mamba remove package doesn't respect --force flag

[dhirschfeld]

Or has different (IMHO undesirable) behaviour from conda.

With conda the --force flag allows you to remove a package without uninstalling every package which depends on it - e.g.

> conda remove --force numpy

## Package Plan ##

  environment location: C:\Users\dhirschf\envs\dev

  removed specs:
    - numpy


The following packages will be REMOVED:

  numpy-1.18.4-py37hae9e721_0


Proceed ([y]/n)? y

Preparing transaction: done
Verifying transaction: done
Executing transaction: done

Whilst it will leave your environment in an inconsistent state, IMHO if you use --force it's up to you to deal with the consequences. This can be very useful in cases where you e.g. want to use a develop/editable install.

In any case, trying the same with mamba basically wants to uninstall the entire environment:

> mamba remove --force numpy

                  __    __    __    __
                 /  \  /  \  /  \  /  \
                /    \/    \/    \/    \
███████████████/  /██/  /██/  /██/  /████████████████████████
              /  / \   / \   / \   / \  \____
             /  /   \_/   \_/   \_/   \    o \__,
            / _/                       \_____/  `
            |/
        ███╗   ███╗ █████╗ ███╗   ███╗██████╗  █████╗
        ████╗ ████║██╔══██╗████╗ ████║██╔══██╗██╔══██╗
        ██╔████╔██║███████║██╔████╔██║██████╔╝███████║
        ██║╚██╔╝██║██╔══██║██║╚██╔╝██║██╔══██╗██╔══██║
        ██║ ╚═╝ ██║██║  ██║██║ ╚═╝ ██║██████╔╝██║  ██║
        ╚═╝     ╚═╝╚═╝  ╚═╝╚═╝     ╚═╝╚═════╝ ╚═╝  ╚═╝

        mamba (0.4.2) supported by @QuantStack

        GitHub:  https://github.com/QuantStack/mamba
        Twitter: https://twitter.com/QuantStack

█████████████████████████████████████████████████████████████

Removing specs: ['numpy']
Transaction

  Prefix: C:\Users\dhirschf\envs\dev

  Removing specs:

   - numpy


  Package               Version  Build                   Channel       Size
-----------------------------------------------------------------------------
  Remove:
-----------------------------------------------------------------------------

  algopy                  0.5.7  py_0                    installed
  altair                  4.1.0  py_1                    installed
  arrow-cpp              0.17.1  py37h1234567_10_cpu     installed
  arviz                   0.9.0  py_0                    installed
  astropy           4.0.1.post1  py37h8055547_0          installed
  autograd                  1.3  py_0                    installed
  bokeh                   2.1.1  py37hc8dfbb8_0          installed
  bottleneck              1.3.2  py37hbc2f12b_1          installed
  catboost               0.23.2  py37hc8dfbb8_0          installed
  cftime                  1.2.0  py37h44b1f71_1          installed
  chainer                 7.4.0  py_0                    installed
  cython-blis             0.4.1  py37h8055547_1          installed
  dask                   2.20.0  py_0                    installed
  dask-glm                0.2.0  py_1                    installed
  dask-ml                 1.5.0  py_0                    installed
  datashader             0.10.0  py_0                    installed
  datashape               0.5.4  py_1                    installed
  deap                    1.3.1  py37h9b59f54_1          installed
  fletcher                0.5.0  pyh9f0ad1d_1            installed
  h5py                   2.10.0  nompi_py37hde23a51_103  installed
  holoviews              1.13.3  pyh9f0ad1d_0            installed
  hvplot                  0.6.0  pyh9f0ad1d_0            installed
  ibis-framework         0.14.0  py37_0                  installed
  imagecodecs         2020.5.30  py37h92c78e3_2          installed
  imageio                 2.9.0  py_0                    installed
  ipympl                  0.5.6  pyh9f0ad1d_1            installed
  lightgbm                2.3.1  py37h6538335_0          installed
  matplotlib              3.2.2  1                       installed
  matplotlib-base         3.2.2  py37hab8957c_1          installed
  mkl_fft                1.0.15  py37h14836fe_0          installed
  mkl_random              1.1.0  py37h675688f_0          installed
  mlxtend                0.17.2  py_0                    installed
  netcdf4                 1.5.3  nompi_py37he83aa3c_105  installed
  numba                  0.50.1  py37h3bbf574_0          installed
  numcodecs               0.6.4  py37h6538335_0          installed
  numdifftools           0.9.39  py_0                    installed
  numexpr                 2.7.1  py37h25d0782_0          installed
  numpy                  1.18.4  py37hae9e721_0          installed
  opencv                  4.3.0  py37_2                  installed
  opt_einsum              3.2.1  py_0                    installed
  pandas                  1.0.5  py37h3bbf574_0          installed
  panel                   0.9.7  py_0                    installed
  parquet-cpp             1.5.1  2                       installed
  patsy                   0.5.1  py_0                    installed
  pomegranate            0.13.3  py37h1fb7aa8_0          installed
  py-opencv               4.3.0  py37h43977f1_2          installed
  py-xgboost               0.90  py37_1                  installed
  pyarrow                0.17.1  py37h1234567_10_cpu     installed
  pygpu                   0.7.6  py37h44b1f71_1001       installed
  pymc3                   3.9.2  py_0                    installed
  pytables                3.6.1  py37h2d87964_2          installed
  python-cufflinks       0.17.3  py_0                    installed
  python-hdfs             2.5.8  py_0                    installed
  pytorch-cpu             1.1.0  py37ha925a31_0          installed
  pywavelets              1.1.1  py37h44b1f71_1          installed
  quandl                  3.5.0  py37_0                  installed
  scikit-image           0.17.2  py37h3bbf574_1          installed
  scikit-learn           0.23.1  py37ha0ab79d_0          installed
  scikit-optimize         0.7.4  py_0                    installed
  scipy                   1.4.1  py37h9439919_0          installed
  seaborn                0.10.1  1                       installed
  seaborn-base           0.10.1  py_1                    installed
  skorch                  0.8.0  pyh5ca1d4c_0            installed
  spacy                   2.3.1  py37h1fb7aa8_0          installed
  statsmodels            0.11.1  py37h4ab8f01_2          installed
  theano                  1.0.4  py37h1834ac0_1002       installed
  thinc                   7.4.1  py37h1fb7aa8_0          installed
  tifffile             2020.7.4  py_0                    installed
  torchvision-cpu         0.3.0  py37_cuNone_1           installed
  tpot                   0.11.5  pyh9f0ad1d_0            installed
  turbodbc                4.1.1  py37h47af31f_0          installed
  umap-learn              0.4.6  py37hc8dfbb8_0          installed
  xarray                 0.16.0  py_0                    installed
  zarr                    2.4.0  py_0                    installed

  Summary:

  Remove: 74 packages

  Total download: 0  B

-----------------------------------------------------------------------------

Confirm changes: [Y/n] n

[JBlaschke]

My workaround at the moment is to install conda in the environment, and to use that for the --force uninstall:

mamba activate $MY_ENV
mamba install conda -c defaults --yes
conda remove --force $OFFENDING_PACKAGE

But ideally mamba will be able to do that soon :)

[wolfv]

I have been giving this some thought and it will require us to re-engineer slightly our current "Transaction" class to be able to handle "explicit" transactions as well.

Explicit transactions will not go through the solver, but rather either use URLs / files or the forced removal of an installed spec.

Having this would also be nice so that we can consolidate our current micromamba explicit environmetn file handling.

[gusostow]

+1 for this

[jonashaag]

Closing this as recent discussion is in #1523

[beckermr]

So the plan is to always have the mamba api differ from conda here? If so I think we need to have mamba error if people try the command.

[jonashaag]

I just meant to say we should continue discussing in the other ticket.

[jonashaag]

If I “close as duplicate” in GitHub it days “closed as not planned” which doesn’t make sense

[beckermr]

Ahhhhh got it. Great and thank you!