Add partition support to BigQuery

[dwmclary]

This adds the ability to create a partitioned table. It would be nice to expose a table.list_partitions() method, but currently there is not an API-based way of returning partitions. So, for now, it simply allows users to interact with the current partitioning API.

[tseaver]

The docs for creating partitioned tables point to a query for listing partitions, Could we use that to implement `table.list_partitions'?

[tseaver]

Thank you for the patch!

[dwmclary]

I've got a method I didn't include for list_partitions that does generate and run that query. What I'm not sure about is the best way to fake a partitioned table for testing said method.

[tseaver]

What I'm not sure about is the best way to fake a partitioned table for testing said method.

Assuming the method looks something like:

deflist_partitions(self):
query=self._client.query(
'SELECT partition_id from [%s.%s$__PARTITIONS_SUMMARY__]'%
(self.dataset_name, self.name))
query.run()
return [row[0] forrowinquery.rows]

I think I would just mock up the query method on the _Client mock to return an instance of a mocked _Query class: it would then have a run method which would populate its rows attribute.