memory dilemma

Hi,


I have dilemma with PG's memory. I'am finishing with to_char()
implementation and I try use internal cache buffer in this routines. 
This cache is used if a to_char() format-picture (which going to the 
internal to_char parser) is equal as previous and to_char's parser is 
skiped. It is very good, because speed rise (20%).

A problem is how implement this cache:
via palloc -  It is standard in PG, but it is problem, because memory contents is not persisten across transactions.
AndI don't know how check when memory is free (lose) and a routine must reallocs memory again (if transaction finish PG
memorymanagementnot zeroizing (reset) memory and any "if( buffer )" still affects as good memory).     
 
via malloc - (Now used). It is good, because buffer is persistent.This variant is (example) use in regexp utils in PG
now.But is it nice?     
 
via a static buffer - but how long? Terrible. Or set any default size for this buffer, and if format-picture will
bigger- usepallocated memory and not use cache buffer. (It is my favourite variant.)  
 
not use cache - hmm.. but I like fast routines (my currentto_char() implementation is faster (20-50%) than current
date_part()).  
 

Any idea? Please.    
                    Karel

PS. IMHO - add to PostgreSQL any *across transactions persistent*    memory managemet?

----------------------------------------------------------------------
Karel Zak <zakkr@zf.jcu.cz>              http://home.zf.jcu.cz/~zakkr/

Docs:        http://docs.linux.cz                    (big docs archive)    
Kim Project: http://home.zf.jcu.cz/~zakkr/kim/        (process manager)
FTP:         ftp://ftp2.zf.jcu.cz/users/zakkr/        (C/ncurses/PgSQL)
-----------------------------------------------------------------------