Database Reference
In-Depth Information
In addition to data, tables have some additional interesting metadata. A
table can specify an expiration time, which enables it to be automatically
garbage collected after it expires. Tables also have a row count and a last
modified time, which can be used to tell whether they have been updated
recently.
Datasets
BigQuery datasets are collections of tables. They are logical groupings of
tables, which can be thought of as similar to a database. Unlike databases,
however, a dataset doesn't provide any indication of where or how the
underlying tables are stored, and there are no restrictions about combining
data (via copy, the SQL
JOIN
operation, and so on) in different datasets.
Dataset Access Control
Datasets are the primary unit of sharing in BigQuery because they control
access to all underlying tables. Datasets have ACLs that specify readers,
writers, and owners for the dataset. Readers are allowed to read data from
tables in a dataset and run queries against those tables; writers have the
same permissions as readers but are also allowed to create and modify
tables; and owners have the same permissions as writers but are also
allowed to change the ACL on the dataset.
By default, the ACL on a dataset specifies only the project permissions set
up in the Google Developers Console. The project owners map to dataset
owners; project editors map to dataset writers; and project viewers map
to dataset readers. These ACLs can be changed, however, and the project
permissions entries can be removed from the dataset. As previously
mentioned, project owners have some special rights. Because the project
owners pay for the dataset, they always have the option to delete it. This
prevents people from having a dataset they are obliged to pay for but cannot
remove.
If a user is not a member of the project team, in order to run queries against
tables in a dataset, they need to be a member of some other project team
that will get billed for their queries. If you allow a user to read your tables,
this will not give them permission to do anything that may cost you money;
for them to do that, they must be a member of your project's team.
