Hi, I'm working on a SAS migration project - migrating codes from SAS 9.3 to SAS 9.4 (SAS studio). In SAS 9.3 the final SAS datasets from the SAS codes are used for data warehousing where as in SAS 9.4 the data warehouse is Redshift. We noticed the result sets produced by left join (sas to sas datasets, sas to redshift tables) has different sort order in SAS 9.3 and SAS 9.4 when we don't have a condition or keys in place to control the entire row values. When we remove duplicates by keys in PROC SORT after left join, it eliminates different records because the left join sorts the result set implicitly in different order within duplicate keys. As a result, PROC SORT picks different records in both 9.3 and 9.4 and we see data mismatch between 9.3 and 9.4 in the final datasets. The client doesn't want to make any changes in 9.3 legacy codes to control the record order. So we tried finding patterns in the left join sort order in 9.3 and 9.4 and it looks like they are sorting them randomly within duplicate keys. Did anyone encounter this issue in your migration project? Is this something expected because I never thought left join in 9.3 and 9.4 works differently?
... View more