Glue data catalog schema not updated

Question

Hi everyone. 
Should crawler update table schema if datasourse schema is changed?
For example, I have some parquet file with data. One field has datatype "double".
Parquet file is created by Glue Job. I've updated Glue Job and set "decimal" instead of "double"
When I run job - it finishs with 'succeded' status. Then I run the crawler.
But the table still has "double" datatype.
Only if I delete the table and run crawler once again - it creates a new table with proper datatype = "decimal"
Is it correct behavior or I'm doing smth wrong?

Accepted Answer

Yes, IF you configured the crawler appropriately. See [https://docs.aws.amazon.com/glue/latest/dg/crawler-configuration.html](here) for details.

Also, it could be possible that you do NOT need to run the crawler separately, and instead update the catalog as a part of your pySpark script. See [https://docs.aws.amazon.com/glue/latest/dg/update-from-job.html]()

Answer

Decimal and double are not really compatible types on the way they are serialized, not even decimals with different precisions are.  
Whichever the crawler decides to use for the table, I think that table will be broken if you have a mixture of those types and some numbers will be incorrect.  
To do that kind of change you should recreate the table with the new type, after deleting the old files.
Also, if you generate the data in your own job, it should update the table during writing, no point on waiting for a crawler to make a guess when the job knows the schema.

Glue data catalog schema not updated

Relevant content