avoid series constructor

Author: TomAugspurger

pandas/core/reshape/reshape.py

@@ -853,6 +853,7 @@ def check_len(item, name):
 
 def _get_dummies_1d(data, prefix, prefix_sep='_', dummy_na=False,
  sparse=False, drop_first=False, dtype=None):
+ from pandas.core.reshape.concat import concat
  # Series avoids inconsistent NaN handling
  codes, levels = _factorize_from_iterable(Series(data))
 
@@ -909,7 +910,7 @@ def _make_col_name(prefix, prefix_sep, level):
  index = None
 
  if sparse:
- sparse_series = {}
+ sparse_series = []
  N = len(data)
  sp_indices = [[] for _ in range(len(dummy_cols))]
  mask = codes != -1
@@ -928,10 +929,9 @@ def _make_col_name(prefix, prefix_sep, level):
  sarr = SparseArray(np.ones(len(ixs), dtype=dtype),
  sparse_index=IntIndex(N, ixs), fill_value=0,
  dtype=dtype)
- sparse_series[col] = Series(data=sarr, index=index)
+ sparse_series.append(Series(data=sarr, index=index, name=col))
 
- out = DataFrame(sparse_series, index=index, columns=dummy_cols,
- dtype=dtype)
+ out = concat(sparse_series, axis=1, copy=False)
  return out
 
  else: