UPDATE SELECT Statements
To perform multiple updates in a single request to Google BigQuery,first use the INSERT INTO syntax to insert a temporary table of data into Google BigQuery. This works by first populating a temporary table with the data you are going to submit to Google BigQuery. Once you have all of the data you want to update, use UPDATE SELECT FROM to pass the temporary table data into the table in Google BigQuery.
This functionality is also available via the standard Batch Processing API available in JDBC.
Populate the Temporary Table
The temporary table you are populating is dynamic and is created at run time the first time you insert to it. Temporary tables are denoted by a # appearing in their name. When using a temporary table to update, the temporary table must be named in the format [TableName]#TEMP, where TableName is the
name of the table you are inserting to. For example:
INSERT INTO [publicdata].[samples].github_nested#TEMP (Id, Name, MyCustomField__c) VALUES ('AX1000001', 'New github_nested', '9000'); INSERT INTO [publicdata].[samples].github_nested#TEMP (Id, Name, MyCustomField__c) VALUES ('AX1000002', 'New github_nested 2', '9001'); INSERT INTO [publicdata].[samples].github_nested#TEMP (Id, Name, MyCustomField__c) VALUES ('AX1000003', 'New github_nested 3', '9002');
This creates a temporary table called [publicdata].[samples].github_nested#TEMP with three columns and three rows of data. Since type cannot be determined on the temporary table itself, all values are stored in memory as strings. The values are later converted to the proper type when they are submitted to the [publicdata].[samples].github_nested table.
Update the Actual Table
Once your temporary table is populated, it is now time to update the actual table in Google BigQuery. You can do this by performing an UPDATE to the actual table and selecting the input data from the temporary table. For example:
UPDATE [publicdata].[samples].github_nested (Id, repository.name, MyCustomField__c) SELECT Id, repository.name, MyCustomField__c FROM [publicdata].[samples].github_nested#TEMPIn this example, the full contents of the [publicdata].[samples].github_nested#TEMP table are passed into the [publicdata].[samples].github_nested table. This results in fewer requests being submitted to Google BigQuery since multiple updates may be submitted with each request, which is much better for performance if you have many records to update.
The results of the query are stored in the LastResultInfo#TEMP temporary table. This table is cleared and repopulated the next time data is modified by passing in a temporary table. Please be aware that the LastResultInfo#TEMP table has no predefined schema. You need to check its metadata at run time before reading data.
Temporary Table Life Span
Temporary tables only last as long as the connection remains open. When the connection to Google BigQuery is closed, all temporary tables are cleared, including the LastResultInfo#TEMP table.