SparseDataFrame from_dict doesn't adhere to sparse input semantics

[matanox]

Code Sample

from pandas import DataFrame, SparseDataFrame

list_of_sparse_semantics = [{0: 1, 4: 1, 22: 1, 1001: 2}, {0: 1, 3: 1}, {0: 1, 55: 1}]

df = DataFrame.from_dict(list_of_sparse_semantics)
sdf = SparseDataFrame.from_dict(list_of_sparse_semantics)

Problem description

It may seem as if SparseDataFrame does not follow the sparse semantics that are followed by DataFrame's from_dict method. It stuffs each dict as a single column/element, rather than treating its key-val tuples as sparse content for building the equivalent sparse row as DataFrame does.

Expected Output

sdf.equals(df)

True

Output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 3.6.3.final.0
python-bits: 64
OS: Linux
OS-release: 4.4.0-130-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.20.3
pytest: 3.2.1
pip: 10.0.1
setuptools: 36.5.0.post20170921
Cython: 0.26.1
numpy: 1.14.3
scipy: 0.19.1
xarray: None
IPython: 6.4.0
sphinx: 1.6.3
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2017.2
blosc: None
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.2
feather: None
matplotlib: 2.1.0
openpyxl: 2.4.8
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 1.0.2
lxml: 3.8.0
bs4: 4.6.0
html5lib: 0.999999999
sqlalchemy: 1.1.13
pymysql: None
psycopg2: None
jinja2: 2.8.1
s3fs: None
pandas_gbq: None
pandas_datareader: None

[matanox]

per my availability when this is ever responded to, happy to contribute a fix/feature for this, if given fair guidance as to where to touch and what to look out for 😃

[matanox]

Also, the following (a workaround) seems to work slower than one could hope for, given the input is already sparse when it arrives at the SparseDataFrame.... but might boil down to my specific scipy usage strategy below.

from scipy.sparse import dok_matrix, coo_matrix
from pandas import SparseDataFrame

%time i, j, data = zip(*((i, t[0], t[1]) for i, row in enumerate(a) for t in row))

%time m = coo_matrix((data, (i, j)), shape=(num_of_docs, lexicon_size))

%time sdf = SparseDataFrame(m).fillna(0)

CPU times: user 80 ms, sys: 4 ms, total: 84 ms
Wall time: 84.5 ms
CPU times: user 28 ms, sys: 4 ms, total: 32 ms
Wall time: 28.5 ms
CPU times: user 41.9 s, sys: 132 ms, total: 42 s
Wall time: 41.8 s