Re: Optimization Question -- Follow up on using the stream read (and write) procedures directly

["Jeff Creem" <jeff@thecreems.com>]

"Florian Weimer" <fw@deneb.enyo.de> wrote in message
news:87elxq0ywd.fsf@deneb.enyo.de...
> "Jeff Creem" <jeff@thecreems.com> writes:
>
> > My least favorite part of the approach is the unchecked_conversion
> > of access types.
>
> You shouldn't use Unchecked_Conversion in this case because the
> in-memory and stream representations of an object are usually
> different.

I basically agree but I was probably not clear about what I was trying to
accomplish. This set of
routines is for when  you want the stream representation of an object to
match the in memory
representation and further you want to avoid the huge overhead of the
individual 'read operations
that would happen from the array_type'read attribute.

Even the alternate approach of using unchecked_conversion on the data (v.s.
the access types) would
suffer from the different memory v.s. default stream representation concern
(valid concern but not exactly appropriate for the intended use).

Note that this package could clearly suffer in cases where element_type'size
is < a stream_element'size (and a
few other size cases as well)

But this could be caught easily with a little more error checking.