Running Queries - Google BigQuery Analytics - page 267

… ).execute()['totalBytesProcessed']
u'552786'

With the Tables.get() method, you can see that the table has 42,522
rows:

>>> service.tables().get(
… projectId=project_id ,
… datasetId='reference' ,
… tableId='zip_codes').execute()['numRows']
u'42522'

If there were no STRING fields read by the query, the number of rows
would be enough to determine the number of bytes processed. However,
the state field has a variable length (it will be different in every row).
The number of bytes processed per row should be the length of the state
field plus 2 bytes, 8 bytes for the population field, and 1 byte for the
decommissioned field. The following query computes the cost by adding
up the field sizes:

>>> cost_query = """
… SELECT state_len + pop_len + decommissioned_len FROM
(
… SELECT SUM(LENGTH(state) + 2) AS state_len ,

… 8 * COUNT(population) AS pop_len ,

… COUNT(decommissioned) AS decommissioned_len

… FROM [bigquery-e2e:reference.zip_codes])

… """
>>> service.jobs().query(
… projectId=project_id ,
… body={'query': cost_query}
… ).execute()['rows'][0]['f'][0]['v']
u'552786'

This query returns the same value as the previous
totalBytesProcessed : 552,876—or just over one-half a megabyte. Of
course, running a query to compute how much a query costs is not
particularly efficient because that query will cost you money, too. What if
you want to find out how much a query will cost before actually running it?