Error in Databricks code?

[THIAM_HUATTAN]

https://www.databricks.com/notebooks/recitibikenycdraft/data-preparation.html

Could someone help to see in that Step 3: Prepare Calendar Info

# derive complete list of dates between first and last dates

dates = (

spark

.range(0,days_between).withColumnRenamed('id','days')

.withColumn('init_date', lit(first_date))

.selectExpr('cast(date_add(init_date, days) as timestamp) as date')

)

what does 'days' refer to in the last selectExpr sentence? it seems to be me not defined. is it meant to be 'days_between'? If i replace 'days' with 'days_between', it also breaks, because it expects a integer value, and not a variable.

Thanks.

[Harun]

@THIAM HUAT TAN​ 

You are renaming the "id" column to "days", that days column should have a integer value. and in the selectExpr you are doing a date addtion of init_date and the values from days column.

I have recreated the same with a sample data set.

Note : Kindly check whether the "id" column is a integer in your case.

[THIAM_HUATTAN]

thanks Harun for your example.. however, I am still stuck.. initially I have thought it is due to the date with time that interfere with the date_add, and I make it with to_date, however, the same error message appear.

With the next statement, it gives error below:

Appreciate if you could point me to the error. Thanks.

[THIAM_HUATTAN]

with this line added, then the error disappears 🙂

[Harun]

@THIAM HUAT TAN​ Kindly cast the "id" column to IntegerType like how i have casted for "arrival_date" column in my above sample record snapshot. That will make the code work.

[THIAM_HUATTAN]

yes, it should work similarly, thanks.

cast('int') vs cast(IntegerType()), I suppose both are identical?

[UmaMahesh1]

Hi @THIAM HUAT TAN​ 

In your notebook, you are creating a integer column days_between with the code

days_between = (last_date - first_date).days + 10

Logically speaking, what the nb trying to do is to fetch all the dates between two dates to do a forecast.

So, to get a complete list of dates between the start date and x number of days into the future (x = days_between), you are using spark.range(0,days_between). What is does is that it fetches a column of integers just like how python range works. The name of the column would be by default ''id".

So you are renaming this id column to "days" just for simplicity to denote that you are adding x days to another reference column (which again you are creating by using .withColumn('init_date', lit(first_date))) and creating a new column date which contains the date calculated.

The date calculated would be nothing but first_date + days.

Hope this helps...

Cheers..