Re: PostgreSQL CHARACTER VARYING vs CHARACTER VARYING (Length)

On Tuesday, April 28, 2020, Rui DeSousa <rui(at)crazybean(dot)net> wrote:

>
> Arbitrarily? What’s a cusip, vin, ssn? Why would you put a btree index on
> a text field? Because it’s not.
>
> What you’re advocating is a NoSQL design — defer your schema design.
> Letting the application code littered in multiple places elsewhere define
> what a cusip, etc. is.
>
>
All of those would be defined as PKs somewhere with a constraint that
limits not only their length but also allowable characters so you don’t get
something like !(at)#$%^&*( as a valid ssn of length 9. A domain is probably
even better though has implementation trade-offs.

A length constraint by itself is insufficient in those examples, which are
still arbitrary though the decision is outside the control of the modeler.
If the supplied values are external, which they likely are, the system
under design should probably just define the values loosely and accept
whatever the source system provides as-is.

David J.