Re: Confusing messages about index row size - Mailing list pgsql-hackers

From David G. Johnston
Subject Re: Confusing messages about index row size
Date
Msg-id CAKFQuwb9HiC4e1nGhwrikHo2B1f3SRbGfuTh+TKUFa0haUcTqg@mail.gmail.com
Whole thread Raw
In response to Confusing messages about index row size  (Jaime Casanova <jcasanov@systemguards.com.ec>)
List pgsql-hackers
On Sunday, September 12, 2021, Jaime Casanova <jcasanov@systemguards.com.ec> wrote:

So, what is it? the index row size could be upto 8191 or cannot be
greater than 2704?

The wording doesn’t change between the two: The size cannot be greater the 8191 regardless of the index type being used.  This check is first, probably because it is cheaper, and just normal abstraction layering, but it doesn’t preclude individual indexes imposing their own constraint, as evidenced by the lower maximum of 2704 in this specific setup.

It may be non-ideal from a UX perspective to have a moving target in the error messages, but they are consistent and accurate, and doesn’t seem worthwhile to expend much effort on usability since the errors should themselves be rare.

David J.

pgsql-hackers by date:

Previous
From: Jaime Casanova
Date:
Subject: Confusing messages about index row size
Next
From: "wangsh.fnst@fujitsu.com"
Date:
Subject: RE: drop tablespace failed when location contains .. on win32