Re: BUG #14048: copy issues with jsonb

Ok. There are 288K rows in the file that was failing yesterday. I'll
play with date ranges until I can narrow it down to a few rows and send
it along.

I wasn't able to get to trying pgdump yesterday, I'll give it a try
today and let you know.

What I'm doing manually is exactly what I posted.

On 2016-03-27 22:25, David G. Johnston wrote:

> On Sun, Mar 27, 2016 at 8:04 AM, <rotten(at)windfish(dot)net> wrote:
>
>> The following bug has been logged on the website:
>>
>> Bug reference: 14048
>> Logged by: Rick Otten
>> Email address: rotten(at)windfish(dot)net
>> PostgreSQL version: 9.5.1
>> Operating system: Ubuntu 14.04
>> Description:
>>
>> I have a few jsonb columns in a table and I've run into several issues with
>> COPY with that column type. I didn't see anything specific about it in the
>> to-do list, and Google didn't turn up anyone else specifically complaining,
>> so I'm filing it here. I apologize if this is a known issue.
>>
>> The table is the backend for a web pixel (tracking) service. It has 4 jsonb
>> columns which contain variable data about the browser and session. (parsed
>> user agent, list of cookies, query parameters, etc...)
>>
>> I wanted to copy a section of the table from one 9.5.1 database to another.
>>
>> For some reason the CSV and TEXT format exports do not quote/escape the
>> nested JSON very well - after several tries, I gave up on that approach.
>> The "copy to" output appears to be too mangled. It was not obvious, at
>> first glance, where it is getting confused.
>
>> ​[...]
>
>> Meanwhile, I'll use pg_dump to get the rows I need. Hopefully that will be
>> able to extract and reload the jsonb data ok. pg_dump hasn't been a problem
>> in the past anyway...
>
> ​I'm reasonably certain pg_dump/pg_restore makes use of "COPY" to perform its work - at least for custom format and plain non-inserts mode​ - so if that works but whatever you are doing manually does not there is some degree of suspicion that what you are doing may be incorrect.
>
> We really need a sample record and a self-contained script that you can make reproduce this behavior.
>
> David J.