On Thu, 20 Sep 2001, Philippe M . Chiasson wrote:
> After a fun week of machines crashing all over me, here is my first attempt at
> a tied implementation of APR::Tables.
>
> Important to note that this will most probably be partly rewritten as I don't believe
> it's "the right way" to do it. It does work for me though.
cool. one reason i held off on this, was because i don't think
APR::Table-s should be tied by default. but hadn't thought of a way to
turn that on, but something along the lines of:
use APR::Table tie => 1;
or if they are tied by default:
use APR::Table tie => 0;
would turn it off.
> The few things that I believe are not done right are:
>
> 1. the tie and untie functions should be macros
because of ? speed? you could just MP_INLINE those functions.
> 2. the T_APRTABLEOBJ typemap stuff is just nasty. The only thing I want to be able
>to do
> is to define input/output typemap for an APR::Table, and right now I had to
>hardcode quite
> a few things
what did you have to hard code? APR::Table? that is an xsubpp variable,
$Package i think.
> 3. All the mp_xs_sv2_* macros are completely separate from the typemap entries, and
>I think they
> should be somehow tied together. Right now I have to
> #define mp_xs_sv2_APR__Table(sv) \
> _mpxs_apr_table_untie(aTHX_ sv)
> if I want to use mpxs_usage_va_2 and friends
you should only need mpxs_usage_foo if a function has a variable list
argument (like $r->print(...)). anything else should be taken care of by
the prototype that is generated in the .xs, like the current
APR::Table::{get,set,etc} currently are.
> 4. XS aliases. if apr_table_get and apr_table_FETCH are identical, there should be
>a way to simply
> alias one to the other instead of duplicating the XS glue
> newXS("APR::Table::FETCH", XS_APR__Table_get, file);
> is all that's required.
see Apache__RequestIO.h:
#define mpxs_Apache__RequestRec_PRINT mpxs_Apache__RequestRec_print
and modperl_functions.map:
int:DEFINE_PRINT | | ...
support for ALIAS would be nicer though.
> 5. FIRSTKEY, NEXTKEY not implemented yet, as I have to figure out a place I can
>stash the state of an iterator somwehere.
should be able todo what 1.x does. see APR::URI for an example of
subclassing a data structure.
one note that i mentioned in another thread, no "_private" names please,
i.e. no _ prefix. implemention of your patch is looking good so far, but
the style is way off, please fix :-)
and check that the patch will compile with 5.6.1, i think you'll at least
need to add to modperl_perl_includes.h:
#ifndef PERL_MAGIC_tied
# define PERL_MAGIC_tied 'P'
#endif
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
