Open jaimergp opened 2 years ago
Even if we will remove pyqt
from the installation, a plugin might still bring it back if it directly depends on it, or a dependency does. We have setup a CI pipeline to analyse potential cases like this, and we are finding a few.
I'll be collecting them here, along with the relevant issues and PRs fixing it.
palmari
napari-time-reader
beetlesafari
, napari-accelerated-pixel-and-object-classification
, napari-assistant
, napari-pyclesperanto-assistant
, napari-segment-blobs-and-things-with-membranes
, napari-workflow-optimizer
, guanine-crystal-analysis
, napari-cupy-image-processing
, napari-simpleitk-image-processing
iacs-ipac-reader
napari-live-recording
platymatch
napari-allencell-segmenter
napari-nD-annotator
@jaimergp another thing to check for might be depending on napari[all]
since this also pulls in PyQt.
See for example PlatyMatch: https://github.com/juglab/PlatyMatch/pull/7
Thanks @psobolewskiPhD! I checked the conda-forge
recipe and it doesn't suffer from that issue, since there's no napari[all]
option in conda, just napari
, so we are covered in this case!
Oh good.
Alas that specific one does have a from PyQt5.QtCore import Qt
as well.
Hopefully the devs take a look at my PR.
Oh I had missed the PR content! Good catch! So far the experimental CI doesn't check the source content. I think we could talk with the napari-hub team to set some kind of static code analysis though.
Pending new version of
constructor
, which provides solver-less installs thanks to https://github.com/conda/constructor/pull/559