haesleinhuepf/devbio-napari

missing napari-czifle2 in conda recipe of devbio-napari

Opened this issue · 10 comments

This
conda create --name devbio-napari-env python=3.9 devbio-napari -c conda-forge conda activate devbio-napari-env
failed to install napari-czifile2
A manual install with

conda install -c conda-forge napari-czifile2 helped.

Hi Antonio @manerotoni,

the conda package should update over night with the dependency included. Thanks again for reaching out! Let me know if there are any further issues!

Best,
Robert

Hi Robert,
I think the mamba recipe also misses the napari-czifile2 (I thought it is the same as conda?)

Hey Antonio @manerotoni ,

it's the same as conda, yes, and it lives here:
https://github.com/conda-forge/devbio-napari-feedstock/blob/main/recipe/meta.yaml#L55

If you still experience issues, would you mind sharing details? E.g. what did you install and how? What was the output of the program and how do you know installation doesn't work? The output of conda list in your environment would be interesting too.

Let me know!

Best,
Robert

The two install are definitely very different. I have a Windows10 miniconda install. Just performed the test today 26.07.2022, 9.25 AM

conda create

conda create --name devbio-napari-env-conda python=3.9 devbio-napari -c conda-forge

Package list is conda env export > dev-napari-env-conda.log 384 packages

devbio-napari-env-conda.log

mamba create

mamba create --name devbio-napari-env-mamba python=3.9 devbio-napari -c conda-forge

Package list is conda env export > dev-napari-env-conda.log 360 packages
devbio-napari-env-mamba.log

The package installed and also the versions installed are quite different!
This is just a short list
image

Very confusing.

mamba (0.24.0) and conda (4.13.0) do not give me any error. I am able to install missing packages using
mamba install ... -c conda-forge

In the report of what is being installed packages are indeed missing:

Here part of the mamba install log

image

and here the corresponding part with conda

image

The package installed and also the versions installed are quite different!

Indeed! Conda installs devbio-napari 0.5.9 and mamba installs devbio-napari 0.5.6 .

What happens if you explicitly install a specific devbio-napari version? Do conda and mamba do the same then?

Furthermore, did you install mamba in the base environment? And did you call both installations from the base?

conda and mamba are in the base environment.
But why do they use a different distribution?
I would think that they should both use the latest.
I will try to install 0.5.9, then I expect to get the same result.

From base environment (mamba installed with conda install mamba)

conda create --name devbio-napari-env-conda-0.5.9 python=3.9 devbio-napari=0.5.9 -c conda-forge
devbio-napari-env-conda-0.5.9.log
384 packages installed

mamba create --name devbio-napari-env-mamba-0.5.9 python=3.9 devbio-napari=0.5.9 -c conda-forge
devbio-napari-env-mamba-0.5.9.log
390 packages installed

We still have a delta of 6 packages, but now in the inverse direction, also some packages have a different versions.

I think these packages are unrelated to devbio-napari.

Hi @manerotoni ,

I suggest to talk to the conda and/or mamba developers about this. Furthermore, we are planning to release devbio-napari in the future with pinned dependencies. It should then become more robust against variations.

Best,
Robert

Will you communicate with conda/mamba developers or should I do it?
It is a little scary that 2 tools that are meant to do the same install do something quite different. Mamba resolve dependencies differently and this explains why you are faster but may also get you two different results.