Re: gnome-vfs/GIOChannel for parsing

[David Helder <dhelder umich edu>]

On 17 Feb 2003, Michael Meeks wrote:
> 	Ideally - having a single I/O API would be best; however GIOChannel as
> you point out is not a GObject but a semi-OO custom thing. This is
> because glib doesn't depend on/use GObject - as a feature.
>
> 	Other issues might be that it's nice to have a GInputStream and a
> GOutputStream separation - and this can simplify the process of writing
> small, correct implementations. Then of course - there is the
> seekability dimension that adds problems.

As long as we're describing our ideal IO object:

The thing that bugs me with GIOChannel is that there is no easy way to
derive from it correctly.  It's implementions, GUnixIOChannel and
GWin32IOChannel, are hidden.  So, I can't include the struct as the first
member of my child struct like I normally would.  I could delegate using a
sort of container GIOChannel, but this seems hacky and I'm not sure if
watches would work right.  Has anyone been able to derive from GIOChannel?
What did you do?

David


-- 
David Helder - dhelder umich edu - <http://www.eecs.umich.edu/~dhelder>