> Code can provide just as much safety as the database
No, database constraints declaratively specify invariants that have to be valid for the entirety of your data, no matter when it was entered.
Checks and constraints in code only affect data that is currently being modified while that version of the code is live. Unless you run regular "data cleanup" routines on all your data, your code will not be able to fix existing data or even tell you that it needs to be fixed. If you deployed a bug to production that caused data corruption, you deploying a fix for it will not undo the effects of the bug on already created data.
Now, a new database constraint will also not fix previously invalid data, but it will alert you to the problem and allow you to fix the data, after which it will then be able to confirm to you that the data is in correct shape and will be continue to be so in the future, even if you accidentally deploy a bug wherein the code forgets to set a reference properly.
I'm fine with people claiming that, for their specific use case or situation, database constraints are not worth it (e.g. for scalability reasons), but it seems disingenuous to me to claim that code can provide "just as much safety".
No, database constraints declaratively specify invariants that have to be valid for the entirety of your data, no matter when it was entered.
Checks and constraints in code only affect data that is currently being modified while that version of the code is live. Unless you run regular "data cleanup" routines on all your data, your code will not be able to fix existing data or even tell you that it needs to be fixed. If you deployed a bug to production that caused data corruption, you deploying a fix for it will not undo the effects of the bug on already created data.
Now, a new database constraint will also not fix previously invalid data, but it will alert you to the problem and allow you to fix the data, after which it will then be able to confirm to you that the data is in correct shape and will be continue to be so in the future, even if you accidentally deploy a bug wherein the code forgets to set a reference properly.
I'm fine with people claiming that, for their specific use case or situation, database constraints are not worth it (e.g. for scalability reasons), but it seems disingenuous to me to claim that code can provide "just as much safety".