Importing Pandas gives error AttributeError: module 'pandas' has no attribute 'core' in iPython Notebook
I am running a iPython notebook via the Anaconda Navigator app (version 1.1.0). When I want to import pandas it gives me a strange error. I thought the Anaconda app included the pandas package?
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-4-af55e7023913> in <module>()
----> 1 import pandas as pd
/Users/bertcarremans/anaconda/lib/python3.5/site-packages/pandas/__init__.py in <module>()
37 import pandas.core.config_init
38
---> 39 from pandas.core.api import *
40 from pandas.sparse.api import *
41 from pandas.stats.api import *
/Users/bertcarremans/anaconda/lib/python3.5/site-packages/pandas/core/api.py in <module>()
8 from pandas.core.common import isnull, notnull
9 from pandas.core.categorical import Categorical
---> 10 from pandas.core.groupby import Grouper
11 from pandas.core.format import set_eng_float_format
12 from pandas.core.index import (Index, CategoricalIndex, Int64Index,
/Users/bertcarremans/anaconda/lib/python3.5/site-packages/pandas/core/groupby.py in <module>()
16 DataError, SpecificationError)
17 from pandas.core.categorical import Categorical
---> 18 from pandas.core.frame import DataFrame
19 from pandas.core.generic import NDFrame
20 from pandas.core.index import (Index, MultiIndex, CategoricalIndex,
/Users/bertcarremans/anaconda/lib/python3.5/site-packages/pandas/core/frame.py in <module>()
37 create_block_manager_from_arrays,
38 create_block_manager_from_blocks)
---> 39 from pandas.core.series import Series
40 from pandas.core.categorical import Categorical
41 import pandas.computation.expressions as expressions
/Users/bertcarremans/anaconda/lib/python3.5/site-packages/pandas/core/series.py in <module>()
33 from pandas.core.internals import SingleBlockManager
34 from pandas.core.categorical import Categorical, CategoricalAccessor
---> 35 import pandas.core.strings as strings
36 from pandas.tseries.common import (maybe_to_datetimelike,
37 CombinedDatetimelikeProperties)
AttributeError: module 'pandas' has no attribute 'core'
Solution 1:
"Have you tried turning it off and on again?" (Roy of The IT crowd)
This happened to me today, which is why I ended up to this page. Seeing that error was weird since, recently, I have not made any changes in my Python environment. Interestingly, I observed that if I open a new notebook and import pandas
I would not get the same error message. So, I did shutdown
the troublesome notebook and started it again and voila it is working again!
Even though this solved the problem (at least for me), I cannot readily come up with an explanation as to why it happened in the first place!
Solution 2:
There's this bug in the latest version of pandas (pandas 0.23) that gives you an error on importing pandas.
But this can be easily fixed by installing an earlier version of pandas (pandas 0.22) using the command pip install pandas==0.22
on Windows Command Prompt.
Solution 3:
Apparently the error I got when trying to import pandas for the first time was ValueError: unknown locale: UTF-8
Trying to import again afterwards, gave another error as described in my question above.
I found the solution to solve the ValueError on IPython Notebook locale error
After updating my bash profile, the error AttributeError: module 'pandas' has no attribute 'core'
did not appear anymore.
Solution 4:
There is an other weird reason this happens. If you have a file called pandas.py or a directory called pandas in the same or nested levels, that library is used instead and fails to work. Rename the folder and restart the env and it started working. Faced this
Solution 5:
I had a similar issue since I installed pandas
using python -m pip install pandas --upgrade --user
which installed a conflicting version in my user python packages directory, masking the Anaconda installed version that other dependencies relied upon.
conda list | grep pandas
pandas == 0.23.4
python -m pip list | grep pandas
pandas == 0.24.0
So uninstalling the user directory masked version cleaned up the issue for me.
python -m pip uninstall pandas
For reference all possible python packages are installed in the directories listed from this command:
python -m site
Might be worth iterating through these and checking for duplicates.
Edit: Since my original answer I learnt you can run:
python -m pip list -v
And it shows the directory the library is installed. This often shows whether the library you want is in a virtual environment, conda environment, user directory, system site packages etc.