Not compatible with numpy>=1.12

[grisaitis]

Numpy 1.12 dropped support for indexing with floats: https://docs.scipy.org/doc/numpy-1.12.0/release.html#deprecationwarning-to-error

I think malis does this in seg_to_affgraph... I'll provide more details later

[hueifang]

Hi, I ran test_malis.py and got the following:

Traceback (most recent call last):
File "test_malis.py", line 36, in
cc = m.connected_components(nVert,node1,node2,idxkeep)
File "malis/malis.pyx", line 51, in malis.malis.connected_components
(seg,segSizes) = prune_and_renum(seg,sizeThreshold)
File "malis/malis.pyx", line 80, in malis.malis.prune_and_renum
renum = np.zeros(int(segId.max()+1),dtype=np.uint64)
TypeError: 'numpy.float64' object cannot be interpreted as an integer

Numpy is 1.12.1. Do you have any idea as to how to fix it? Thanks!

[necrodancer]

I am suffering from the same problem as @hueifang for test_malis.py.
In the code as below:

renum = np.zeros(int(segId.max()+1),dtype=np.uint64)

It looks like there is no problem since segId.max()+1 is cast to integer.
But the error TypeError: 'numpy.float64' object cannot be interpreted as an integer occurs.
I think the error indicates that int() casting didn't work, or there is something wrong in higher version of Numpy.

p.s. My Numpy version is 1.14.5.

[hueifang]

@necrodancer: You can resolve this issue by building from source: ./make.sh. It appeared that the pip would install an old version that has the type error.

[necrodancer]

@hueifang
It worked after I built from `./make.sh'.
I really appreciate your comments. Thank you very much.

p.s. Should I close the issue?
Actually, this is the first time to ask sth in github.

[hueifang]

@necrodancer Glad it worked!

ps. hmm... I don't know. It's my first time, too. :)