How to convert column type from str to date in sparksql when the format is not yyyy-mm-dd?

[semihcandoken]

I imported a large csv file into databricks as a table.

I am able to run sql queries on it in a databricks notebook.

In my table, I have a column that contains date information in the mm/dd/yyyy format :

12/29/2015

12/30/2015 etc...

Databricks imported this column with type str, instead of date. Forcing a 'timestamp' type in the Table UI did not have any effect.

How can I convert this column type to a date inside sql?

I tried to do

select cast(arrival_date as date) from my_data_table

however, this requires that the str column is in YYYY-mm-dd format. And mine is mm/dd/yyyy format as mentioned above.

select to_date('15/1/09') as date;

does not work either for the same reason.

What can I do to have a column of dates?

Is it true that SparkSQL does not support 'update' operations? In that case, I cannot rearrange the string to fit the format either? What options do I have?

[JoeConley]

The following worked for me:

df.withColumn("tx_date", to_date(unix_timestamp($"date", "M/dd/yyyy").cast("timestamp")))

[Bill_Chambers]

also to_date($"date", "format") works.

[shalli]

I tried the above problem with a string and solved it as below:

val df = sc.parallelize(Seq("08-26-2016")).toDF("Id")

df.createOrReplaceTempView("table1")

val bdf = spark.sql("""select from_unixtime(unix_timestamp(Id, 'MM-dd-yyyy')) as new_format from table1""")

bdf.printSchema

bdf.show

val bbdf = bdf.withColumn("dt",$"new_format".cast("date"))

bbdf.printSchema

bbdf.show

bbdf.select(year($"dt")).show

[ShubhamGupta187]

@josephpconley would it be safe to cast a column that contains null values?