Don't return query results from BigQuery DML/DDL statements

[nfallen]

This client library uses tabledata.list to fetch query results, and so even for a DML/DDL statement it fetches rows. Instead, you could use the jobs.getQueryResults API, which correctly does not return any rows for DML/DDL.

[quartzmo]

@nfallen Thank you for bringing up this issue. This library originally used getQueryResults, but was changed last year to use tabledata.list for better performance. I'm not sure if the performance difference still exists, however. I'll see what I can find out.

How many rows are you seeing?

[nfallen]

If you want to continue using tabledata.list for performance reasons you could use the presence of the field statistics.query.numDmlAffectedRows in the jobs.get response to indicate that the query was DML, and the presence of statistics.query.ddlTargetTable to indicate that the query was DDL.

Since there is no maxResults value provided by default the call to tabledata.list will attempt to fetch all rows, and will only be limited by the size of the table or the maximum response size of tabledata.list.

[quartzmo]

Thanks @nfallen, this is very helpful!

[quartzmo]

@nfallen I believe the library actually provides a way to execute DML/DDL without fetching rows, as demonstrated in dataset_ddl_test.rb.

There is no test for DML operations, but I believe it would work similarly. The library does not currently expose statistics.query.numDmlAffectedRows, but I can add it, along with the test.

Does this sound OK to you?