How to add a check constraint in sql azure data warehouse?

Question

I'd like to add a check constraint to a table in azure data warehouse.

I have the below script:

alter table [FactTbl_Test] add constraint Ck_Test_id check ([Test_Id] != 'Unknown')

This causes the below error:

Parse error at line: 2, column: 74: Incorrect syntax near 'check'.

Edit:

Also appears to fail with the same error if the constraint is created inside the table declaration.

Answer 1

This and constraints in general are currently unsupported in Azure SQL Data Warehouse. See https://docs.microsoft.com/en-us/azure/sql-data-warehouse/sql-data-warehouse-tables-overview#unsupported-table-features for full list.

If the constraint is not enforced at the source or if you have multiple sources, you can implement during the ETL or ELT process. For small datasets, ETL would probably work better. With large datasets, you're likely to get better performance via ELT-L. As in extract then load to a temp table (heap is fast), perform the check/cleansing then load to the final table.

Answer 2

The documentation of the CREATE TABLE syntax for Azure SQL Data Warehouse does not mention CHECK constraints at all, nor any other type of constraints except for NOT NULL and DEFAULT constraints.

I assume that CHECK constraints are not supported in this product.

Simple talk article: Taking Azure SQL Data Warehouse for a Test-Drive by Robert Sheldon, says it clearly:

Working with tables

When working with the SQL Data Warehouse version of a T-SQL statement, you’ll often find that the syntax is similar to how the statement is implemented in SQL Server or SQL Database. However, the SQL Data Warehouse version of a statement is usually not as extensive and sometimes supports options specific to SQL Data Warehouse, as we saw with the CREATE DATABASE statement.

The same can be said of the CREATE TABLE statement. It does not support many options available to SQL Server or SQL Database, and there are a couple options specific to SQL Data Warehouse. For example, the statement does not support primary keys, foreign keys, check constraints, unique constraints, unique indexes, computed columns, sparse columns, user-defined data types, indexed views, identities, sequences, triggers, or synonyms. The statement also does not support certain data types, such as geometry, geography, hierarchyid, and so on. At the same time, the CREATE TABLE statement provides table options that are unique to SQL Data Warehouse or handled differently from standard T-SQL.

Answer 3

I recreated this test using an empty table and did not encounter this error. For confirmation's sake, I would suggest formatting the statement as such:

ALTER TABLE [FactTbl_Test]
ADD CONSTRAINT Ck_Test_id CHECK ([Test_Id] != 'Unknown');

Let us know if you encounter the same error, and ensure only this statement is ran in your batch.

While unlikely the issue here, remember if [FactTbl_Test] already contains a [Test_Id] value of 'Unknown', the ADD CONSTRAINT will fail. You can use the option WITH NOCHECK after your table name, which will create the constraint without validating it against the existing data.