Classify validations

Users do not always do things exactly as they should. They often make mistakes. This implies that we always first check whether they have requested "correctly" and then proceed with it. The correctness check, or validations as we call them, are following types.

• Request validation checks whether the request is valid or not irrespective of current state of persisted domain model. These are individual field level or validation across two or more fields with in the request. Most the individual request validations are quite generic and composable in nature, for example, empty/max-size etc. This characteristic is another reason why such validations should be treated differently.
• Domain validation checks whether the request is valid in the context of current domain state. This is performed after activating the domain entities and are geared towards checking business rules. for example, whether I can cancel my ticket x hours before the journey start time.
• Data invariance checks the business rules which cannot be ascertained in the domain layer for concurrency reasons (or because of race condition). for example, overdraft from an account.

Quite often data invariance checks are incorrectly performed during domain validation, as the race condition is over-looked during development. Lets take the example of verifying whether a user with same email-address exists in the system.

class RegistrationService {

    UserRepository userRepository

    public User register(NewUserRequest newUserRequest) {

        User user = mapToUser(newUserRequest)

        if (userRepository.findByEmail(user.emailId()) != null) {

            throw new UserAlreadyExistsException()

        }

        return userRepository.save(user)

    }

}

If two threads (caused by user requests), try to request for same emailId then it is possible that findByEmail might return null for both of them we would two users with same email address. Such validation can be implemented by putting a unique constraint in the database and handling the exception in code.

class RegistrationService {

    public User register(NewUserRequest newUserRequest) {

        try {

            User user = mapToUser(newUserRequest)

            return userRepository.save(user)

        catch (SQLException sqlException) {

            if (sqlExceptionCausedByUniqueConstraint("UC_USER_EMAILID"))

                throw new UserAlreadyExistsException()

            throw

        }

    }

}

In distributed database one might need to do more than this but the what still holds true is that such checks userRepository.findByEmail(user.emailId()) != null are not very useful. Infact as code review of domain validation one should always keep an eye for data invariance checks.