Pass through buffer

From: Tom Howard (tomhoward_at_[hidden])
Date: 2002-07-29 07:46:30

Hi,

I'm wondering if someone can help me with the pass through buffer 
(basic_passthrubuf) that is part of more_io.

>From what I can understand the pass though buffer passes 1 character at a 
time to (and from) the underlying stream buffer and I'm wondering what 
happens with flushes.

In the code I'm working on (gzostreambuf which writes to an ostream) I need 
to know is overflow() is being called because there is no more room in the 
buffer I'm using or because of an explicit flush.  In the later case I need 
to also flush the compression (Z_SYNC_FLUSH), but doing so all the time will 
degrade the compression performance plus increase the number of flushes to 
the underlying stream.

This is particularly bad as the underlying stream in my case is a socket 
stream an the increased flushes breaks up the data into many small packets.  
Because of the software at the other end of the socket, data must be sent as 
one message unless it's over 8096 bytes.

So basically I only want to flush the underlying stream if the gzostreambuf 
has been flushed (either by endl or flush). is this possible?

My current workaround is to make sure the buffersize of my gzostreambuf is 
greater than or equal to the buffersize of the underlying stream, but I was 
hoping there was a neater solution and thats why I was looking at the pass 
through buffer.

>From what I unbderstand, If I use the pass through buffer as part of my 
gzostreambuf, then flushing will not be passed to the underling streambuf at 
all.  This means that data sent to the stream may be compressed, but result 
in no output for quite some time later and this sort of latency would be 
unacceptable in the software I'm working on (client-server multi-agent 
simulation).

Cheers,

Tom Howard