The Problems With Validating ActiveRecord Models and Why State Validation Is a Bad Id
In the typical Rails application, you can find the most of the validations in the ActiveRecord models, which is nothing surprising - ActiveRecord models are used for multiple things. Whether it is a good thing, or a bad thing (in most cases it’s the latter) deserves a separate book or at least blog post-series as it’s not a simple problem, there is one specific thing that can cause a lot of issues that are difficult to solve and go beyond design decisions and ease of maintenance of the application, something that impacts the behavior of the model - the validations.
Just to give you a real-world example of what validation in ActiveRecord model looks like (as impossible as it seems, it really did happen) - when updating the check-in time of the reservation, which is a simple attribute on Reservation model, the record turned out to be invalid because… the format of guest’s phone didn’t match some regexp.
There are multiple ways to bypass this problem: use
validate: false flag with
save(validate: false) or use
update_columns method, but this is definitely not something that can be applied in a “normal” use case. In a typical scenario, this will be the error message displayed in the UI/returned to API consumer, and it will be confusing.
However, this is the expected behavior of ActiveRecord (or in general, ActiveModel-style) validations, which is a validation of the state of the model. And judging from this example, it’s evident that it leads to problematic scenarios. What kind of design then would be the most appropriate to prevent such issues?