On Feb 11, 2010, at 7:07 PM, Andrew Dunstan wrote:

> 
> 
> Alexey Klyukin wrote:
>> On Feb 11, 2010, at 6:24 PM, Andrew Dunstan wrote:
>> 
>>  
>>> Alexey Klyukin wrote:
>>>    
>>>> Hello,
>>>> 
>>>> When developing pl/perlu functions common definitions and methods are often stored in external .pm modules. During deployment the modules should be installed somewhere in @INC to be reachable by the perl interpreter. However, installing the modules to a location outside of the PG installation makes it hard to have a consistent environment when running multiple PG versions on the same host. What do you think about defining a canonical place for pl/perlu .pm modules (i.e. PKGLIBDIR) and adding this location to @INC during the interpreter initialization ? Another idea is to allow a user to specify such location by adding a new custom GUC variable.
>>>> 
>>>> 
>>>>       
>>> Why won't setting this in the new on_perl_init setting work? It's even included in to documented examples using the standard lib module: <http://developer.postgresql.org/pgdocs/postgres/plperl-under-the-hood.html#PLPERL-CONFIG>
>>>    
>> 
>> The lack of support for SPI functions makes this hardly an adequate solution. I do have both modules and SPI calls in several pl/perlu functions.
>> 
>>  
> 
> 
> That has nothing to do with what you asked about, namely setting the include path. You can set the include path in on_perl_init with "use lib" and then "use" your modules in your plperlu functions, at which point SPI will be available.

Ah, it seems I misinterpreted the documentation. This is much better, but still it requires setting the path explicitly. 
What about having a default location for the modules that is automatically added to @INC and is recommended in the documentation as a place to put .pm files  ?

--
Alexey Klyukin				    http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc

In response to

• Re: a common place for pl/perlu modules at 2010-02-11 17:07:46 from Andrew Dunstan

pgsql-hackers by date

Next:	From: Karl Denninger	Date: 2010-02-11 17:38:42
	Subject: Re: [GENERAL] [HACKERS] Bug on pg_lesslog
Previous:	From: Marko Tiikkaja	Date: 2010-02-11 17:35:18
	Subject: Re: Writeable CTEs and empty relations