Scala: Spark SQL to_date(unix_timestamp) returning NULL

问题

Spark Version: spark-2.0.1-bin-hadoop2.7 Scala: 2.11.8

I am loading a raw csv into a DataFrame. In csv, although the column is support to be in date format, they are written as 20161025 instead of 2016-10-25. The parameter date_format includes string of column names that need to be converted to yyyy-mm-dd format.

In the following code, I first loaded the csv of Date column as StringType via the schema, and then I check if the date_format is not empty, that is there are columns that need to be converted to Date from String, then cast each column using unix_timestamp and to_date. However, in the csv_df.show(), the returned rows are all null.

def read_csv(csv_source:String, delimiter:String, is_first_line_header:Boolean, 
    schema:StructType, date_format:List[String]): DataFrame = {
    println("|||| Reading CSV Input ||||")

    var csv_df = sqlContext.read
        .format("com.databricks.spark.csv")
        .schema(schema)
        .option("header", is_first_line_header)
        .option("delimiter", delimiter)
        .load(csv_source)
    println("|||| Successfully read CSV. Number of rows -> " + csv_df.count() + " ||||")
    if(date_format.length > 0) {
        for (i <- 0 until date_format.length) {
            csv_df = csv_df.select(to_date(unix_timestamp(
                csv_df(date_format(i)), "yyyy-­MM-­dd").cast("timestamp")))
            csv_df.show()
        }
    }
    csv_df
}

Returned Top 20 rows:

+-------------------------------------------------------------------------+
|to_date(CAST(unix_timestamp(prom_price_date, YYYY-­MM-­DD) AS TIMESTAMP))|
+-------------------------------------------------------------------------+
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
|                                                                     null|
+-------------------------------------------------------------------------+

Why am I getting all null?

回答1:

To convert yyyyMMdd to yyyy-MM-dd you can:

spark.sql("""SELECT DATE_FORMAT(
  CAST(UNIX_TIMESTAMP('20161025', 'yyyyMMdd') AS TIMESTAMP), 'yyyy-MM-dd'
)""")

with functions:

date_format(unix_timestamp(col, "yyyyMMdd").cast("timestamp"), "yyyy-MM-dd")

来源：https://stackoverflow.com/questions/40433065/scala-spark-sql-to-dateunix-timestamp-returning-null