PyPlot error on Julia 1.0.1

Tooling
1

Hi all,

I am having some trouble getting PyPlot to work with Julia 1.0.1. Starting with a fresh install, running from command line, here’s what I got (sorry it’s a little long):

   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.0.1 (2018-09-29)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org/ release
|__/                   |

┌ Warning: Terminal not fully functional
└ @ Base client.jl:342
julia> Pkg.status()
ERROR: UndefVarError: Pkg not defined
Stacktrace:
 [1] top-level scope at none:0

julia> using Pkg

julia> Pkg.status()
    Status `~/.julia/environments/v1.0/Project.toml`

julia> Pkg.add("PyPlot")
   Cloning default registries into /Users/kkylin/.julia/registries
   Cloning registry General from "https://github.com/JuliaRegistries/General.git"
  Updating registry at `~/.julia/registries/General`
  Updating git-repo `https://github.com/JuliaRegistries/General.git`
 Resolving package versions...
 Installed Reexport ────────── v0.2.0
 Installed VersionParsing ──── v1.1.3
 Installed FixedPointNumbers ─ v0.5.3
 Installed JSON ────────────── v0.19.0
 Installed LaTeXStrings ────── v1.0.3
 Installed PyPlot ──────────── v2.6.3
 Installed PyCall ──────────── v1.18.5
 Installed Conda ───────────── v1.1.1
 Installed MacroTools ──────── v0.4.4
 Installed Compat ──────────── v1.3.0
 Installed Colors ──────────── v0.9.4
 Installed ColorTypes ──────── v0.7.5
  Updating `~/.julia/environments/v1.0/Project.toml`
  [d330b81b] + PyPlot v2.6.3
  Updating `~/.julia/environments/v1.0/Manifest.toml`
  [3da002f7] + ColorTypes v0.7.5
  [5ae59095] + Colors v0.9.4
  [34da2185] + Compat v1.3.0
  [8f4d0f93] + Conda v1.1.1
  [53c48c17] + FixedPointNumbers v0.5.3
  [682c06a0] + JSON v0.19.0
  [b964fa9f] + LaTeXStrings v1.0.3
  [1914dd2f] + MacroTools v0.4.4
  [438e738f] + PyCall v1.18.5
  [d330b81b] + PyPlot v2.6.3
  [189a3867] + Reexport v0.2.0
  [81def892] + VersionParsing v1.1.3
  [2a0f44e3] + Base64 
  [ade2ca70] + Dates 
  [8bb1440f] + DelimitedFiles 
  [8ba89e20] + Distributed 
  [b77e0a4c] + InteractiveUtils 
  [76f85450] + LibGit2 
  [8f399da3] + Libdl 
  [37e2e46d] + LinearAlgebra 
  [56ddb016] + Logging 
  [d6f4376e] + Markdown 
  [a63ad114] + Mmap 
  [44cfe95a] + Pkg 
  [de0858da] + Printf 
  [3fa0cd96] + REPL 
  [9a3f8284] + Random 
  [ea8e919c] + SHA 
  [9e88b42a] + Serialization 
  [1a1011a3] + SharedArrays 
  [6462fe0b] + Sockets 
  [2f01184e] + SparseArrays 
  [10745b16] + Statistics 
  [8dfed614] + Test 
  [cf7118a7] + UUIDs 
  [4ec0a83e] + Unicode 
  Building Conda ─→ `~/.julia/packages/Conda/uQitS/deps/build.log`
  Building PyCall → `~/.julia/packages/PyCall/0jMpb/deps/build.log`

julia> using PyPlot
[ Info: Precompiling PyPlot [d330b81b-6aea-500a-939a-2ce795aea3ee]
[ Info: Installing matplotlib via the Conda matplotlib package...
[ Info: Running `conda install -y matplotlib` in root environment
Solving environment: done

## Package Plan ##

  environment location: /Users/kkylin/.julia/conda/3

  added / updated specs: 
    - matplotlib


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    python-dateutil-2.7.3      |           py37_0         260 KB
    pytz-2018.5                |           py37_0         231 KB
    matplotlib-3.0.0           |   py37h54f8f79_0         6.8 MB
    pyparsing-2.2.2            |           py37_0          97 KB
    kiwisolver-1.0.1           |   py37h0a44026_0          56 KB
    freetype-2.9.1             |       hb4e5f40_0         864 KB
    libpng-1.6.35              |       ha441bb4_0         318 KB
    cycler-0.10.0              |           py37_0          14 KB
    tornado-5.1.1              |   py37h1de35cc_0         661 KB
    ------------------------------------------------------------
                                           Total:         9.3 MB

The following NEW packages will be INSTALLED:

    cycler:          0.10.0-py37_0       
    freetype:        2.9.1-hb4e5f40_0    
    kiwisolver:      1.0.1-py37h0a44026_0
    libpng:          1.6.35-ha441bb4_0   
    matplotlib:      3.0.0-py37h54f8f79_0
    pyparsing:       2.2.2-py37_0        
    python-dateutil: 2.7.3-py37_0        
    pytz:            2018.5-py37_0       
    tornado:         5.1.1-py37h1de35cc_0


Downloading and Extracting Packages
python-dateutil-2.7. | 260 KB    | ################################################################################################################ | 100% 
pytz-2018.5          | 231 KB    | ################################################################################################################ | 100% 
matplotlib-3.0.0     | 6.8 MB    | ################################################################################################################ | 100% 
pyparsing-2.2.2      | 97 KB     | ################################################################################################################ | 100% 
kiwisolver-1.0.1     | 56 KB     | ################################################################################################################ | 100% 
freetype-2.9.1       | 864 KB    | ################################################################################################################ | 100% 
libpng-1.6.35        | 318 KB    | ################################################################################################################ | 100% 
cycler-0.10.0        | 14 KB     | ################################################################################################################ | 100% 
tornado-5.1.1        | 661 KB    | ################################################################################################################ | 100% 
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
ERROR: InitError: PyError (PyImport_ImportModule) <class 'RuntimeError'>
RuntimeError("The current Numpy installation ('/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py') fails to pass simple sanity checks. This can be caused for example by incorrect BLAS library being linked in.")
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/matplotlib/__init__.py", line 141, in <module>
    from . import cbook, rcsetup
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/matplotlib/cbook/__init__.py", line 33, in <module>
    import numpy as np
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 227, in <module>
    _sanity_check()
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 225, in _sanity_check
    raise RuntimeError(msg.format(__file__))

Stacktrace:
 [1] pyimport(::String) at /Users/kkylin/.julia/packages/PyCall/0jMpb/src/PyCall.jl:486
 [2] macro expansion at ./logging.jl:309 [inlined]
 [3] pyimport_conda(::String, ::String, ::String) at /Users/kkylin/.julia/packages/PyCall/0jMpb/src/PyCall.jl:653
 [4] pyimport_conda at /Users/kkylin/.julia/packages/PyCall/0jMpb/src/PyCall.jl:649 [inlined]
 [5] __init__() at /Users/kkylin/.julia/packages/PyPlot/fZuOQ/src/init.jl:174
 [6] _include_from_serialized(::String, ::Array{Any,1}) at ./loading.jl:630
 [7] _require_from_serialized(::String) at ./loading.jl:681
 [8] _require(::Base.PkgId) at ./logging.jl:317
 [9] require(::Base.PkgId) at ./loading.jl:855
 [10] macro expansion at ./logging.jl:311 [inlined]
 [11] require(::Module, ::Symbol) at ./loading.jl:837
during initialization of module PyPlot

As the error suggests, this seems to be a problem involving libopenblas being linked somewhere. But I couldn’t quite figure out a fix. Any pointers would be greatly appreciated!

KL

2

It’s useful to check if it works with Python alone first:

using Pkg
Pkg.add("PyCall")
using PyCall
# Check if it's importable by Python
run(`$(PyCall.pyprogramname) -c "import numpy"`)
run(`$(PyCall.pyprogramname) -c "import matplotlib"`)
# Check if it's importable by PyCall
pyimport("numpy")
pyimport("matplotlib")
# Finally try PyPlot
using PyPlot

Note: the error is coming from

https://github.com/numpy/numpy/blob/8aa121415760cc6839a546c3f84e238d1dfa1aa6/numpy/__init__.py#L193-L214

which mentions

https://github.com/numpy/numpy/issues/8577

3

Thanks for the suggestion. Here’s what I got:

julia> using Pkg

julia> using PyCall

julia> run(`$(PyCall.pyprogramname) -c "import numpy"`)
Process(`/Users/kkylin/.julia/conda/3/bin/python -c 'import numpy'`, ProcessExited(0))

julia> run(`$(PyCall.pyprogramname) -c "import matplotlib"`)
Process(`/Users/kkylin/.julia/conda/3/bin/python -c 'import matplotlib'`, ProcessExited(0))

julia> pyimport("numpy")
ERROR: PyError (PyImport_ImportModule) <class 'RuntimeError'>
RuntimeError("The current Numpy installation ('/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py') fails to pass simple sanity checks. This can be caused for example by incorrect BLAS library being linked in.")
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 227, in <module>
    _sanity_check()
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 225, in _sanity_check
    raise RuntimeError(msg.format(__file__))

Stacktrace:
 [1] pyimport(::String) at /Users/kkylin/.julia/packages/PyCall/0jMpb/src/PyCall.jl:486
 [2] top-level scope at none:0

julia> pyimport("matplotlib")
ERROR: PyError (PyImport_ImportModule) <class 'RuntimeError'>
RuntimeError("The current Numpy installation ('/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py') fails to pass simple sanity checks. This can be caused for example by incorrect BLAS library being linked in.")
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/matplotlib/__init__.py", line 126, in <module>
    from . import cbook
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/matplotlib/cbook/__init__.py", line 34, in <module>
    import numpy as np
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 227, in <module>
    _sanity_check()
  File "/Users/kkylin/.julia/conda/3/lib/python3.7/site-packages/numpy/__init__.py", line 225, in _sanity_check
    raise RuntimeError(msg.format(__file__))

Stacktrace:
 [1] pyimport(::String) at /Users/kkylin/.julia/packages/PyCall/0jMpb/src/PyCall.jl:486
 [2] top-level scope at none:0

So indeed, Julia cannot import numpy. However, the numpy installation seems to be fine when called from within python:

julia> run(`/Users/kkylin/.julia/conda/3/bin/python`)
Python 3.7.0 (default, Jun 28 2018, 07:39:16) 
[Clang 4.0.1 (tags/RELEASE_401/final)] :: Anaconda, Inc. on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy as np
>>> a = np.arange(15).reshape(3, 5)
>>> a
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14]])
>>> a.shape
(3, 5)
>>> a.ndim
2
>>> 
Process(`/Users/kkylin/.julia/conda/3/bin/python`, ProcessExited(0))

julia>

(I do have other pythons on my system in various places, but this looks like the one Julia is using.)

I’ll keep looking.

Thanks,
KL

4

Alright, looks like the culprit is having multiple python distributions. In particular, I have Anaconda installed (separately, manually) elsewhere in my home directory, and that seemed to interfere with the version in .julia/conda. When I set PyCall to use Anaconda, the problems went away.

For anyone else having similar issues, try:

ENV["PYTHON"] = "/Users/foo/anaconda/bin/python"  ## put the path to your python program here
Pkg.build("PyCall")

And that may (hopefully!) take care of it.

KL

5

Actually, the test for matplotlib should be something like

run(`/Users/kkylin/.julia/conda/3/bin/python -c "from matplotlib.pyplot import plot, show; plot([1,2,3]); show()"`)

since just importing matplotlib won’t activate the GUI. Or maybe use Python REPL as you did.

julia> run(`/Users/kkylin/.julia/conda/3/bin/python`)
Python 3.7.0 ...
>>> from matplotlib.pyplot import plot, show
>>> plot([1,2,3])
>>> show()

Yes, that’s correct. In general, whatever the value of PyCall.pyprogramname is the Python used by Julia.

If it works with ~/anaconda/bin/python I guess there is a combination of numpy, matplotlib (+ other Python packages) and PyCall that does not work on macOS. Judging from the value of your PyCall.pyprogramname, I guess you installed/updated Conda.jl after installing/updating ~/anaconda? If that’s the case, you may want to be careful updating ~/anaconda until the bug is solved.

6

If it works with ~/anaconda/bin/python I guess
there is a combination of numpy, matplotlib (+ other
Python packages) and PyCall that does not work on
macOS. Judging from the value of your
PyCall.pyprogramname, I guess you
installed/updated Conda.jl after installing/updating
~/anaconda? If that’s the case, you may want to
be careful updating ~/anaconda until the bug is
solved.

Yes, I think you are right. My installation with
regard to python is a bit of a mess. I do need to
clean it up.

Thanks for the help!

KL