[interfacekit] Re: Flattened BMessages
- From: "Cedric Degea" <cdegea@xxxxxxxxxxxxxxxx>
- To: interfacekit@xxxxxxxxxxxxx
- Date: Wed, 22 May 2002 00:26:18 +0200
>
>Has anyone looked into the flattened BMessage format yet?
A little bit;
> I wrote a BMessage
>implementation, but flattening and unflattening is still not
compatible.
>
Neither is mine; things looked good for simple cases (one-level,
ie. not Addmessage()ing a message within another) but I realized
later it does not "scale up".. here's what I have anyway in case
it helps; let's start with a simple
overview, the BMessage::FlattenedSize() method:
{
// declare accumulator, and init right away to header size
ssize_t size =
sizeof('1BOF')
+4 ///??
+sizeof(ssize_t)
+sizeof(uint32)
+1 ///??
;
// add up each slot into the accumulator
for( uint32 i = 0 ; i < nameList.size() ; i++ )
{
size += 1; ///??
size += sizeof( type_code );
size += sizeof( uint8 );
size += sizeof( uint8 ) +strlen(nameList[i].slotName);
size += dataLengthAt( i );
}
// trailing null byte
size += sizeof('\0');
// finito!
return size;
}
This does not reflect some more thinking I had to do afterwards
when realizing I was missing a big piece of the puzzle:
(here i've had to censor some of the fairly ugly comments in French
I had written that outline my ignorance) :
status_t BMessage::Flatten( BDataIO * stream, ssize_t * size_into = NULL
) const
{
ssize_t flattened_size = FlattenedSize();
stream->Write( "1BOF", 4 ); ///maybe this is actually 'FOB1' in intel
notation?? (Flattened Object format version 1..)
stream->Write( "????", 4 ); /// ressembles a flattened_size &
0x10000000 sometimes, sometimes not.. is that a hash or what?
stream->Write( &flattened_size, sizeof(ssize_t) );
stream->Write( &what, sizeof(uint32) );
stream->Write( "?", 1 ); ///hard-coded to 0x01 maybe?
for( uint32 i = 0 ; i < nameList.size() ; i++ )
{
// ..remember the size..
ssize_t payload_len = dataLengthAt( i );
#warning nah, for integers (int8, int32..) ok, but it seems for a STRING
it becomes 0x03....
#warning ** TODO refaire mon code, Be's ' BMessage slot header is
actually
/*
0x01
0x0f or 0x03
GNOL or RTSC or..
num_entries if string, num bytes if integers (! ***)
num bytes i nthe WHOLE section (! ****) i.e., AddString + AddString
will produce a 0x18 here!
[snip - how the heck are STRINGS stored??]
*/
// Be's BMessage seems to have a 0x0f before each field..??
uint8 weird_header = 0x0f;
stream->Write( &weird_header, sizeof(uint8) );
// slotType, e.g. 'LONG'
stream->Write( &nameList[i].slotType, sizeof(type_code) );
// numItems * sizeof one atom.
stream->Write( &payload_len, sizeof(uint8) ); //use an uint8
rather
than ssize_t? Nope it works as is even on Intel...
// slotName, e.g. "bytes"
uint8 len = strlen( nameList[i].slotName );
stream->Write( &len, sizeof (uint8) );
stream->Write( nameList[i].slotName, len );
// dataAddy
stream->Write( nameList[i].dataAddy, payload_len );
}
// Be's BMessage appends a blank byte at the end...??
stream->Write( "", sizeof('\0') );
// Also process our second param
//
if( size_into )
*size_into = flattened_size;
// success
return B_OK;
}
Cedric (thankful to be lurking here and every now and then
trying to justify being allowed to with meager tidbits :-)
--
http://cdegea.free.fr/ | BeDev E-16870
"What's oil got to do, got to do with it" -- F02 Chorus
Other related posts:
