Database Reference
In-Depth Information
In this case you can see that the job failed because you lack sufficient
permissions to create a table in the specified dataset.
To see what kind of data errors can be generated, there's an input file that is
riddled with errors. The following job configuration loads this file:
load_config['schema'] = {
'fields': [
{'name':'string_f', 'type':'STRING'},
{'name':'boolean_f', 'type':'BOOLEAN'},
{'name':'integer_f', 'type':'INTEGER',
'mode':'REQUIRED'},
{'name':'float_f', 'type':'FLOAT'},
{'name':'timestamp_f', 'type':'TIMESTAMP'}
]
}
load_config['sourceUris'] = [
'gs://bigquery-e2e/chapters/06/sample_bad.csv',
]
The example also changes the schema to include a field that is declared
a
REQUIRED
field to illustrate that an error is reported when no value is
supplied for a required field. When you run the sample that loads this data,
you can observe seven errors that look like this:
{
"reason": "invalid",
"message": "Could not interpret bool from string.",
"location": "File: 0 / " "Line:2 / Field:2"
}
The errors are reporting bad records that were rejected. The
reason
code is
always
invalid
and the
message
field describes the problem. In this case
you passed the value “nottrue” for a boolean field, which was not recognized.
These errors also contain a location entry describing where the error was
encountered. The location may refer to a specific field but in some cases—for
example, too many or too few columns—it just refers to the line or position
at which the error occurred. When the input file is processed in parallel, the
location will be reported as the byte offset of the start of the line rather than
