2012/11/14 Larry Gritz <[email protected]>

> On Nov 14, 2012, at 1:04 AM, Michel Lerenard wrote:
>
> > I did noticed it, but it doesn't bother me that much (i am using these
> functions). I thought there was a good reason behind it.
>
> Do you use them to get arbitrary ranges, or do you use the "channel
> subset" version only to extract a single channel at a time?
>
> I extract channels one by one, we work by channel in our application.



> > I think it would be better to harmonize the parameters, and drop the
> range version.
>
> Drop the range version?  The "range" version is begin/end.  Or did you
> mean use the range notation everywhere?
>
> Sorry i thought the range version was start/count. I meant use begin/end
everywhere.


> > Having both will complicated things, the signatures being the same we'd
> need to have different function names.
>
> Ick, I hate to add more functions and pick a less convenient name.  Did
> you see the response I wrote to John's mail?  I proposed that if
> chend<=chbegin, it means to read one channel.  That would make the "all
> channels" and "one channel" case strictly back-compatible.  Do you think
> that would cover enough cases that it would be ok to change the meaning of
> those parameters, even though the type signature has remained the same?
>

I agree with the idea, but it won't be compatible with single channel read
for the first channel: the call will have parameters : beginchan = 0 /
endchan = 1 and it will read two channels instead of one.
But i like the idea of reading at least one channel. It means that if you
set endchan to zero you're sure to read one channel.
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org

Reply via email to