Skip to content

UPDATE FROM SELECT Statement

Description

The UPDATE FROM SELECT statement updates data in existing table. The data comes from a subselect query. It can be used as alternative to 'create table' and 'insert into' for store predictions in distinct columns of existing rows

Syntax

Here is an example:

UPDATE 
    int2.table2 
SET
    predicted = source.result, 
FROM   
 (  
     SELECT p.result, p.prod_id, p.shop_id   
      FROM int1.table1 as t 
     JOIN mindsdb.pred1 as p
 ) AS source 
WHERE 
    prod_id = source.prod_id 
    and shop_id = source.shop_id

And the steps followed by the syntax:

  • It executes query from 'FROM' block to get the output dataset. In our example it is join of table table1 (from integration int1) with predictor pred1. It also can be select from integration
  • source is the alias for fetched data
  • then it updates table2 from int2 using conditions from WHERE block and fields for update from SET block
  • under the hood it splits input data to rows and execute this query for every row: 
    UPDATE 
    table2 
SET
    predicted = <row.result>, 
WHERE 
    prod_id = <row.prod_id> 
    and shop_id = <row.shop_id>

Note: in WHERE block it is better to use primary key for table or set of rows that can be a primary key (and identifies every row in table). Otherwise, it can lead to unexpected results when one row in destination table was updated several times from different rows in source table (because conditions from different rows are fit).