You can report an issue to the Jdaviz GitHub issues.
Some currently known but unresolved common issues that users encounter are as follow in their respective categories. This list is not exhaustive, so please also consult existing Jdaviz GitHub issues as well if you are unable to find your issue here:
known_issues_installation
known_issues_app
known_issues_cubeviz
known_issues_imviz
known_issues_specviz
This can be fixed by reinstalling scikit-image:
pip uninstall scikit-image
conda install scikit-image
The reason for this issue is that prebuilt binaries for scikit-image don't work on Mac versions of 10.13 or older and conda installs an older version of scikit-image that works with those versions. Another way to get the up-to-date scikit-image version is:
pip install -U --no-binary scikit-image scikit-image.
Although this solution takes much longer (~5 minutes) to install than the first solution.
The 0.6.4 version of vispy fails to build for some combinations of platform/OS and Python versions. vispy 0.6.5 has resolved this, but a workaround if you have an older version of vispy is to ensure you have a compatible version:
conda create -n jdaviz python=3.8
conda activate jdaviz
pip install vispy>=0.6.5
pip install jdaviz --no-cache-dir
See Issue #305 for updates on this topic.
In a conda environment, where numpy was installed using conda, installing jdaviz using pip will attempt to pull bottleneck from PyPI. This might result in bottleneck trying to build numpy from source and crash, stalling the installation altogether. When this happens, exit the installation, install bottleneck with conda, and try to install jdaviz again.
In some environments, you occasionally might not able to start Jdaviz in Jupyter Lab due to this error:
IndexError: pop from an empty deque
This is an upstream issue at jupyterlab/jupyterlab#11934 that is not related to Jdaviz. The workaround is to find another environment where you do not see this error or use Jupyter Notebook instead.
Running Cubeviz from the command line sometimes results in a failure to initialize the app in the browser due to a RuntimeError
in tornado/ioloop.py
. We are investigating, but in the meantime reinstalling fresh in a new conda environment may help. Alternatively, running Cubeviz in a Jupyter notebook instead of from the command line will circumvent the problem.
When running Jdaviz on a Linux virtual machine (VM), the spectrum may not appear in the spectrum viewer. This is a known bug in an underlying package. Until it is fixed, the workaround is to run the following in a Jupyter notebook cell before importing jdaviz
:
from glue_jupyter.bqplot.profile import layer_artist
layer_artist.USE_GL = False
When trying to do a second collapse with the same spectral region, but with resized bounds: change to Region=None, resize the region, then reselect Region 1, the region bounds are correct. However, applying Collapse again, it errors out and the image viewer that contained the initial collapse goes blank.
In order to see the full Cubeviz app in a Jupyter notebook, one can click on the side of the cell output to collapse or expand the scrollable window. This has the unintended consequence of changing the contrast of the image displayed in the Cubeviz cube viewer.
In some OS/browser combinations, imviz.add_markers(...)
might take a few tries to show the markers, or not at all. This is a known bug reported in glue-viz/glue-jupyter#243 . If you encounter this, try a different OS/browser combo.
Due to a known bug reported in glue-viz/glue-astronomy#52 , aperture photometry and radial profile will report inaccurate results when you calculate them on dithered images linked by WCS unless you are on the reference image (this is usually the first loaded image).
See the identically named issue in known_issues_cubeviz
.
Giving a redshift value will report a converted radial velocity, which if entered manually will not convert to the exact same redshift value. Note that the redshift value is always treated as the true value and used when plotting lines, etc.