Add utility function for converting geotiffs to area definitions

[djhoese]

Users some time want to get an area definition for a geotiff file. This isn't a difficult thing to do necessarily, but is easy to get wrong. @loerum has already added this type of functionality to mipp for reading SARS data, but it could likely be rewritten or made more flexible: https://github.com/pytroll/mipp/blob/master/mipp/read_geotiff.py

I think if the function just takes a gdal object gdal.Open(fn) and uses the GDAL API for getting the projection information it may work for more than just geotiff.

[drnextgis]

Hi @djhoese I'm on it.

[pnuu]

We already have it, although it needs also the CRS object from rasterio: https://github.com/pytroll/satpy/blob/master/satpy/readers/generic_image.py#L103

[djhoese]

@drnextgis Sounds good. I think you could borrow the logic that @pnuu links to, but I'd still like this type of utility in pyresample itself. Some of the logic for this will overlap with what is being done in #138 but that's ok since that PR isn't ready quite yet.

I'm assuming you are trying to get hacktoberfest PRs?

I think the utility function should accept either a filename, a gdal object, or a rasterio object. If a filename is specified then import rasterio and open it. Then it will do whatever calculations are necessary to generate an AreaDefinition.

[drnextgis]

Yes I found this issue by hackctoberfest tag. But it is a good opportunity to know more about pyresample.

[drnextgis]

I start to work on this issue. In example provided by @pnuu the rotation of source raster is not taken into account. But I have questions about this. For example rasterio's bounds method takes into account rotation rasterio/rasterio#1420 (fixed). Is it expected by AreaDefinition constructor that area_extent parameter calculated with taking into account a rotation? I see at least two possible option, but maybe I miss something:

• pass to area_extent bbox calculated by rasterio + affine rotation to rotation argument
• pass to area_extent and rotation values of affine of source image

What is the right approach?

[djhoese]

@drnextgis Do you have an example?

[drnextgis]

Example of rotated raster?

[djhoese]

Sorry, I misunderstood your question. The rotation argument should be the degrees of rotation:

    rotation: float
        rotation in degrees (negative is cw)

So I guess you could calculate that from the affine rotation. I don't use a lot of rotated area definitions so I can speak on the best way to do this. @mraspaud or @pnuu or @loreclem may have other opinions.

[drnextgis]

I've made a PR, please take a look.