Database Reference
In-Depth Information
$
bq show ch11.time_lapse
Table 317752944021:ch11.time_lapse
Last modified Schema Total Rows
Total Bytes Expi>
----------------- -------------------- ------------
------------- ----->
13 Nov 14:27:04 |- index: integer 10
160
|- millis: integer
$ bq show ch11.time_lapse@${AROUND_3}-${AROUND_7}
Table
317752944021:ch11.time_lapse@1384381608691-1384381619201
Last modified Schema Total Rows
Total Bytes Expi>
----------------- -------------------- ------------
------------- ----->
13 Nov 14:27:04 |- index: integer 4 64
|- millis: integer
The number of rows and bytes in the slice are a fraction of the rows and
bytes in the full table. In addition, a dry run query on a slice can be used
to probe how many bytes a query will scan. This feature is helpful for
applications that need to frequently query the most recently added data to a
table. They can avoid paying for the cost of scanning the full table by using
a table reference that limits the query to changes that occurred in the time
window of interest. The sample application uses this feature to generate
graphs of recent logs.
Table Recovery
When we discussed the table copy job we described how it could be used
to prevent corruption due to a bad load job. With time slices of tables, we
have a tool to recover from inadvertent table corruption even if no copy was
created prior to the load operation. The idea is straightforward: Copy the
table, as it existed prior to the job, to a new table. Then the corrupted table
can be deleted (or truncated) and replaced with the recovered table.
