[GRASS-dev] r.out.gdal rewritten - won't scale beyond 2GB

Glynn Clements glynn at gclements.plus.com
Thu Oct 26 14:47:57 EDT 2006

M-Dàris Narti-B¹s wrote:-A

> Sorry, I'm not reading related code, but does not many current GRASS modules 
> relay on loading whole raster map into ram? Not all GRASS modules scale 
> well :)

Some do, mostly for algorithms which can't readily be implemented
using sequential I/O. Most of them aren't exactly "core" functionality
though (and r.out.gdal *is* core functionality).

> IMLO better working version today than excellent version tomorrow. Talks about 
> replacing r.out.gdal script w C module I hear for long time, but no one does. 
> Now we have an option, why not use it?

Because failing to work on maps which won't fit into memory is a major
problem for an export module. As it stands, the module is almost
useless, in the sense that it has far more significant drawbacks than
the r.out.gdal script which we already have.

It's not reasonable to call the current version a "working" version. 
GRASS is supposed to be able to handle maps which are too large to fit
into memory. If it could be assumed that a map could be read into
memory, everything would be so much simpler. The reason why modules
have to go to the trouble of using a get-row/put-row loop, managing
row buffers etc is because reading an entire map into memory isn't
considered acceptable.

Glynn Clements <glynn at gclements.plus.com>

More information about the grass-dev mailing list