Re: mysql_fdw trouble - Mailing list pgsql-general

From Adrian Klaver
Subject Re: mysql_fdw trouble
Date
Msg-id 5632839B.5070309@aklaver.com
Whole thread Raw
In response to Re: mysql_fdw trouble  (Dane Foster <studdugie@gmail.com>)
Responses Re: mysql_fdw trouble  (Dane Foster <studdugie@gmail.com>)
List pgsql-general
On 10/29/2015 12:56 PM, Dane Foster wrote:
> On Thu, Oct 29, 2015 at 3:30 PM, Adrian Klaver
> <adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>> wrote:
>
>     On 10/29/2015 12:10 PM, Dane Foster wrote:
>
>         On Thu, Oct 29, 2015 at 3:01 PM, John R Pierce
>         <pierce@hogranch.com <mailto:pierce@hogranch.com>
>         <mailto:pierce@hogranch.com <mailto:pierce@hogranch.com>>> wrote:
>
>              On 10/29/2015 11:20 AM, Dane Foster wrote:
>
>                  ​I think you are correct about mysql_fdw "... sending
>             the trim()
>                  checks for remote execution" because according to the docs:
>
>                  "The latest version will push-down the foreign table
>             where clause
>                  to the foreign server. The where condition on the
>             foreign table
>                  will be executed on the foreign server hence there will
>             be fewer
>                  rows to to bring across to PostgreSQL. This is a
>             performance feature."
>
>
>              the alternative would be to fetch the whole table across
>         the FDW
>              interface, then run the where locally, for a large table where
>              you're only selecting a few rows, this would be very painful.
>
>                  I guess using mysql_fdw is a no-go for my data
>             migration needs.
>
>
>              or, rewrite that WHERE clause to be mysql compatible.
>
>         Easier said than done because the LENG​TH and TRIM functions
>         both exist
>         in MySQL but I guess under the covers in PostgreSQL btrim is being
>         invoked when TRIM is called therefore that is what is being "pushed
>         down" to the MySQL and there is nothing I can do about that.
>
>         I guess I could leave out the call to trim, and copy the data into a
>         temp table on the PostgreSQL side, and blah blah blah. My point
>         being
>         why should I have to jump through hoops because mysql_fdw is broken?
>         I'll just go back to writing the migration script as a PHP program
>         because if mysql_fdw didn't exist that's what I would have to do
>         anyway.
>
>
>     Remember you are using a Beta version of Postgres, so it is not
>     entirely unexpected that things might be broken, especially when
>     working with non-core extensions. In the spirit of testing, that
>     Beta implies, why not help fix mysql_fdw by filing an issue? If you
>     already have, my apologies.
>
> ​I'm fully aware of that fact and gladly accept my responsibility which
> is why I have opened an
> issue:https://github.com/EnterpriseDB/mysql_fdw/issues/70

Great and thanks.

>
> For me reporting the issue in the hopes that they will fix it is a
> separate issue from expending energy working around the bug because the
> great thing about the procedural code is that it's littered w/ the same
> SQL that a pure SQL migration script would contain. So if they fix it in
> reasonable amount of time then all that's required to create a pure SQL
> migration script is copy/paste.


>
> Dane​
>
>
>
>              --
>              john r pierce, recycling bits in santa cruz
>
>         ​Dane​
>
>
>
>     --
>     Adrian Klaver
>     adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>
>
>


--
Adrian Klaver
adrian.klaver@aklaver.com


pgsql-general by date:

Previous
From: Jim Longwill
Date:
Subject: Re: Configure Different Databases on One Server
Next
From: Adrian Klaver
Date:
Subject: Re: Configure Different Databases on One Server