[gdal-dev] basic raster math

[Christopher Barker]

Ari Jolma wrote:
>> numpy_array = band.ReadAsArray()
> 
> Just to make sure, I guess this reads the whole thing into memory?

yes.

> I.e.  don't try this with huge images.

well, you can TRY it ;-)

ReadAsArray() supports accessing a subset of the band as well, if you 
want to work with monster images in chunks.

WolfgangZ wrote:
> 1. accessing the cell value of a (float) tiff. I know how to load the 
> map into an array, but this seems to be overkill.

As above, ReadAsArray can load a subset of the band, presumable 
including a single cell. You could use ReadRaster too, as there isn't 
much point to using arrays for a single value.

> 2. can I get the surrounding cells or better directly (bilinear/cubic) 
> interpolated data?

same as above! ReadAsArray will also re-scale for you.

My Question: can I control what re-scaling algorithm is used?

> 4. can I change the resolution of the input map directly in python (as 
> replacement of gdal_translate -tr)?

I think you can use ReadAsArray/WriteAsArray to do this too.

Jose Luis Gomez Dans wrote:
> data = numpy.array(g.GetRasterBand(1).ReadAsArray().tolist())
> #The previous line converts the oldgen python Numeric array into a 
> #spanking numpy array.

This is far less efficient than you need!

data = numpy.asarray(g.GetRasterBand(1).ReadAsArray())

should to it, and without any memory copying!

If that fails, then numpy.fromstring( array.tostring() )would be a lot 
better than tolist()


> 	dst_ds.GetRasterBand(1).WriteArray( raster )

This is the key line -- thanks!

This really points out that we need some Recipe pages in the Wiki for 
Python/GDAL -- anyone want to get that started?

While we're on the topic -- what are the plans for numpy vs. Numeric? It 
looks like the current utilities and samples all allow either one -- I'd 
like to see a switch to only-numpy at some point -- the sooner the 
better. It's what's supported now, and it does have small but subtle and 
useful differences.

Also -- are there any Python API docs out there that I haven't found? If 
not, we could use them -- maybe doc strings and some auto-generation 
tool, like pydoc or Doxygen. What is being used for the C/C++?

Robin Dunn (of wxPython) has done some nice work of auto-generating docs 
strings as part of the SWIG process -- perhaps we could borrow that 
approach.

-Chris








-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chris.Barker at noaa.gov