Batch Request to Google API making calls before the batch executes in Python? - api

Question

I am working on sending a large batch request to the Google API through Admin SDK which will add members to certain groups based upon the groups in our in-house servers(a realignment script). I am using Python and to access the Google API I am using the [apiclient library][1]. When I create my service and batch object, the creation of the service object requests a URL.
batch_count = 0
batch = BatchHttpRequest()
service = build('admin', 'directory_v1')
logs
INFO:apiclient.discovery:URL being requested:
https://www.googleapis.com/discovery/v1/apis/admin/directory_v1/rest
which makes sense as the JSON object returned by that HTTP call is used to build the service object.
Now I want to add multiple requests into the batch object, so I do this;
for email in add_user_list:
if batch_count != 999:
add_body = dict()
add_body[u'email'] = email.lower()
for n in range(0, 5):
try:
batch.add(service.members().insert(groupKey=groupkey, body=add_body), callback=batch_callback)
batch_count += 1
break
except HttpError, error:
logging_message('Quota exceeded, waiting to retry...')
time.sleep((2 ** n)
continue
Every time it iterates through the outermost for loop it logs (group address redacted)
INFO:apiclient.discovery:URL being requested:
https://www.googleapis.com/admin/directory/v1/groups/GROUPADDRESS/members?alt=json
Isn't the point of the batch to not send out any calls to the API until the batch request has been populated with all the individual requests? Why does it make the call to the API shown above every time? Then when I batch.execute() it carries out all the requests (which is the only area I thought would be making calls to the API?)
For a second I thought that the call was to construct the object that is returned by .members() but I tried these changes:
service = build('admin', 'directory_v1').members()
batch.add(service.insert(groupKey=groupkey, body=add_body), callback=batch_callback)
and got the same result still. The reason this is an issue is because it is doubling the number of requests to the API and I have real quota concerns at the scale this is running at.