Re: field error on refreshed materialized view

2024-01-27 Thread Michael Nolan
Followup:  The problem turned out to be a field in the mysql server
that was not size-limited to 255 characters and had several rows with
as many as 299 characters in them.  Apparently when using an FDW and
materialized view in postgresql 10.4, field size limits aren't
checked, because a query on the matview in the 10.4 system finds those
records with more than 255 characters in that field.

So, not a character encoding issue at all.
--
Mike Nolan




Re: field error on refreshed materialized view

2024-01-27 Thread Tom Lane
Michael Nolan  writes:
> Followup:  The problem turned out to be a field in the mysql server
> that was not size-limited to 255 characters and had several rows with
> as many as 299 characters in them.  Apparently when using an FDW and
> materialized view in postgresql 10.4, field size limits aren't
> checked, because a query on the matview in the 10.4 system finds those
> records with more than 255 characters in that field.

> So, not a character encoding issue at all.

Hm ... if the foreign table's field was declared as varchar(255),
older PG might've believed that and not rechecked the length while
storing into the matview.  Not sure if this behavior change was
intentional or an accidental product of refactoring.

regards, tom lane