take_1d yields surprising results when working with SparseArray

[hexgnu]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np
import pandas.core.algorithms as algos

algos.take_1d(pd.SparseArray([0,0,1], fill_value=0), [0,1,2]) #=> array([       1, 23618416,       32])

# VS

algos.take_1d(np.array([0,0,1]), [0,1,2]) #=> array([0,0,1])

Problem description

This to me smells like a problem with SparseArray sending over a sparse representation to take_1d in the C code. I would expect these values to be the same.

Expected Output

I would expect them to be the same.

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit: be09289 python: 3.6.1.final.0 python-bits: 64 OS: Linux OS-release: 4.13.16-202.fc26.x86_64 machine: x86_64 processor: x86_64 byteorder: little LC_ALL: en_US.UTF-8 LANG: en_US.UTF-8 LOCALE: en_US.UTF-8

pandas: 0.23.0.dev0+205.gbe0928903
pytest: 3.3.1
pip: 9.0.1
setuptools: 28.8.0
Cython: 0.27.3
numpy: 1.13.1
scipy: 0.19.1
pyarrow: 0.8.0
xarray: 0.10.0
IPython: 6.1.0
sphinx: 1.6.5
patsy: None
dateutil: 2.6.1
pytz: 2017.2
blosc: 1.5.1
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.4
feather: 0.4.0
matplotlib: 2.0.2
openpyxl: 2.4.9
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 1.0.2
lxml: 4.1.1
bs4: 4.6.0
html5lib: 1.0b10
sqlalchemy: 1.1.15
pymysql: 0.7.11.None
psycopg2: None
jinja2: 2.9.6
s3fs: 0.1.2
fastparquet: 0.1.3
pandas_gbq: None
pandas_datareader: None

[jorisvandenbossche]

Those algos are not meant to be used by users, so please don't use them directly. And they assume numpy arrays as input. Although a sparse array is a subclass, if you do np.array(sparse_array) you only get the actual values, without any fills. That is the reason that you see those strange values (they are leftovers of making an 'empty' result array to be filled by the take operation, but only the first item got filled).
There is an issue for that np.array(..) behaviour: #14167

[jorisvandenbossche]

Ah, I see now you were actually working on that issue I linked to :-)

[hexgnu]

Yea I had a feeling it was related. That problem is a lot harder than I expected on first take... have a PR that's somewhat done but a lot of weird edge cases.

[hexgnu]

Also I would never use algos as a user. I was just looking for advice on fixing something else related to merging sparse frames.

[jorisvandenbossche]

Also I would never use algos as a user.

Yes, sure! That's only not always clear from the issue post :)

[TomAugspurger]

Ideally algos.take_1d would dispatch to arr.take for SparseArray input. But we have some places in the library that seem to rely on take_1d(sparse_array) returning an ndarray.

[TomAugspurger]

This is fixed by #22325.