(Hopefully I did not find this issue in the list, because there was none...) I have experienced the following issue for a while (wonder if it ever worked actually):
I create a Series or a DataFrame with well-defined, floating-point indices. Then I try to slice some rows using a list. For some valid index values I get NaN instead of the value itself. An example:
>>> import numpy as np
>>> import pandas as pd
>>> i = np.linspace(0, 10, 101) # create a floating-point (i)ndex-range
>>> d = np.random.random(101) # add some random numbers as (d)ata
>>> s = pd.Series(d, index=i) # create a Series from the generated data
>>> type(s)
<class 'pandas.core.series.Series'>
>>> s.head()
0.0 0.969683
0.1 0.463655
0.2 0.879788
0.3 0.122202
0.4 0.466166
dtype: float64
>>> xs = [0.8, 1.1, 1.9, 2.2, 3.3] # we make a list with a few indexes of interest
>>> s[xs] # try to list Series-values for those indices, but for some we see only NaN!
0.8 0.996011
1.1 0.699921
1.9 NaN
2.2 0.919059
3.3 NaN
dtype: float64
R:\WinPython\python-3.4.3.amd64\lib\site-packages\pandas\core\format.py:2012: RuntimeWarning: invalid value encountered in greater
has_large_values = (abs_vals > 1e8).any()
R:\WinPython\python-3.4.3.amd64\lib\site-packages\pandas\core\format.py:2013: RuntimeWarning: invalid value encountered in less
has_small_values = ((abs_vals < 10 ** (-self.digits+1)) &
R:\WinPython\python-3.4.3.amd64\lib\site-packages\pandas\core\format.py:2014: RuntimeWarning: invalid value encountered in greater
(abs_vals > 0)).any()
>>> s[1:2] # we can see, that the values are really there, also at index 1.9
1.0 0.905714
1.1 0.699921
1.2 0.105968
1.3 0.809368
1.4 0.227002
1.5 0.825698
1.6 0.911150
1.7 0.027182
1.8 0.046001
1.9 0.456927
2.0 0.592820
dtype: float64
>>>
The warnings are actually came from the console, the IPython Notebook that I normally use does not even issue them. Same result using e.g. s.loc[xs]. The problem also exists for DataFrames, which can be reproduced in the same manner.
I'm using the latest 64-bit WinPython with Python 3.4.3 and Pandas 0.16.1:
>>> pd.show_versions()
INSTALLED VERSIONS
------------------
commit: None
python: 3.4.3.final.0
python-bits: 64
OS: Windows
OS-release: 7
machine: AMD64
processor: Intel64 Family 6 Model 42 Stepping 7, GenuineIntel
byteorder: little
LC_ALL: None
LANG: hu_HU
pandas: 0.16.1
nose: 1.3.6
Cython: 0.22
numpy: 1.9.2
scipy: 0.15.1
statsmodels: 0.6.1
IPython: 3.1.0
sphinx: 1.3.1
patsy: 0.3.0
dateutil: 2.4.2
pytz: 2015.2
bottleneck: None
tables: 3.2.0
numexpr: 2.4.3
matplotlib: 1.4.3
openpyxl: None
xlrd: 0.9.3
xlwt: None
xlsxwriter: 0.7.2
lxml: None
bs4: 4.3.2
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: 1.0.4
pymysql: None
psycopg2: None
Comment From: jorisvandenbossche
This is a floating point issue:
In [72]: s.index[19]
Out[72]: 1.9000000000000001
In [74]: xs[2]
Out[74]: 1.9
In [76]: s.index[19] == xs[2]
Out[76]: False
See #9817.
Comment From: jorisvandenbossche
Closing as a duplicate of #9817
Comment From: cousin333
I also suspected, that the issue is about floating point numbers. Thanks for linking it to a seemingly unconnected issue, I would never find that one. Actually they really have the same root.