Installing add-ons on Windows updates numpy (and makes Orange useless)

[echavezangel]

I am getting a the following problem with data set files ("tab" files) after installing some adds
Read error:
'''Cannot parse dataset C:/Program Files/Orange/Lib/site-packages/Orange/datasets/housing.tab: Cannot cast ufunc 'contains (vectorized)' output from dtype('O') to dtype('bool') with casting rule 'same_kind' '''

I don't know exactly which is the add causing this problem.
This appears after the installation of some adds (network, time series or any except spectroscopy).
I updated the orange version 3.30.2 -> 3.31.0 but the problem didn't disappear. The only way to read again "tab" files is uninstalling orange and installing again.

• Operating system: Win 10 64 bits
• Orange version: 3.30.2 and 3.31.0
• How you installed Orange: *.exe

[markotoplak]

On Dec 31 a new version of numpy was released which seemed to get updated during your add-on installation process. The new version of numpy introduced some changes which are incompatible with the current Orange code. We plan to update Orange to support newer numpy versions shortly (the fix was already committed, see #5760).

To confirm that this is the problem could you please open the Orange command prompt and paste the outputs of conda list and conda --version here?

You can solve your problem by installing the previous numpy version manually. Open the Orange command prompt and run conda install numpy=1.21. Please report if that solved your problem. Do not forget to run Orange command prompt as admin.

Does anyone have any idea why would numpy get upgraded when installing add-ons? All but spectroscopy are installed with conda, so It is likely something conda install related. If numpy gets upgraded for at every add-on install (I suspect none has 1.22 as an explicit dependency) we have a problem...

[echavezangel]

Thank you for the answer the installation of previous numpy worked well.

here the full Conda list:

_py-xgboost-mutex 2.0 cpu_0
anyio 3.4.0 py38haa244fe_0
anyqt 0.0.13 pyh6c4a22f_0
asteval 0.9.25 pypi_0 pypi
backcall 0.2.0 pyh9f0ad1d_0
backports 1.0 py_2
backports.functools_lru_cache 1.6.4 pyhd8ed1ab_0
baycomp 1.0.2 py_1
blas 2.112 openblas
blas-devel 3.9.0 12_win64_openblas
bottleneck 1.3.2 py38h6f4d8f0_5
brotli 1.0.9 h8ffe710_6
brotli-bin 1.0.9 h8ffe710_6
brotlipy 0.7.0 py38h294d835_1003
ca-certificates 2021.10.8 h5b45459_0
cachecontrol 0.12.10 pyhd8ed1ab_0
catboost 1.0.3 py38haa244fe_1
certifi 2021.10.8 py38haa244fe_1
cffi 1.15.0 py38hd8c33c5_0
chardet 4.0.0 py38haa244fe_2
charset-normalizer 2.0.8 pyhd8ed1ab_0
colorama 0.4.4 pyh9f0ad1d_0
colorcet 3.0.0 pypi_0 pypi
commonmark 0.9.1 py_0
cryptography 36.0.0 py38hb7941b4_0
cycler 0.11.0 pyhd8ed1ab_0
debugpy 1.5.1 py38h885f38d_0
decorator 5.1.0 pyhd8ed1ab_0
dictdiffer 0.9.0 pyhd8ed1ab_0
docutils 0.18.1 py38haa244fe_0
entrypoints 0.3 pyhd8ed1ab_1003
et_xmlfile 1.0.1 py_1001
extranormal3 0.0.3 pypi_0 pypi
fonttools 4.28.2 py38h294d835_0
freetype 2.10.4 h546665d_1
future 0.18.2 py38haa244fe_4
h11 0.12.0 pyhd8ed1ab_0
h2 4.1.0 py38haa244fe_0
h5py 3.6.0 pypi_0 pypi
hpack 4.0.0 pyh9f0ad1d_0
httpcore 0.13.7 pyhd8ed1ab_0
httpx 0.19.0 py38haa244fe_0
hyperframe 6.0.1 pyhd8ed1ab_0
icu 68.2 h0e60522_0
idna 3.1 pyhd3deb0d_0
importlib-metadata 4.8.2 py38haa244fe_0
importlib_metadata 4.8.2 hd8ed1ab_0
ipykernel 6.5.1 py38h595d716_0
ipython 7.30.1 py38haa244fe_0
ipython_genutils 0.2.0 py_1
jbig 2.1 h8d14728_2003
jedi 0.18.1 py38haa244fe_0
joblib 1.1.0 pyhd8ed1ab_0
jpeg 9d h8ffe710_0
jupyter_client 7.1.0 pyhd8ed1ab_0
jupyter_core 4.9.1 py38haa244fe_1
keyring 23.4.0 py38haa244fe_0
keyrings.alt 4.0.2 pyhd3deb0d_0
kiwisolver 1.3.2 py38hbd9d945_1
lcms2 2.12 h2a16943_0
lerc 3.0 h0e60522_0
libblas 3.9.0 12_win64_openblas
libbrotlicommon 1.0.9 h8ffe710_6
libbrotlidec 1.0.9 h8ffe710_6
libbrotlienc 1.0.9 h8ffe710_6
libcblas 3.9.0 12_win64_openblas
libclang 11.1.0 default_h5c34c98_1
libdeflate 1.8 h8ffe710_0
libflang 5.0.0 h6538335_20180525
liblapack 3.9.0 12_win64_openblas
liblapacke 3.9.0 12_win64_openblas
libopenblas 0.3.18 pthreads_hc469a61_0
libpng 1.6.37 h1d00b33_2
libsodium 1.0.18 h8d14728_1
libtiff 4.3.0 hd413186_2
libxgboost 1.5.0 h0e60522_1
libzlib 1.2.11 h8ffe710_1013
llvm-meta 5.0.0 0
lmfit 1.0.3 pypi_0 pypi
lockfile 0.12.2 py_1
lz4-c 1.9.3 h8ffe710_1
m2w64-gcc-libgfortran 5.3.0 6
m2w64-gcc-libs 5.3.0 7
m2w64-gcc-libs-core 5.3.0 7
m2w64-gmp 6.1.0 2
m2w64-libwinpthread-git 5.0.0.4634.697f757 2
matplotlib-base 3.5.0 py38h1f000d6_0
matplotlib-inline 0.1.3 pyhd8ed1ab_0
msgpack-python 1.0.3 py38hbd9d945_0
msys2-conda-epoch 20160418 1
munkres 1.1.4 pyh9f0ad1d_0
nest-asyncio 1.5.4 pyhd8ed1ab_0
networkx 2.6.3 pyhd8ed1ab_1
numpy 1.19.5 py38h09042cb_2 (This version was before the installation of the Add)
olefile 0.46 pyh9f0ad1d_1
openblas 0.3.18 pthreads_h543f93c_0
openjpeg 2.4.0 hb211442_1
openmp 5.0.0 vc14_1 [vc14]
openpyxl 3.0.9 pyhd8ed1ab_0
openssl 1.1.1l h8ffe710_0
opentsne 0.6.1 py38he5193b3_1
orange-canvas-core 0.1.23 pyhd8ed1ab_0
orange-spectroscopy 0.6.3 pypi_0 pypi
orange-widget-base 4.15.1 pyhd8ed1ab_0
orange3 3.30.2 py38h5d928e2_0
packaging 21.3 pyhd8ed1ab_0
pandas 1.3.4 py38h5d928e2_1
param 1.12.0 pypi_0 pypi
parso 0.8.3 pyhd8ed1ab_0
pebble 4.6.3 pypi_0 pypi
pickleshare 0.7.5 py_1003
pillow 8.4.0 py38h794f750_0
pip 21.3.1 pyhd8ed1ab_0
prompt-toolkit 3.0.22 pyha770c72_0
py-xgboost 1.5.0 py38haa244fe_1
pycparser 2.21 pyhd8ed1ab_0
pyct 0.4.8 pypi_0 pypi
pygments 2.10.0 pyhd8ed1ab_0
pyopenssl 21.0.0 pyhd8ed1ab_0
pyparsing 3.0.6 pyhd8ed1ab_0
pyqt 5.12.3 py38haa244fe_8
pyqt-impl 5.12.3 py38h885f38d_8
pyqt5-sip 4.19.18 py38h885f38d_8
pyqtchart 5.12 py38h885f38d_8
pyqtgraph 0.12.2 pyhd8ed1ab_0
pyqtwebengine 5.12.1 py38h885f38d_8
pysocks 1.7.1 py38haa244fe_4
python 3.8.8 h7840368_0_cpython
python-dateutil 2.8.2 pyhd8ed1ab_0
python-louvain 0.15 pyhd3deb0d_0
python_abi 3.8 2_cp38
pytz 2021.3 pyhd8ed1ab_0
pywin32 302 py38h294d835_2
pywin32-ctypes 0.2.0 py38haa244fe_1004
pyyaml 6.0 py38h294d835_3
pyzmq 22.3.0 py38h09162b1_1
qasync 0.22.0 pyhd8ed1ab_0
qt 5.12.9 h5909a2a_4
qtconsole 5.2.1 pyhd8ed1ab_0
qtpy 1.11.2 pyhd8ed1ab_0
renishawwire 0.1.15 pypi_0 pypi
requests 2.26.0 pyhd8ed1ab_1
rfc3986 1.5.0 pyhd8ed1ab_0
scikit-learn 0.23.2 py38ha09990b_3
scipy 1.5.3 py38h5f893b4_0
serverfiles 0.3.0 py_0
setuptools 59.4.0 py38haa244fe_0
six 1.16.0 pyh6c4a22f_0
sniffio 1.2.0 py38haa244fe_2
spectral 0.22.4 pypi_0 pypi
sqlite 3.37.0 h8ffe710_0
threadpoolctl 3.0.0 pyh8a188c0_0
tk 8.6.11 h8ffe710_1
tornado 6.1 py38h294d835_2
traitlets 5.1.1 pyhd8ed1ab_0
typing_extensions 4.0.1 pyha770c72_0
ucrt 10.0.20348.0 h57928b3_0
uncertainties 3.1.6 pypi_0 pypi
unicodedata2 13.0.0.post2 py38h294d835_4
urllib3 1.26.7 pyhd8ed1ab_0
vc 14.2 hb210afc_5
vs2015_runtime 14.29.30037 h902a5da_5
wcwidth 0.2.5 pyh9f0ad1d_2
wheel 0.37.0 pyhd8ed1ab_1
win_inet_pton 1.1.0 py38haa244fe_3
xgboost 1.5.0 py38haa244fe_1
xlrd 2.0.1 pyhd8ed1ab_3
xlsxwriter 3.0.2 pyhd8ed1ab_0
xz 5.2.5 h62dcd97_1
yaml 0.2.5 he774522_0
zeromq 4.3.4 h0e60522_1
zipp 3.6.0 pyhd8ed1ab_0
zlib 1.2.11 h8ffe710_1013
zstd 1.5.0 h6255e5f_0

Conda --ver : conda 4.9.2

[markotoplak]

@echavezangel, thanks! Reopening because it needs to be seen why (or if) installing add-ons on Windows upgrades numpy. If it does, we also need new Windows packages.

[markotoplak]

I tried 3 add-ons on a test Windows computer. I confirm, Text really forced the numpy upgrade. Interesting, it seems that one of its dependencies requires numpy newer that 1.19, at least on conda-forge.

To somewhat fix this issue or at least make Orange users suffer from it less often:

• We need to make new packages (almost done, waiting for MacOS signatures).
• The new release of Orange-Text should require Orange 3.31.1.

[PrimozGodec]

Now the new release of Orange3-text is made. Updating it to the newest version 1.6.2 should also fix Orange since it updates it.