In this short practice, you will be exploring the differences between database-level constraints and model-level validations in Sequelize. You will also be writing some intermediate and custom validations.

Download starter.

Execute the setup-commands.sh script from the root directory of this practice. View the contents of the script to see the commands that are executed, and view the results in the terminal to confirm that setup was successful.

sh setup-commands.sh

The Express endpoints for this practice will not yield results from a database but are included in order to demonstrate how to initialize Sequelize with an existing server. This can be seen from the plaintext responses given by these endpoints.

In future practices you will utilize routes in order to interact with a connected database.

In this practice, you will be modifying a migration and model for the Colors table. Ultimately, you would like the name attribute of Colors to obey the following rules:

• No NULL values should be allowed
• No duplicate values should be allowed
• The length of name should be between 2 and 20 characters
• The name must not end in the character y

Some of these restrictions will be appropriately implemented using database-level constraints, while some will utilize model-level validations. By the end of this exercise you should be more comfortable with determining when it is appropriate to use each as well as how these constraints and validations affect your interaction with the Sequelize model and the database.

Use sequelize-cli to run the migration as is, creating the Colors table. You'll notice that the migration currently only indicates a type for the name column, with no other constraints.

Use sqlite3 to make the following insertion statements.

-- First statement
INSERT INTO Colors (name) VALUES ('red');
-- Second statement
INSERT INTO Colors (name) VALUES ('red');
-- Third statement
INSERT INTO Colors (name) VALUES (NULL);
-- Fourth statement
INSERT INTO Colors (name) VALUES ('a');
-- Fifth statement
INSERT INTO Colors (name) VALUES ('orangey');

These statements should all execute without error. Selecting for all entries on the Colors table should yield all five values as results.

Think about the requirements that were laid out above. Which of these can be implemented on the database-level? How will these constraints affect the results of the insertion statements above?

Use sequelize-cli to undo the migration.

The first two requirements that were described above can be accomplished with database-level constraints. Add constraints to name in the migration file that disallow NULL values and that require uniqueness.

Rerun the migration with sequelize-cli. Try to do the same insertions with sqlite3 as in the previous step. The second and third insertion statements should result in an error due to the database-level constraints you created. The fourth and fifth insertion statements should still be accepted by the database.

In the following step you will be writing validations on the Sequelize Color model.

In the models/color.js file, notice the current validations on the Color model's name attribute. It currently checks that the data is a string and it disallows NULL values.

In the test.js file, five test cases are created. When run, the file will attempt to create instances of Color that have names of: NULL, [], a, orangey, and orange. Run these tests with npm test. The test file will create its own test database, run the migration, and delete the database afterwards.

With the current validations, the first two scenarios will not be allowed by the Color model. The database will not be interacted with and no SQL should be output for these cases. The last three cases do not violate any validations, so you should see SQL printed for inserting these values.

Modify the init function to add the following model-level validations to the name attribute:

• The length should be between 2 and 20 characters. Include a custom error message, name must be between 2 and 20 characters, or similar.
• A custom function should throw an error if the name ends in the letter 'y'. The error should indicate name must not end in 'y', or similar.

Hint: In order to add additional validations, a validate key will need to be added to the name object. Take a look at the Sequelize validation docs for more examples of built-in and custom validators.

An example implementation of the custom function is given below. It is up to you to determine how best to incorporate this function into your validations:

function noEndingInY(value) {
  if(value.slice(-1) === 'y') {
    throw new Error('name must not end in \'y\'');
  }
}

When these validations have been created, run the test file again with npm test. These validations should now prevent the third and fourth test cases from interacting with the database since a and orangey are no longer valid names. You should only see SQL generated for the final test case, orange, since it is the only one that does not violate a model-level validation.

Open up sqlite3 again and try inserting the values a and orangey directly into the database. You should not see any errors with these insertion statements.

Can you explain why these statements are accepted in the database but result in errors when the model is used to create them?

You are now able to generate Sequelize models and migrations, as well as implement database constraints and model-level validations.