Re: [UFRaw-Devel] [PATCH] Arbitrary rotation support for ufraw-batch

[Martin L. <mar...@ea...>]

On Tue, Mar 25, 2008 at 08:34:41PM +0100, Udi Fuchs wrote:
> 
> Why did you change rowStride to rawStride? rowStride counts how much
> you need to jump from row to row. The same name is also used in
> GdkPixbuf.

It was just for readability, to distinguish from the row stride of the
developed image which was being rotated. I don't have another variable
with 'stride' in the name now though.

> The whole rotation logic should not be in ufraw_writer.c. It belongs
> in ufraw_ufraw.c. Still, assuming that this code is temporary, it
> could stay for now where it is.

It's temporary, yes. Once the dust has settled I envisage it moving to a
separate function, where it could also be shared with preview code.

> I would try changing sine and cosine from double to "rational". We
> could have sineCol with a dominator of width and sineRow with a
> dominator of height. This way the calculation should be still exact
> but much faster. Again, this is something that could be improved after
> the initial patch is applied.

I haven't really been pushing it for performance at this stage and I'm
sure it could be made much faster. I'm more worried about the integration
of the feature right now.

I'll make the other changes you suggested and send a new patch.


Martin