MS SQL Server: Upsert Mutation
Introduction
An upsert query ensures the given set of rows are present in the database, inserting new rows or updating existing rows as necessary, subject to certain criteria.
Convert insert mutation to upsert
Only tables with update permissions are upsertable. i.e. a table's update permissions are respected before updating an existing row in case of a match.
To convert an insert mutation into an upsert, you need to use the if_matched
argument to specify:
- the columns to be matched for each row using the
match_columnsfield. - the columns to be updated in the case of a match using the
update_columnsfield. - a condition for updating the column using the
wherefield, and
The value of the update_columns field determines the behavior of the upsert request as shown via the use cases below.
Upsert is not a substitute for update
The upsert functionality is sometimes confused with the update functionality. However, they work slightly differently.
An upsert mutation is used in the case when it's not clear if the respective row is already present in the database. If
it's known that the row is present in the database, update is the functionality to use.
For an upsert, all columns that are necessary for an insert are required.
How it works
- MS SQL Server tries to insert a row (hence all the required columns need to be present)
- If this fails because of some match, it updates the specified columns
If not all required columns are present, an error like NULL value unexpected for <not-specified-column> can occur.
Update selected columns on match
The update_columns field can be used to specify which columns to update in case a match occurs.
Example: Insert a new object in the article table or, if the value of the title column matches a a title value
in an existing row, update the content column of the existing article:
Note that the published_on column is left unchanged as it wasn't present in update_columns.
If match_columns is an empty array there is no basis for comparing new rows to existing rows. Thus the mutation
will always insert the values and will never update any rows.
Example: Insert a new object into the article table, will not match on columns because none are specified.
This query is equivalent to a simple insert without an if_matched clause.
Update selected columns on match subject to a filter
A where condition can be added to the if_matched clause to check a condition before making the update in case a
match occurs.
Example: Insert a new object in the article table, or if the value of the title column matches a a title value
in an existing row, update the published_on column specified in update_columns only if the previous published_on
value is lesser than the new value:
Ignore request on match
If update_columns is an empty array then on match the changes are ignored.
Example: Insert a new author object into the author table, unless there already exists an author with the same name.
In this case, the insert mutation is ignored because there is a match and update_columns is empty.
