Re: Interface for composite types

On Jun 7, 2008, at 16:12, Greg Sabino Mullane wrote:

> It does add some overhead, so I'm not comfortable having it  
> automatically
> enabled quite yet. This method basically slurps all the custom type
> information from the system catalogs and stores in internally in a
> C/XS LL struct. When we are fetching and find an unknown type, we walk
> the list of custom types to find a match. If we do find one, we make
> a temporary sql_info_type struct for it and move on. Then we call a
> second internal function for each row, pg_build_hashref(), which  
> recursively
> converts the string returned by Postgres into a glorious nest of  
> hashrefs and
> arrayrefs. Most of all that is written. What's also needed is the  
> reverse
> mapper, to change a bunch of hashrefs into a stringified form to  
> pass to
> PQexec and friends, and some minor bind_param hacks.

Nice. Would this work for hstore types, then?

Best,

David