Teradata numeric overflow for count validation

[oapa]

Encountered a Teradata numeric overflow error below when trying to run "count" validation on a table.  According to this post, it looks like count(*) can overflow on Teradata since it implicitly casts to INT.

pandas.io.sql.DatabaseError: Execution failed on sql 'SELECT count(*) AS "count", count("id") AS "count__id",
 ......
FROM {td_schema}.{td_table}
SAMPLE 10000': [Version 17.0.0.8] [Session 31075829] [Teradata Database] [Error 2616] Numeric overflow occurred during computation

[dhercher]

To summarize the issue which is occurring: Teradata implicitly casts any COUNT to INT. Running a normal count could be written more verbosely as
SELECT CAST(COUNT(*) AS INT) AS "count" ...

The overflow does not occur during the count, but rather during the cast which implicitly happens. There are Teradata configs which could override this behavior on a per installation basis (in TD15.10 there's a new dbscontrol flag to change the default to BIGINT or NUMBER).

We should likely automatically CAST all Teradata COUNT operations to BIGINT or NUMBER to avoid the error

[nehanene15]

This is now possible with the --cast-to-bigint flag and via YAML. This will cast any int32 columns to int64 before running an aggregation. Closed with PR #461