[geos-devel] geom_accum and memory management

[Paul Ramsey]

We are using the CVS versions in a production environment right now.
Frankly, I could have packaged a release last week, mostly it is just a 
matter of time for me... I like the versioned releases to be internally 
consistent, and that means going through the documentation, etc, and 
making sure everything is up to date, which takes a while.

Paul

Heitzso wrote:

> Hi, I'm monitoring the geos list and am grateful for all of the hard
> work that everyone is putting into this package.  But sometimes the
> details of the whirlwind changes obscure the bigger picture for me.
> Are you nearing a release or are all of the big/good changes that
> are going into cvs going to require a cooling-off period for a few
> months to shake out the last memory/other bugs?
> Probably a better way to phrase my question is .... when
> will these mods be settled in and ready for use in a production
> environment?
> 
> Thanks!
> 
> strk wrote:
> 
>> The introduced geom_accum function will force a copy of DETOASTED
>> geometries to allow for final functions release of memory associated
>> with each array element. Without an explicit copy, PG_DETOAST_DATUM
>> might or not allocate new memory, so we will not be safe explicitly
>> pfree'ing result.
>>
>> Now.
>> Final aggregate functions (fastcollect, fastunion) will release
>> memory associated with each array element to reduce old memory
>> usage while allocating new memory for the resulting objects.


-- 
       __
      /
      | Paul Ramsey
      | Refractions Research
      | Email: pramsey at refractions.net
      | Phone: (250) 885-0632
      \_