Build numpy with ucs4 issue - linux

I've installed successfully python2.7 with ucs4 format.
If I install the numpy with: sudo pip install --upgrade numpy, it works and I can use it, but installing by this way only put numpy in my dist-packages, however it's needed to be in site-packages. (because Jepp can find them only there)
Thus I downloaded the numpy tar file, I buildt it (sudo python setup.py build) and installed it (sudo python setup.py install), but it gives me undefined symbol: PyUnicodeUCS2_FromUnicode. It seems for me, it installed numpy with ucs2.
How could I force to install numpy with ucs4?
(I've read this article: Numpy needs the ucs2 but EPD is not the solution for me)
Any idea may help me!
Thanks in advance.

I simply copied the installed files from dist-packages to site packages. If someone has a better solution, please inform me.

Related

PIP UPGRADE - Trouble with versions in personal setup package

I'm working on my first python package, so I'm new with all this.
I have my own setup package which I install through pip install git+<URL>.git.
When I try to upgrade it with the --upgrade flag, it always get reinstalled, even if I haven't changed the version in my setup.py file.
Does anyone know why this is happening? o maybe have a better approach to this?
I have made my research before asking, but I think this is a very specific problem.
I have tried in a venv and out.
python 3.7, pip 19.2, setuptools 40.8
# setup.py
from test import __version__
from setuptools import find_packages, setup
setup(
name="test",
version=__version__,
packages=find_packages(),
)
I expect to get something similar to:
Requirement already up-to-date: test in ./<cwd>/.venv/lib/python3.7/site-packages (0.0.1)
but instead I get this:
Installing collected packages: test
Found existing installation: test 0.0.1
Uninstalling test-0.0.1:
Successfully uninstalled test-0.0.1
Running setup.py install for test ... done
Successfully installed test-0.0.1
Reading the documentation it seems like the expected behaviour to me.
Note that if a satisfactory version of the package is already installed, the VCS source will not overwrite it without an --upgrade flag.
Read a somewhat related discussion here:
https://github.com/pypa/pip/issues/2837
Looks like pip does upgrades in a much more binary fashion when VCS is involved. It seems to always upgrade when --upgrade is set and never when it's not set.
You could try the work-around that is suggested in this discussion, but be aware of its side effects (for example, you will have a full clone of the git repository on your file system):
pip install --editable --upgrade git://somewhere/something.git

<Import statement issue>ImportError: Missing required dependencies ['numpy'] [duplicate]

Since yesterday I've had this error when I try to import packages on anaconda :
ImportError: Missing required dependencies ['numpy']
I have tried un-installing Anaconda and Python, switching to Python 2.7 but nothing works it's still the same error, here is the code I get :
Any help is really appreciated thanks !
I had this same issue immediately after upgrading pandas to 0.19.2. I fixed it with the following install/uninstall sequence from the windows cmd line:
pip uninstall pandas -y
pip uninstall numpy -y
pip install pandas
pip install numpy
This also broke my matplotlib install so I uninstalled/installed that as well.
Very odd behavior for a seemingly routine upgrade.
What happens if you try to import numpy?
Have you tried'
pip install --upgrade numpy
pip install --upgrade pandas
I had to install this other package:
sudo apt-get install libatlas-base-dev
Seems like it is a dependency for numpy but the pip or apt-get don't install it automatically for whatever reason.
I had this problem with last version of numpy 1.16.x
Problem resolved with
python3 -m pip uninstall numpy
python3 -m pip install numpy==1.14.0
Did you install miniconda and pandas without dependencies?
Try installing numpy first with conda install numpy or pip install numpy.
If you're on Windows you can get pre-compiled versions of most libraries that require compilation from here.
On Windows 10 Anaconda3-5.3.0-Windows-x86_64 I had the Missing required dependencies ['numpy'] error when running scripts as so, %HOMEPATH%\AppData\Local\Continuum\anaconda3\python.exe pandas_script_foo.py.
In my case the error was caused by missing Anaconda package PATH definitions when running Anaconda python.exe in a windows cmd.exe session. The numpy package is not missing. It just can't be found on the PATH.
The Anaconda installation includes windows shortcuts that give examples of configuring the PATH per script run. See the shortcuts in the %HOMEPATH%\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Anaconda3 (64-bit) directory for examples.
See the %HOMEPATH%\AppData\Local\Continuum\anaconda3\cwp.py script to see how Anaconda configures PATH.
Below is an example windows BAT file that calls cwp.py to setup PATH, and then run a python script. Its a copy of the commands the Anaconda jupyter-lab shortcut executes.
%HOMEPATH%\AppData\Local\Continuum\anaconda3\python.exe ^
%HOMEPATH%\AppData\Local\Continuum\anaconda3\cwp.py ^
%HOMEPATH%\AppData\Local\Continuum\anaconda3 ^
%HOMEPATH%\AppData\Local\Continuum\anaconda3\python.exe ^
%HOMEPATH%\AppData\Local\Continuum\anaconda3\Scripts\jupyter-lab-script.py
If you need to execute python scripts on Anaconda with the conveniance of running a BAT file, the above BAT file example should do the trick.
The data manipulation capabilities of pandas are built on top of the numpy library. In a way, numpy is a dependency of the pandas library. If you want to use pandas, you have to make sure you also have numpy. When you install pandas using pip, it automatically installs numpy. If it doesn't, try the following
pip install -U numpy pandas
For conda
conda install numpy pandas
I also faced the same issue. It happened to me after I upgraded my numpy library.
It was resolved in my case by upgrading my pandas library as well after upgrading my numpy library using the below command:
pip install --upgrade pandas
Try:
sudo apt-get install libatlas-base-dev
It should work now.
Else, try uninstall and reinstall numpy and pandas.
I had the same issue. It was because I had multiple versions of numpy installed. Remove all versions by repeatedly using:
pip uninstall numpy
Then re-install it with the command:
pip install numpy
First, try to import numpy on it's own, like so:
import numpy as np
I got this message:
ImportError: Something is wrong with the numpy installation. While importing
we detected an older version of numpy in
['/home/michael/.local/lib/python3.6/site-packages/numpy']. One method of
fixing this is to repeatedly uninstall numpy until none is found, then
reinstall this version.
So do what it says, keep uninstalling numpy until there is none, and then reinstall.
This worked for me.
I had the same issue while using Microsoft Visual Code with Python 3.7.3 64-bit('base':conda)as my python interpreter. Before running any code type the following three commands:
C:/ProgramData/Anaconda3/Scripts/activate #activate conda Scripts directory
conda activate base #activate conda
& C:/ProgramData/Anaconda3/python.exe #to run python
I have same problem.
I have got two version of numpy 1.16.6 and 1.15.4, fresh installed pandas did not work correctly.
I fixed it by uninstalling all versions of numpy and pandas and install the last versions.
$ pip uninstall numpy pandas -y
Uninstalling numpy-1.16.6:
Successfully uninstalled numpy-1.16.6
Uninstalling pandas-0.24.2:
Successfully uninstalled pandas-0.24.2
$ pip uninstall numpy pandas -y
Uninstalling numpy-1.15.4:
Successfully uninstalled numpy-1.15.4
Cannot uninstall requirement pandas, not installed
$ pip uninstall numpy pandas -y
Cannot uninstall requirement numpy, not installed
$ pip install numpy pandas
I had the same issue with anaconda package, it got updated.
anaconda {4.3.1 -> custom} ## I am not sure if this was the issue
Hit below command to know
conda list --revisions
what i did is just uninstall pandas with conda and re-install it
conda install pandas
Some new libs may also get installed with it.
It worked for me hope will do the same for you.
Uninstall all pip packages that you're having problems with. Manually remove all site-packages files. If you're using MacPorts, sudo port clean .
Then try reinstalling. Sometimes, there are files that should have been removed, but weren't if the installation was abruptly interrupted or something.
There could be an issue with conflicting versions of the package(s), as well as potentially issues with Pathing. Are you sure you've set the correct Path for your binaries? (/opt/local/bin, /anaconda2/bin, etc.)
Another issue could be some PYTHONPATH that's explicitly looking in the wrong place for the file.
I had a same issue recently with Anaconda with Python 3.7.
I solved this problem by downgrading python version to 3.6:
conda install python=3.6
and then by updating all the packages:
conda update --all
pandas is built on top of numpy so you need to have numpy to use the data manipulation feature, so install numpy first.
pip install numpy
This worked in my anaconda environment, but I do not know why conda does not work. For some reason conda uninstall was not sufficient. This only worked with conda remove.
conda remove pandas
conda remove numpy
conda install pip
pip install pandas
*With help from this answer
This raises the following import warning in python 3.6 and 3.7:
ImportWarning: can't resolve package from __spec__ or __package__, falling back on __name__ and __path__
If you with to ignore this warning (and maybe other ImportWarnings), add the following to your script before importing pandas:
import warnings
warnings.filterwarnings('ignore', category=ImportWarning, module='_bootstrap.py')
In my case even though I was using the above options of uninstall and installing using pip the code was still giving me same errors.
Finally, I created a vritual environment and Installed numpy and pandas using pip in my virtual env. Now the code is running.
Steps: for Anaconda3 - Please change according to your installation type:
[if you dont have virtual env package installed]
$ pip install virtualenv
[from command prompt go to the directory by c:\anadonda3\scripts
[write the following command to use virtual env to create a virtual env for you in your desired location]
$virtualenv c:\anaconda3\envs\my_virtual_env
[once created you will have to activate your virtual env]
$c:\anaconda3\envs\my_virtual_env\scripts activate
[now pip install numpy and pandas and other required packages using pip]
[once installations are done exit from the virtual env]
$c:\anaconda3\envs\my_virtual_env\scripts deactivate
now use the python.exe inside your virtual env folder to run the script and it will run even with python 3.7.
I am using Win10 and Conda, and this issue just append to me when upgrading python 3.7.2-h8c8aaf0_0 --> 3.7.2-h8c8aaf0_2.
I solved it by return to the previous version with
conda install python=3.7.2=h8c8aaf0_0
If you're running your program on PyCharm on Windows, there is a known bug, because PyCharm simply doesn't add env-related paths to PATH.
The issue is fixed in the 2019.1 Early Access Preview (EAP) build.
For me installing the EAP fixed the issue.
nothing worked for me ... except when I found this
I suspect that you have a local file called unittest.py that is getting imported instead of the standard module.
I was trying to upgrade my anaconda 2 with anaconda 3. I tried installing Anaconda3-2018.12-Windows-x86 and Anaconda3-2019.03-Windows-x86_64 on my Windows 10 machine and failed with this error. For me, using Anaconda3-4.4.0-Windows-x86_64 for anaconda 3 worked the trick after trying everything listed in answers here.
I fixed this using Anaconda by going to Environments > base(root), searching for numpy in the installed modules and clicking the tickbox alongside it and choosing > Mark for specific version installation > 1.14.0 (as suggested by another user on this thread). Then clicking Apply. Once it downgraded numpy I stopped getting errors when running py files on the command line.
Throughout this saga, I was still able to use https://pypi.org/project/auto-py-to-exe/ even when I was getting the numpy errors on the command line, but it was a hassle to create an exe every time I wanted to test a change. It's all sorted now. I guess there was a problem with numpy 1.16.4.
Anyway, I hope this helps someone who's using Anaconda as well.
The following worked for me.
Deleted the folders for numpy and pandas together with their contents completely from the site-packages folder. Check depending on whether you are using python2 or python3. Check the exact path as per your machine.
N.B handle with care "rm -rf" command. If you are not sure of what you are doing, please do it manually using any file explorer of your choice!!
rm -rf ~/anaconda2/envs/myenv/lib/pythonX/site-packages/pandas*
rm -rf ~/anaconda2/envs/myenv/lib/pythonX/site-packages/numpy*
Then i installed clean packages for pandas and numpy as usual with
pip install numpy
pip install pandas
I've got the same error recently. Before applying uninstall or install tools, try to update your Jupyter.
How? Go to 'Environments' and type on the Search Packages box 'pandas'. Afterwards, check the version (if that column shows a blue number with a diagonal arrow, it means that your pandas is out of date). Click on 'pandas' and a option will pop up (choose 'Apply' and wait for a couple of minutes to update the package). And then, make a quick test on any notebook to make sure that your Jupyter is running smoothly.
For those who couldn't solve with the above answers:
Ensure that you are running python3 with
$ python version
If not, install python3.
Then change default python to python3 with
$ alias python=python3
Next, close your jupyter lab/notebook environment and re-launch it with default python being python3.
build_exe_options = {"packages": ["os",'pandas','numpy']}
It works.
you are running python 3.7
create environment for python 3.6
python3.6 filename.py

python install wheel leads to import error

I'd like to make a wheel binary distribution, intstall it and then import it in python. My steps are
I first create the wheel: python ./my_package/setup.py bdist_wheel
I install the wheel: pip install ./dist/*.whl
I try to import the package: python -c"import my_package"
This leads to the error:
ImportError: No module named 'my_package'
Also, when I do pip list, the my_package is listed.
However, when I run which my_packge, nothing is shown.
When I run pip install ./my_package/ everything works as expected.
How would I correctly build and install a wheel?
python version 3.5
pip version 10.1
wheel version 0.31.1
UPDATE:
When I look at the files inside my_package-1.0.0.dist-info, there is an unexpected entry in top_level.txt. It is the name of the folder where I ran
python ./my_package/setup.py bdist_wheel in. I believe my setup.py is broken.
UPDATE WITH REGARDS TO ACCEPTED ANSWER:
I accepted the answer below. Yet, I think it is better to simply cd into the package directory. Changing to a different directory as suggested below leads to unexpected behavior when using the -d flag, i.e. the target directory where to save the wheel. This would be relative to the directory specified in the setup.py file.
I had the very same error, but it was due to my setup.py not specifying the entry "packages=setuptools.find_packages()".
Everythings builds nicely without that but you can't import anything even though pip shows it to be installed.
If you need to execute the setup script from another directory, ensure you are entering the project dir in the script.
from setuptools import setup
root = os.path.normpath(os.path.join(os.path.abspath(__file__), os.pardir))
os.chdir(root)
# or using pathlib (Python>=3.4):
import patlib
root = pathlib.Path(__file__).parent
os.chdir(str(root))
setup(...)
In my case, in order to solve it I just had to upgrade pip (since Docker installed pip 9).
python3 -m pip install --upgrade pip
I have experienced the same situation, maybe not for the same reason, here just for reference.
The package name should not contain the dash "-", there's no error pop out, but after installing your wheel, though it is shown in pip list, you can't find that package.
/src/your-package-name # should not
/src/your_package_name # should like this
In the setup.py, you can use the name with dash "-" without limitation:
setuptools.setup(
name="instrument-lab",
...

ImportError:'save_weights' requires h5py

When I save weights during training my CNN model using keras, it says ImportError:'save_weights' requires h5py, but I have already installed h5py.
I would greatly appreciate if someone could explain how to fix this issue.
Just install necessary packages
sudo apt-get install libhdf5-dev
pip install h5py
If you are using windows and python IDE, open cmd and input following commands:
pip install h5py
pip install cython
I hope it helps.
I was getting the same error as you.
I installed all the requirements listed here: https://github.com/fchollet/keras/issues/3426
Finally just needed to reboot and it started working.
As suggested by others:
pip install h5py
Note that this may not immediately resolve the issue in your active session and you may need to reload keras.models either through the following commands or by just creating a new session/re-opening your jupyter notebook.
In Python3:
from importlib import reload
reload(keras.models)
In Python2:
use importlib.import_module instead. See docs for a reference.
These additional steps may be necessary because of the try/except ImportError in keras sourcecode that assigns h5py = None when it's unable to locate it the first time it's executed.
In my case, re-installing did the trick:
pip uninstall -y cython h5py
pip install cython h5py
(Windows 10, Conda, Keras 2.4.3)
I think you may miss this
from keras.applications import imagenet_utils
I got the same problem even though I have imported the h5py.
It is the load error with the keras. It has to be reloaded.
import keras
from importlib import reload
reload(keras.models)
It has worked for me.
h5py==2.10.0 works well with TF >= 2.1 so try 'pip install h5py==2.10.0'
Have you tried directly installing h5py? http://docs.h5py.org/en/latest/build.html
Try running:
pip install h5py
or
sudo apt-get install libhdf5

Install NumPy for Python 3.5

I am using python 3.5 and I am trying to install NumPy but when I try to install from command prompt using command: pip install numpy
I get a whole lot of errors. The main error though seems to be at the end: error: Unable to find vcvarsall.bat
I have also tried downloading the numpy binaries from http://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy and installing using command: pip install numpy-1.10.4+mkl-cp35-cp35m-win_amd64.whl
But I get an error that says: numpy-1.10.4+mkl-cp35-cp35m-win_amd64.whl is not a supported wheel on this platform.
Then I found a numpy-1.10.2-win32-superpack-python3.4.exe file here https://sourceforge.net/projects/numpy/files/NumPy/1.10.2/ but when I try to run that I get an error that I require Python version 3.4.
I looked and cannot find an installer for python 3.5.exe, do I need to uninstall python 3.5 and install 3.4 instead? Or is there anyway to install numpy for Python 3.5?
I am using Windows 10/64 bit.
An easy way to obtain numpy, scipy, pandas, ... is to install anaconda.
Anaconda will automatically download and install the latest modules.
https://www.continuum.io/downloads
I hope this will help you along.

Resources