Support sparse input in interfaces_getitem when num_best is not None

[tmylk]

Sparse input is ony supported when num_best=None, otherwise errors.
Raised on the mailing list https://groups.google.com/d/msg/gensim/-PhUHorj9-E/N6apxcqnHQAJ

from scipy.sparse import random
from gensim.similarities import SparseMatrixSimilarity
from gensim.matutils import Sparse2Corpus

X = random(200e3, 15e3, density=2e-4, format="csc")
index = SparseMatrixSimilarity(Sparse2Corpus(X, documents_columns=False), num_features=X.shape[1], num_terms=X.shape[0], maintain_sparsity=True, num_best=100)

for s in index:
    print s

returns

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-2-04d8dd3277db> in <module>()
----> 1 for s in index:
      2         print s
      3 

/home/lev/miniconda2/lib/python2.7/site-packages/gensim/interfaces.pyc in __iter__(self)
    264                 chunk_end = min(self.index.shape[0], chunk_start + self.chunksize)
    265                 chunk = self.index[chunk_start : chunk_end]
--> 266                 for sim in self[chunk]:
    267                     yield sim
    268         else:

/home/lev/miniconda2/lib/python2.7/site-packages/gensim/interfaces.pyc in __getitem__(self, query)
    226         # most similar for each document in turn
    227         if matutils.ismatrix(result):
--> 228             return [matutils.full2sparse_clipped(v, self.num_best) for v in result]
    229         else:
    230             # otherwise, return top-n of the single input document

/home/lev/miniconda2/lib/python2.7/site-packages/gensim/matutils.pyc in full2sparse_clipped(vec, topn, eps)
    241     if topn <= 0:
    242         return []
--> 243     vec = np.asarray(vec, dtype=float)
    244     nnz = np.nonzero(abs(vec) > eps)[0]
    245     biggest = nnz.take(argsort(abs(vec).take(nnz), topn, reverse=True))

/home/lev/miniconda2/lib/python2.7/site-packages/numpy/core/numeric.pyc in asarray(a, dtype, order)
    529 
    530     """
--> 531     return array(a, dtype, copy=False, order=order)
    532 
    533 

ValueError: setting an array element with a sequence.

[souravsingh]

@tmylk I am interested in working on the issue. We basically have to add a check for num_best, correct?

[tmylk]

@souravsingh Need to add a new code path when num_best is not None and input is sparse

[manneshiva]

@tmylk I have fixed this issue and am submitting a PR for the same. Instead of having a separate code path for sparse input and num_best not None, I have added another function, any2sparse_clipped() in matutils and replaced full2sparse_clipped with this function in interfaces.