。

任何帮助，将不胜感激。由于

来源

2017-10-16 Arjun

您可以使用RLIKE

SELECT * 
FROM table_name 
WHERE CAST(PRIME_KEY AS STRING) RLIKE '([0-9]+[a-z]+)'

更通用的字母数字滤波器匹配。

WHERE CAST(PRIME_KEY AS STRING) RLIKE '([a-z]+[0-9]+)|([0-9]+[a-z]+)'

编辑：根据评论

必要的进口和UDF

val spark = SparkSession.builder 
    .config(conf) 
    .getOrCreate 

import org.apache.spark.sql.functions._ 
val extract_pkey = udf((x: String) => x.replaceAll("null|\\]|\\[|,", "").trim) 

import spark.implicits._

设置样本数据进行测试，并与UDF

val df = Seq(
    ("123", "IND", "25635525", "[11243749772,null]", "2017-10-17"), 
    ("123", "IND", "25632349", "[null,112322abcd]", "2017-10-17"), 
    ("123", "IND", "25635234", "[null,11243kjsd434]", "2017-10-17"), 
    ("123", "IND", "25639822", "[1124374343,null]", "2017-10-17") 
).toDF("ID", "CNTRY_CD", "SUB_ID", "PRIME_KEY", "DATE") 
    .withColumn("PRIME_KEY", extract_pkey($"PRIME_KEY")) 


df.registerTempTable("tbl") 

spark.sql("SELECT * FROM tbl WHERE PRIME_KEY RLIKE '([a-z]+[0-9]+)|([0-9]+[a-z]+)'") 
    .show(false) 

+---+--------+--------+------------+----------+ 
|ID |CNTRY_CD|SUB_ID |PRIME_KEY |DATE  | 
+---+--------+--------+------------+----------+ 
|123|IND  |25632349|112322abcd |2017-10-17| 
|123|IND  |25635234|11243kjsd434|2017-10-17| 
+---+--------+--------+------------+----------+

清洗

来源

2017-10-16 18:14:13 mrsrinivas

如何检查列的数值是否包含通过SQL查询的字母

回答

相关问题