Re: [ADMIN] Processing very large TEXT columns (300MB+) using C/libpq - Mailing list pgsql-admin

From Cory Nemelka
Subject Re: [ADMIN] Processing very large TEXT columns (300MB+) using C/libpq
Date
Msg-id CAMe5Gn0T7OtneYReFCvpSqhaTSGvO0Ujf+RjKVX_eeaRRgEPQg@mail.gmail.com
Whole thread Raw
In response to Re: [ADMIN] Processing very large TEXT columns (300MB+) using C/libpq  (Geoff Winkless <pgsqladmin@geoff.dj>)
List pgsql-admin
I'll take out all the code that isn't directly related to reading the data and see if that helps.  That was next step I intended anyway.  

thank you for the reply

--cnemelka

On Fri, Oct 20, 2017 at 9:43 AM, Geoff Winkless <pgsqladmin@geoff.dj> wrote:
It's probably worth removing the iterating code Just In Case.

Apologies for egg-suck-education, but I assume you're not doing something silly like

for (i=0; i < strlen(bigtextstring); i++) {
....
}

I know it sounds stupid, but you'd be amazed how many times that crops up, and for small strings it doesn't matter, but for large strings it's catastrophic.

Geoff

On 20 October 2017 at 16:16, Cory Nemelka <cnemelka@gmail.com> wrote:
All I am am doing is iterating through the characters so I know it isn't my code.

--cnemelka

On Fri, Oct 20, 2017 at 9:14 AM, Cory Nemelka <cnemelka@gmail.com> wrote:
Yes, but I should be able to read them much faster.  The psql client can display an 11MB column in a little over a minute, while in C using libpg library, it takes over an hour.  

Anyone have any experience with the same issue that can help me resolve?

--cnemelka

On Thu, Oct 19, 2017 at 5:20 PM, Aldo Sarmiento <aldo@bigpurpledot.com> wrote:
I believe large columns get put into a TOAST table. Max page size is 8k. So you'll have lots of pages per row that need to be joined with a size like that: https://www.postgresql.org/docs/9.5/static/storage-toast.html


On Thu, Oct 19, 2017 at 2:03 PM, Cory Nemelka <cnemelka@gmail.com> wrote:
I have getting very poor performance using libpq to process very large TEXT columns (300MB+).   I suspect it is IO related but can't be sure.

Anyone had experience with same issue that can help me resolve?

--cnemelka





pgsql-admin by date:

Previous
From: Geoff Winkless
Date:
Subject: Re: [ADMIN] Processing very large TEXT columns (300MB+) using C/libpq
Next
From: Tom Lane
Date:
Subject: Re: [ADMIN] Processing very large TEXT columns (300MB+) using C/libpq