[GRASS-user] calculating perimeter per patch

Hamish hamish_b at yahoo.com
Sat Feb 12 08:29:00 EST 2011


Milton:
> But, unfortulatelly, as I am working with a very large raster
> map (48,000 x 60,000 pixels; 250,000 clumps), and GRASS fail
> when I try r.to.vect under Windows/Vista.
...
> I will try that. The GRASS_VECTOR_LOWMEM=1 works only
> for G7?

Markus Metz wrote:
> Yes, only for G7. Also Large File Support for wingrass is
> only available in G7.

for what it's worth, the problem here is most likely to be
trying to work with large rasters in Windows, not large vectors.
So GRASS_VECTOR_LOWMEM=1 is probably not needed, but the large
file support in GRASS 7.svn is.

i.e. the same command will probably work on a Mac or Linux using
GRASS 6.4, but if it were truly a large vector file problem and
not a large raster one it would require GRASS 7.svn on all
platforms.

(and LFS vector maps are a helluva lot of data if you consider
how much more efficient vector data is than raster [in general]
...)

mostly I just mean to make sure the right problem gets blamed. :)

also of course you'll need 64bit Vista or Windows7 to get access
to all the RAM you need.


dev note: we need to figure out the sqlite DB file-per-mapset or
file-per-map (like dbf) before we go much further. file-per-
mapset hits filesystem/sqlite library limits & problems much
sooner than per-map, although that's just mitigating the problem
of large-files, not solving it. also per-map is a bit less
damaging if the DB file gets corrupted.



Hamish

ps- oblig comment on perimeters: it's a fractal problem and so
has no real answer. It's only useful to consider as compared
as a ratio to values from the same dataset, and only then if you
can guarantee that the data is consistently measured/created
across the entire map; which is very hard to do. search for the
"length of UK coastline" on wikipedia for reasons. I'd argue we
should consider removing any option which reports perimeter as a
favour to our users, rather than knowingly giving them dubious,
albeit popular, metrics which can, in the long run, only act to
harm their results and mislead their understanding of the data.
or at minimum add a WARNING: perimeter is a garbage metric when
used outside of pure mathematics, use at your own risk.



      


More information about the grass-user mailing list