More robust DMatrix creation from a sparse matrix (#1606)

* [CORE] DMatrix from sparse w/ explicit #col #row; safer arg types

* [python-package] c-api change for _init_from_csr _init_from_csc

* fix spaces

* [R-package] adopt the new XGDMatrixCreateFromCSCEx interface

* [CORE] redirect old sparse creators to new ones

tests/python/test_sparse_dmatrix.py

@@ -0,0 +1,33 @@
+import numpy as np
+import xgboost as xgb
+from scipy.sparse import rand
+
+rng = np.random.RandomState(1)
+
+param = {'max_depth': 3, 'objective': 'binary:logistic', 'silent': 1}
+
+
+def test_sparse_dmatrix_csr():
+    nrow = 100
+    ncol = 1000
+    x = rand(nrow, ncol, density=0.0005, format='csr', random_state=rng)
+    assert x.indices.max() < ncol - 1
+    x.data[:] = 1
+    dtrain = xgb.DMatrix(x, label=np.random.binomial(1, 0.3, nrow))
+    assert (dtrain.num_row(), dtrain.num_col()) == (nrow, ncol)
+    watchlist = [(dtrain, 'train')]
+    bst = xgb.train(param, dtrain, 5, watchlist)
+    bst.predict(dtrain)
+
+
+def test_sparse_dmatrix_csc():
+    nrow = 1000
+    ncol = 100
+    x = rand(nrow, ncol, density=0.0005, format='csc', random_state=rng)
+    assert x.indices.max() < nrow - 1
+    x.data[:] = 1
+    dtrain = xgb.DMatrix(x, label=np.random.binomial(1, 0.3, nrow))
+    assert (dtrain.num_row(), dtrain.num_col()) == (nrow, ncol)
+    watchlist = [(dtrain, 'train')]
+    bst = xgb.train(param, dtrain, 5, watchlist)
+    bst.predict(dtrain)