Python stops working on Windows when calling layer.GetFeatureCount() of OGR

Question

My question concerns OGR in Python3 on Windows. So I've installed gdal from the wheel file, and I started playing with OGR functions, but there's a problem I'm encountering that I can't understand. If GetFeatureCount() is called on a layer that was opened in a separate file Python stops working, but not if all my code is in the same script (reading & get count of features).

The code below works:

from osgeo import ogr

filename = '<path>'
driver = ogr.GetDriverByName("ESRI Shapefile")
datasource = driver.Open(filename)
layer = datasource.GetLayer()
print('Features count:', layer.GetFeatureCount())

But if I move the opening of the shapefile's layer to a separate file, it doesn't work anymore (Python stops working somehow):

class IO:
    @staticmethod
    def read_vector(filename):
        # load vector
        driver = ogr.GetDriverByName("ESRI Shapefile")
        datasource = driver.Open(filename)
        layer = datasource.GetLayer()

        return layer

And it called from this file:

from gdal_io import IO


filename =  '<path>'
layer = IO.read_vector(filename)
print('Features count:', layer.GetFeatureCount())

Is something necessary after reading the shapefile vector with OGR?

Answer 1

You have been bitten by a well known GDAL/OGR Python "gotcha".

Issue: Your datasource reference has gone out of scope (it's only referenced in the static method) and has orphaned your layer object which causes a crash.

Note: This isn't because you've opened the datasource from a class in an imported module, you could reproduce by simply defining a function in your main script that opens a datasource, gets a layer and returns the layer only. Or even by using something like layer = ogr.Open(filename).GetLayer(). Either way, the datasource gets dereferenced and python will crash if you try to use the orphaned layer object.

Solution: Return the datasource not the layer. Or use fiona, a much more pythonic interface to OGR.