Description
Example Code
print
666A
(pd.DataFrame({'a': np.zeros(1000, 'V4')}))
Results
Non-deterministic behavior. Sometimes you get all zeros, sometimes you get garbage like this:
a
0 [1, 0, 0, 0]
1 [0, 0, 0, 0]
2 [8, 0, 0, 0]
3 [0, 0, 0, 0]
4 [48, -92, 71, -60]
5 [-27, 127, 0, 0]
6 [5, 0, 0, 0]
7 [0, 0, 0, 0]
...
That is despite the fact that the bytes are actually all zero, and NumPy prints all rows as [0, 0, 0, 0].
Sometimes when printing a wider DataFrame containing such a column, it segfaults with this stack trace:
#0 BYTE_copyswap (dst=0x7fffebdbe868, src=0x7fffe9f0d054, __NPY_UNUSED_TAGGEDswap=0, __NPY_UNUSED_TAGGEDarr=0x7fffea15f990)
at numpy/core/src/multiarray/arraytypes.c.src:1911
#1 0x00007ffff5d82075 in PyArray_Scalar (data=0x7fffe9f0d054, descr=0x7fffea1612d0, base=<optimized out>) at numpy/core/src/multiarray/scalarapi.c:835
#2 0x00007ffff5d52df0 in array_item (self=0x7fffea15f990, i=<optimized out>) at numpy/core/src/multiarray/mapping.c:1371
Expected Output
All rows [0, 0, 0, 0] - just as NumPy prints it.
Output of pd.show_versions()
## INSTALLED VERSIONS
commit: None
python: 3.5.1
python-bits: 64
OS: Linux
OS-release: 3.13.0
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
pandas: 0.18.1
pip: 8.1.2
setuptools: 27.2.0
numpy: 1.11.1