Converting text to date in ETL

rollthejewels

Hi all,

 

I'm trying to convert a text into a date field in magic ETL, and I'm running into errors and unexpected outputs.  Screenshots below.

When trying to just convert the (numbers as text) into dates, I'm receiving the error "An invalid date was found".  When I conver that column first from text to numbers there's no issue, but then when I go to convert that new column to dates, I'm receiving all dates as 1/1/1970.

I've already accounted for non-numerical text fields in the filters previously.

Any help would be appreciated.

 

Structure of Data

Output of changing the above field to date

 

Output of changing field to number first then to date

 

MarkSnodgrass

Ok. This should work for you. I've added some additional columns just for learning purposes so you can see how it is put together. The column that I named NewDate is really the only column that you would need in your output. I've added some comments to explain what each part is doing.

SELECT "datelen",
--Gets the year
RIGHT("datelen",4) Yr,
--Gets the day
LEFT(RIGHT("datelen",6),2) Dy,
--Gets the month
--check to see if this is a 2-digit month or not
CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	--add a leading zero for one-digit month
	LPAD(LEFT("datelen",1),2,'0')
END Mo,
--concatenate the parts together to a convertible format
CONCAT(
  CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	LPAD(LEFT("datelen",1),2,'0')
END,
CONCAT(LEFT(RIGHT("datelen",6),2),RIGHT("datelen",4))) FullDate,
--put it all together and convert it to a date format
to_Date(
CONCAT(
  CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	LPAD(LEFT("datelen",1),2,'0')
END,
CONCAT(LEFT(RIGHT("datelen",6),2),RIGHT("datelen",4))),'MMDDYYYY') NewDate

FROM "df_ga_int_emarsys_campaign_attribution_sql_prep"

Hopefully, this works for your data, but it depends if there is some bad data that isn't taken into account.

MarkSnodgrass

I seem to recall running into a similar issue and I could never get the ETL to cooperate. I ending up doing a SQL dataflow to get the dates in an actual date format and then used that output dataset as my new input dataset in the ETL. Here is an example of what I did in my SQL dataflow 

CASE WHEN "EXTRACTDTE" < 19000101 THEN NULL ELSE date(to_char(TRIM("EXTRACTDTE"),'99999999')) END AS DateEXTRACTDTE

I had about 10 different date fields in my dataset and I applied this same logic to each field. 

Hope this helps.

 

rollthejewels

Thanks - I'm trying that and running into an error:

The database reported a syntax error: FUNCTION to_char does not exist

See attached screenshot.

 

 

MarkSnodgrass

It looks like I was using the Redshift SQL for this conversion. Do you have the option to use that? If not, we should be able to track down the equivalent functions in MySQL.

rollthejewels

Using redshift, this is my code:

SELECT
`Campaign`,
`Date`,
`Revenue`,
CASE WHEN `DateLen` < 19000101 THEN NULL ELSE date(to_char(TRIM(`DateLen`), "99999999")) END AS DateEXTRACTDTE
FROM
`df_ga_int_emarsys_campaign_attribution_sql_prep`

---

And I'm getting the error:  

The database reported a syntax error: [Amazon](500310) Invalid operation: syntax error at or near "`" Position: 231;

Thanks for your help, I'm not very familiar with using these types of dataflows.

MarkSnodgrass

Redshift and MySQL have different preferences on their use of single quotes and double quotes. 

Try this:

SELECT
"campaign",
"date",
"revenue",
CASE WHEN "datelen" < 19000101 THEN NULL ELSE date(to_char(TRIM("datelen"), '99999999')) END AS DateEXTRACTDTE
FROM
"df_ga_int_emarsys_campaign_attribution_sql_prep"

rollthejewels

The SQL worked but that field just came back blank...

MarkSnodgrass

I would say that means it is falling into the first part of the case statement and it thinks the values are less than 19000101. Can you add another screenshot that shows the datelen column without any logic applied to it?

rollthejewels

MarkSnodgrass

Ok. This should work for you. I've added some additional columns just for learning purposes so you can see how it is put together. The column that I named NewDate is really the only column that you would need in your output. I've added some comments to explain what each part is doing.

SELECT "datelen",
--Gets the year
RIGHT("datelen",4) Yr,
--Gets the day
LEFT(RIGHT("datelen",6),2) Dy,
--Gets the month
--check to see if this is a 2-digit month or not
CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	--add a leading zero for one-digit month
	LPAD(LEFT("datelen",1),2,'0')
END Mo,
--concatenate the parts together to a convertible format
CONCAT(
  CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	LPAD(LEFT("datelen",1),2,'0')
END,
CONCAT(LEFT(RIGHT("datelen",6),2),RIGHT("datelen",4))) FullDate,
--put it all together and convert it to a date format
to_Date(
CONCAT(
  CASE WHEN LENGTH("datelen") = 8 THEN
	LEFT("datelen",2) 
ELSE 
	LPAD(LEFT("datelen",1),2,'0')
END,
CONCAT(LEFT(RIGHT("datelen",6),2),RIGHT("datelen",4))),'MMDDYYYY') NewDate

FROM "df_ga_int_emarsys_campaign_attribution_sql_prep"

Hopefully, this works for your data, but it depends if there is some bad data that isn't taken into account.

rollthejewels

Works perfectly, thank you for your help!