Running Queries - Google BigQuery Analytics - page 243

More than 98 percent of BigQuery queries return in less than 3 seconds
(obviously this is highly dependent on the query being run); if you set the
timeoutMs value to a large value, you're almost certain to get a response
without a timeout. Queries that take longer than a few seconds are generally
complicated with lots of JOIN or GROUP BY operations; simpler queries
should always complete within a reasonable timeout, unless they are
operating over truly massive amounts of data.
When Jobs.query() times out, it doesn't return an HTTP response error.
Instead, it returns a normal query response, but the jobComplete flag will
be set to false . It also returns a jobId for the Query job. You can use
this jobId to call Jobs.getQueryResults() to both wait for the job to
complete and return the results.
You might wonder why the Jobs.query() request is a POST request and
not a GET request. After all, a number of systems that you might want to
use to run queries support only GET requests, and there is a convenient
standard “ q=<url encoded query> ” syntax that you could use. One
technical issue is that GET requests require passing all the data in the URL,
and many web-capable systems have a limit on the length of URL they
can use in a GET request. Internet Explorer has a hard-limit for URLs at
approximately 2048 characters; Microsoft Excel limits URL query strings to
256 characters. Many SQL queries in BigQuery are several kilobytes long;
any system with a URL length limit would be unable to issue a considerable
proportion of BigQuery queries.