docker cannot build scipy

Question

I am trying to create a docker container for my python application (this is the first time I am trying to use docker). I looked at the online tutorials and created a DockerFile as follows:

FROM python:3.6-alpine
COPY . /app
WORKDIR /app

RUN apk --update add --no-cache \ 
    lapack-dev \ 
    gcc \
    freetype-dev

# Install dependencies
RUN apk add --no-cache --virtual .build-deps \
    gfortran \
    musl-dev \
    g++

RUN pip3 install -r requirements.txt

RUN python3 setup.py install

RUN apk del .build-deps

ENTRYPOINT python3 testapp.py

My project requirements are:

numpy==1.13.3
Cython==0.28.2
nibabel==2.2.1
scipy==1.0.0

I build the docker file as: docker build -t myimg .

So, the docker file progresses but scipy fails to build with the following error:

Collecting numpy==1.13.3 (from -r requirements.txt (line 1))
  Downloading https://files.pythonhosted.org/packages/bf/2d/005e45738ab07a26e621c9c12dc97381f372e06678adf7dc3356a69b5960/numpy-1.13.3.zip (5.0MB)
Collecting Cython==0.28.2 (from -r requirements.txt (line 2))
  Downloading https://files.pythonhosted.org/packages/79/9d/dea8c5181cdb77d32e20a44dd5346b0e4bac23c4858f2f66ad64bbcf4de8/Cython-0.28.2.tar.gz (1.9MB)
Collecting nibabel==2.2.1 (from -r requirements.txt (line 3))
  Downloading https://files.pythonhosted.org/packages/d7/de/1d96fd0b118c9047bf35f02090db8ef8fd3927dfce635f09a6f7d5b572e6/nibabel-2.2.1.zip (4.2MB)
Collecting scipy==1.0.0 (from -r requirements.txt (line 4))
  Downloading https://files.pythonhosted.org/packages/d0/73/76fc6ea21818eed0de8dd38e1e9586725578864169a2b31acdeffb9131c8/scipy-1.0.0.tar.gz (15.2MB)
Collecting six>=1.3 (from nibabel==2.2.1->-r requirements.txt (line 3))
  Downloading https://files.pythonhosted.org/packages/67/4b/141a581104b1f6397bfa78ac9d43d8ad29a7ca43ea90a2d863fe3056e86a/six-1.11.0-py2.py3-none-any.whl
Building wheels for collected packages: numpy, Cython, nibabel, scipy
  Running setup.py bdist_wheel for numpy: started
  Running setup.py bdist_wheel for numpy: still running...
  Running setup.py bdist_wheel for numpy: still running...
  Running setup.py bdist_wheel for numpy: still running...
  Running setup.py bdist_wheel for numpy: finished with status 'done'
  Stored in directory: /root/.cache/pip/wheels/b6/10/65/189b772e73b4505109d5a1e6671b07e65797023718777295e0
  Running setup.py bdist_wheel for Cython: started
  Running setup.py bdist_wheel for Cython: still running...
  Running setup.py bdist_wheel for Cython: still running...
  Running setup.py bdist_wheel for Cython: still running...
  Running setup.py bdist_wheel for Cython: finished with status 'done'
  Stored in directory: /root/.cache/pip/wheels/6f/24/5d/def09ad0aed8ba26186f2a38070906f70ab4b2287bf64d4414
  Running setup.py bdist_wheel for nibabel: started
  Running setup.py bdist_wheel for nibabel: finished with status 'done'
  Stored in directory: /root/.cache/pip/wheels/46/50/8d/bcb0b8f7c030da5bac1752fbe9cc375cbf5725fa93ba79ad84
  Running setup.py bdist_wheel for scipy: started
  Running setup.py bdist_wheel for scipy: finished with status 'error'
  Complete output from command /usr/local/bin/python -u -c "import setuptools, tokenize;__file__='/tmp/pip-install-boosbyfg/scipy/setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" bdist_wheel -d /tmp/pip-wheel-cczhwdqj --python-tag cp36:
  Traceback (most recent call last):
    File "<string>", line 1, in <module>
    File "/tmp/pip-install-boosbyfg/scipy/setup.py", line 418, in <module>
      setup_package()
    File "/tmp/pip-install-boosbyfg/scipy/setup.py", line 398, in setup_package
      from numpy.distutils.core import setup
  ModuleNotFoundError: No module named 'numpy'

Not sure why it is having issue finding numpy as it was installed as part of the requirements?

score 3 · Accepted Answer · answered Jul 19 '18 at 05:15

3

Because to build scipy wheel you need to have numpy installed. However, only numpy wheel build was complete by the time pip attempts to build scipy wheel.

You will have to install the dependencies first. There are multiple ways to do this.

1) Use a shell script like the one below, copy it and RUN it instead of RUN pip install -r requirements.txt:

#!/bin/sh
while read module; do
  pip install $module
done < requirements.txt

2) Install scipy in a seperate RUN command.

3) apk add py-numpy@community as discussed in this answer.

answered Jul 19 '18 at 05:15

Jay

1,980
1
13
23

Thanks! That was it. – Luca Jul 19 '18 at 14:02
2 and 3 didn't work for me. Also, the explanation doesn't seem right as SciPy fails trying to install numpy and it doesn't matter if its install before SciPy it just doesn't work...? – CpILL Apr 25 '19 at 06:55
@CpILL, for 2, I should have specified the order. RUN pip install libraries including numpy and then RUN pip install SciPy. As for the explanation, SciPy doesn't fail trying to install numpy. Scipy installation fails because it doesn't find numpy installed, which at that point was not installed (only wheel built). Hope that clears things. – Jay May 03 '19 at 02:19

docker cannot build scipy

1 Answers1