Python3.6 and wanish issues

Question

I am trying to run the Wanish library in python and getting the following instead of the result expected.

>>> from wanish import Wanish
>>> wanish = Wanish()
>>> wanish.perform_url("http://www.bbc.com/news/uk-england-london-40269625")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\__init__.py", line 167, in perform_url
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\summarizer.py", line 55, in get_plain_text
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\summarizer.py", line 91, in create_referat
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\summarizer.py", line 68, in textrank
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\langid.py", line 175, in classify
  File "C:\Python36\lib\site-packages\wanish-0.6.3-py3.6.egg\wanish\langid.py", line 148, in instance2fv
TypeError: 'float' object cannot be interpreted as an integer

I have tried this using Python 3.6 and even with python 2.7

I came across the same issues.

Answer 1

Actually if you visit pypi wanish page you will see that it located only in Python 3.x categories:

Programming Language :: Python

Programming Language :: Python :: 3.3

Programming Language :: Python :: 3.4

Programming Language :: Python :: 3.5

It won't work in Python 2 because in code there is from urllib.parse import urlparse, urljoin line. But from docs:

The urlparse module is renamed to urllib.parse in Python 3

Answer 2

The offending line is this one* in LanguageIdentifier.instance2fv() :

arr = np.zeros((self.nb_numfeats,), dtype='uint32')

In LanguageIdentifier.from_modelstring() , the factory method used to instantiate this class, you can find this:

nb_numfeats = len(nb_ptc) / len(nb_pc)

In python3, division using the / operator always returns a float, where in python2 it returns an int if both operants are ints. So this is a bug.

The langid.py file seems to actually come from a different project, where this error was fixed a while ago .

---

*I would normally link to the code directly, but someone decided to drop a ~2.4MB blob of base64 in the source file, so github doesn't allow linking.