Documentation

Normalize a row by adding missing columns with DEFAULT value and sort by column name to make sure all rows are consistent in column values and order.

Example: A table "author" is defined as:

CREATE TABLE author ([id] INTEGER NOT NULL PRIMARY KEY, name TEXT NOT NULL, age INTEGER)

Consider the following mutation:

mutation {
  insert_author(
    objects: [{id: 1, name: "Foo", age: 21}, {id: 2, name: "Bar"}]
  ){
    affected_rows
  }
}

We consider DEFAULT value for age column which is missing in second insert row.

The corresponding INSERT statement looks like:

INSERT INTO author (id, name, age)
  OUTPUT INSERTED.id
  VALUES (1, 'Foo', 21), (2, 'Bar', DEFAULT)

Construct a MERGE statement from AnnotatedInsert information. A MERGE statement is responsible for actually inserting and/or updating the data in the table.

As part of an INSERT/UPSERT process, insert VALUES into a temporary table. The content of the temporary table will later be inserted into the original table using a MERGE statement.

We insert the values into a temporary table first in order to replace the missing fields with DEFAULT in normalizeInsertRows, and we can't do that in a MERGE statement directly.