BigQuery: Client.query modifies job_config object

[tswast]

Steps to reproduce

1. Create a QueryJobConfig object.
2. Pass the object to the Client.query method.
3. Observe that the configuration object has changed.

Code example

from google.cloud import bigquery

client = bigquery.Client()
config = bigquery.QueryJobConfig(dry_run=True)
job = client.query(
 """
 SELECT name, SUM(number) AS total_people
 FROM `bigquery-public-data.usa_names.usa_1910_current`
 GROUP BY name
 """,
 job_config=config
)
print(job.total_bytes_processed)

config.dry_run = False
job = client.query(
 """
 SELECT name, SUM(number) AS total_people
 FROM `bigquery-public-data.usa_names.usa_1910_current`
 GROUP BY name
 """,
 job_config=config
)
job.result()

Stack trace

---------------------------------------------------------------------------
BadRequest Traceback (most recent call last)
<ipython-input-17-97f4f20faa6a> in <module>
----> 1 job.result()

~/miniconda3/envs/scratch/lib/python3.7/site-packages/google/cloud/bigquery/job.py in result(self, timeout, page_size, retry, max_results)
 2937 """
 2938 try:
-> 2939 super(QueryJob, self).result(timeout=timeout)
 2940 
 2941 # Return an iterator instead of returning the job.

~/miniconda3/envs/scratch/lib/python3.7/site-packages/google/cloud/bigquery/job.py in result(self, timeout, retry)
 732 self._begin(retry=retry)
 733 # TODO: modify PollingFuture so it can pass a retry argument to done().
--> 734 return super(_AsyncJob, self).result(timeout=timeout)
 735 
 736 def cancelled(self):

~/miniconda3/envs/scratch/lib/python3.7/site-packages/google/api_core/future/polling.py in result(self, timeout)
 125 # pylint: disable=raising-bad-type
 126 # Pylint doesn't recognize that this is valid in this case.
--> 127 raise self._exception
 128 
 129 return self._result

BadRequest: 400 Cannot explicitly modify anonymous table swast-scratch:_def2aa82a75fc33513bfb65968607e1f49148f83.anon561d4b35377f56c5960a4a49be253d3c90c120db

(job ID: 2692472c-1213-4716-8393-d9c49fd7b678)

 -----Query Job SQL Follows----- 

 | . | . | . | . | . |
 1:
 2: SELECT name, SUM(number) AS total_people
 3: FROM `bigquery-public-data.usa_names.usa_1910_current`
 4: GROUP BY name
 5: 
 | . | . | . | . | . |

When you run a job, the configuration object may be modified. In this example, it sets the destination table. It is unexpected that the original Python object would be modified.

I propose that the query (and other "job" methods on client) be updated to make a deep copy of the the job_config argument before passing it on to the job constructor.