[Gdal-dev] RFC 15: Band Masks

[Simon Perkins]

This all seems reasonable to me. Just to clarify: in the case where 
there is a genuine binary valued bitmaks per raster band, will 
GetMaskFlags() return 0? i.e. is it intended that there is no specific 
flag for GMF_PER_BAND?

Cheers,

Sy

Frank Warmerdam wrote:
> Folks,
>
> After some thinking, and discussions with Howard, I have revised RFC 15
> a bunch.  Highlights:
>
>  o Now I plan to return alpha bands as the mask if they apply.
>  o I have changed the name of the methods to just GetMaskBand() and
>    get GetMaskFlags() to avoid confusion.
>  o I have written in more detail about the default implementation.
>  o I have changed the GetMaskType() to GetMaskFlags() with flags bitwise
>    or'ed together for expandability and so multiple conditions can 
> apply at
>    once.
>
> I have also explicitly indicated that the API is not being extended at 
> this
> time for the creation of masks.  Driver CreateCopy() implementations can
> capture masks from the source dataset if applicable, but that is the only
> mechanism GDAL can use to create them for now.  Related to this, I'm
> avoiding addressing band masks in PAM (the persistant auxilary metadata
> mechanism).  I'm doing this mostly to keep things managable for now, and
> because I think some careful thinking needs to be applied before trying
> to provide explicit creation.
>
>   http://trac.osgeo.org/gdal/wiki/rfc15_nodatabitmask
>
> I'm very pleased with the move to expressing nodata, alpha bands and
> explicit bitmasks all through the mask model.  I hope others will feel
> the same!
>
> Best regards,