Message-ID: <gh-pgjdbc-pgjdbc-2656-c1439200147@github.com>
From: "ash211 (@ash211)" <noreply+ash211@github.com>
To: "pgjdbc/pgjdbc" <noreply+pgjdbc-pgjdbc@github.com>
Date: Tue, 21 Feb 2023 22:57:54 +0000
Subject: Re: [pgjdbc/pgjdbc] issue #2656: Adaptive Fetch doesn't enable ResultSet streaming
In-Reply-To: <gh-pgjdbc-pgjdbc-2656@github.com>
References: <gh-pgjdbc-pgjdbc-2656@github.com>
List-Id: <gh-pgjdbc-pgjdbc.github.com>
X-GitHub-Author-Login: ash211
X-GitHub-Comment-Id: 1439200147
X-GitHub-Comment-Type: issue_comment
X-GitHub-Issue: 2656
X-GitHub-Repo: pgjdbc/pgjdbc
X-GitHub-Type: comment
X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/issues/2656#issuecomment-1439200147
Content-Type: text/plain; charset=utf-8

If you remove the maxResultBuffer setting then I think adaptiveFetch is disabled.

Per https://github.com/pgjdbc/pgjdbc/blob/2b90ad04696324d107b65b085df4b1db8f6c162d/README.md?plain=1#L145, `adaptiveFetch`

> Requires declaring maxResultBuffer and defaultRowFetchSize for first iteration.  

----------------

I recently ran into this, where I had set `adaptiveFetch=true` and `defaultRowFetchSize=10` and `maxResultBuffer=10percent` and yet still a user of my tool observed this failure message: `Result set exceeded maxResultBuffer limit. Received: 103887670; Current limit: 103887667`

It seems that currently this JDBC driver requests "give me N rows" and gets back those N rows, regardless of their size.  If postgres server supported such a thing, we may prefer "give me N rows or up to B bytes" and receiving 1..N rows instead of always N.  This would enable smoother streaming of a ResultSet, without OOM or hitting `maxResultBuffer` limits while still maintaining reasonably large fetch sizes (not just 1 all the time).