Hi, Jeff! I should have walked down the hall to ask you! :)
No go on closing first. fiona.open returns an unconsumed iterator so
that it can handle huge featuresets. If you close it before iterating
over it, it raises a ValueError "I/O operation on closed collection."
The data is not actually "open in two places." It's only open on the
fiona collection which is normally used within a context manager where
it stays open until the collection iterator is exhausted. Look at the
examples in the fiona docs.
I'll look at mapnick. But what I'm really trying to do is put together
an ipython notebook showing full vector to raster trip using fiona and
rasterio. The docstring for rasterio.features.rasterize advertizes this:
":param shapes: an iterator over Fiona style geometry objects (with a
default value of 255) or an iterator over (geometry, value) pairs.
Values must be unsigned integer type (uint8)."
And I think right there I see my problem. If "values" in the statement,
"Values must be unsigned integer type (uint8)," means the feature
properties, then I have a problem because:
>>> shp.schema['properties']
OrderedDict([(u'vmax', 'float'), (u'vmin', 'float')])
>>>
Looks like I need to walk my shapes and change those ranges of floats
into uint8 categories. Now I understand why Sean told me yesterday that,
"Rasterio isn't map-making software so this feature will always be
cartographically limited..." If the data resolution is limited to 8
bits, then yeah, cartographically limited. I should use mapnick. The
rasterio use case seems to be about decorating pre-existing rasters with
shapes.
Just for a sanity check, I consumed the collection returned by fiona by
wrapping it in list(), closed the shapefile, and then fed the list of
features to rasterize, without first changing the OrderedDict to a uint8:
import fiona
from rasterio import features
collection = fiona.open("myfeatures.shp")
shapes = list(collection)
image = features.rasterize(shapes, out_shape=(1000,1000))
And that resulted in ValueError "Unsupported geometry type Feature"
because my shapefile represents a collection of features where each
feature is a single range for a value (i.e., four features, one for each
of the ranges 0.0 to 1.0, 1.0 to 2.0, 2,0 to 3.0, and 3.0 to 4.0). I
should walk the features and burn each into its own band.
But if I rasterize just one of the features with, say, shape[0], then
I'm back to the seg fault, presumably because of the bad value type.
Walking the features to convert the data values to uint8 is going to be
a pain because each feature is a collection of multipolygons (polygons
with holes, with islands in the holes, with holes in the islands, and so
on recursively). I sure hope mapnick can handle multipolygons. Even
better would be if it can handle arbitrary data types other than
scalars, or at least handle more data resolution than 8 bits.
Thank you for the mapnick suggestion.
--
Sincerely,
Chris Calloway, Applications Analyst
UNC Renaissance Computing Institute
100 Europa Drive, Suite 540, Chapel Hill, NC 27517
(919) 599-3530
On 10/14/2014 6:35 PM, Jeff Heard wrote:
Chris, I'd close the source dataset first. OGR/GDAL can have serious
issues with having data open in two places, and sometimes it's
surprising what causes it.
You might also try outputting via mapnik2
On Tue, Oct 14, 2014 at 6:35 PM, Chris Calloway <c...@unc.edu
<mailto:c...@unc.edu>> wrote:
I'm trying to programmatically generate image data from a shapefile
with something like this:
import fiona
from rasterio import features
collection = fiona.open("myfeatures.shp")
image = features.rasterize(collection, out_shape=(1000,1000))
And nearly immediately I get:
"Segmentation fault: 11"
I'm using fiona and rasterio from anaconda on OSX 10.9.5:
$ conda list | grep rasterio
rasterio 0.8 py27_0
$ conda list | grep fiona
fiona 1.1.6 np19py27_0
Has anyone gone full trip from shapefile to image data through fiona
and rasterio.features.rasterize? If so, do you have advice? I would
expect this is just wrapping gdal's rasterize.
--
Sincerely,
Chris Calloway, Applications Analyst
UNC Renaissance Computing Institute
100 Europa Drive, Suite 540, Chapel Hill, NC 27517
(919) 599-3530 <tel:%28919%29%20599-3530>
_______________________________________________
Community mailing list
Community@lists.gispython.org
http://lists.gispython.org/mailman/listinfo/community
