[Pan-users] Technical question about how Pan is implemented (hopefully not OT)

[Edith Gross]

I just wonder how data are received in Pan, that is how many times are the 
data copied until the are displayed or written to disk.

When you receive the data from the socket, you "read" them and then you 
copy all data first. But you cannot read a line directly 
(g_io_chanel_read_line_string beeing a rather high level function, offering 
a functionality the socket does not provide), so probably they are copied 
the second time, etc.

Is not the best way to allocate 50M of memory and then read all the data 
into that memory from the socket. Then the whole formatting and scanning of 
lines and searching for the last ".\r\n" could be done in situ, without 
moving the data.

Or has this solution its drawbacks too? 50M is nothing nowadays (my machine 
has for example 2G of ram)? Or maybe the copying is so fast (at 3.0GHz) 
that this is not necessairy?

Then as the data are received, another thread could scan and decode them at 
the same time for the fastest possible performance?

I also noticed that the most time consuming operation is the final 
filtering that may take seconds.

TIA,
--
EG