2005.12.23 20:27 "Re: [Tiff] filtering TIFF tags", by Andrey Kiselev
On Thu, Dec 22, 2005 at 05:02:33PM -0500, Frank Warmerdam wrote:
I am concerned about this too. My application can end up re-reading directories quite a bit (as different levels of overviews are accessed for instance) and it seems like the directory reading cost could be quite high.
In addition to "big metadata", I also see this happening with big tile/strip size/offset arrays. In a BigTIFF world, for instance, it would be easy to have files with a million tiles.
It should be pretty easy to make the code that reads custom (ie. generic) tags do so in a lazy fashion. But it could be pretty hard for the tags that are loaded directly into the tiff directory structure as other code then just assumes they can be accessed at will.
Most tags work through the custom path now. Only few essential tags loaded in the directory structure.
Perhaps we could look at doing something for the custom/generic tags first which would solve your problem with big metadata values.
I like the lazy read idea. I will look how it may be implemented, but it will be for the next release, of course.
Andrey V. Kiselev