rtree intersection returns only zeros when filled with via stream from generator function

Question

I have trouble getting anything useful out of a rtree (Rtree==0.8.2) intersection. Instead of returning the indices/IDs that I passed into the index on creation, I get zeros.

I have some data:

for element in data:
    print(element)

POINT (497582.829135131 5419474.690534886)
POINT (565856.4271351419 5988338.580648975)
POINT (539734.212266123 5614400.468566706)
POINT (683523.3673841552 5398102.521964455)
POINT (633258.6526692112 5405182.75073655)
POINT (353675.3945317736 5747638.443753068)
POINT (480313.0993177936 5324783.150275591)
POINT (476802.8443284882 5390481.26880122)
POINT (392880.6204936574 5829710.142526797)
POINT (483370.4752331344 5792650.901288186)

I have an area of interest that spans over some of those points:

aoi = data[4].buffer(100000)
# GeometryCollection((aoi, box(*aoi.bounds), MultiPoint(data)))  # image below

I create an index based on rtree's stream load example and the documentation on the __init__ method:

def generator_function(geometries):
    for i, geometry in enumerate(geometries):
    # None instead of objects as I don't want to store the objects in the index
        yield (i, geometry.bounds, None)
idx = index.Index(generator_function(data))

When I query the index with the aoi's bounds, I get two hits but their IDs are both "0".

candidates = list(idx.intersection(aoi.bounds))
print("We got {} results".format(len(candidates)))
print("Their indexes are: {}".format(candidates))

We got 2 results
Their indexes are: [0, 0]

My problem: I expected to get the IDs as I passed them from the generator function. Why do I not get those?

If I fill the index step by step instead, I get the result I expected. I see no real logical difference in what should end up in the index though.

idx = index.Index()
for i, element in enumerate(data):
    idx.insert(i, element.bounds)

candidates = list(idx.intersection(aoi.bounds))
print("We got {} results".format(len(candidates)))
print("Their indexes are: {}".format(candidates))

We got 2 results
Their indexes are: [3, 4]

# GeometryCollection((box(*aoi.bounds), MultiPoint([data[i] for i in candidates])))  # Image below

Answer 1

Strange, because with Python 2.7.12 or 3.5.1; Shapely 1.5.16 and rtree 0.8.2, I have no problem

from shapely.geometry import Point
data = [Point(497582.829135131,5419474.690534886),Point(565856.4271351419,5988338.580648975),Point(539734.212266123,5614400.468566706),Point(683523.3673841552, 5398102.521964455),Point(633258.6526692112,5405182.75073655),Point(353675.3945317736,5747638.443753068) ,Point(480313.0993177936,5324783.150275591),Point(476802.8443284882, 5390481.26880122),Point(392880.6204936574,5829710.142526797) ,Point(483370.4752331344,5792650.901288186)]
aoi = data[4].buffer(100000)
from rtree import index
def generator_function(geometries):
    for i, geometry in enumerate(geometries):
        # None instead of objects as I don't want to store the objects in the index
        yield (i, geometry.bounds, None)
idx = index.Index(generator_function(data))
candidates = list(idx.intersection(aoi.bounds))
print("We got {} results".format(len(candidates)))
print("Their indexes are: {}".format(candidates))
print("the Points:",data[3].wkt, data[4].wkt)
We got 2 results
Their indexes are: [4, 3]
the Points: POINT (683523.3673841552 5398102.521964455) POINT (633258.6526692112 5405182.75073655)

What are your versions ?

(edit for 6 char limit)

Answer 2

It looks like this was a bug that has been fixed in rtree 0.8.3.

https://github.com/Toblerity/rtree/issues/71

$ python2 minimalexample.py 
We got 2 results
Their indexes are: [0, 0]

$ pip2 freeze | grep Rtree
Rtree==0.8.2

$ pip2 install --user rtree==0.8.3
(...)
Installing collected packages: rtree
  Found existing installation: Rtree 0.8.2
    Uninstalling Rtree-0.8.2:
      Successfully uninstalled Rtree-0.8.2
Successfully installed rtree-0.8.3
$ python2 minimalexample.py 

We got 2 results
Their indexes are: [4, 3]