Conversation
bsipocz
left a comment
There was a problem hiding this comment.
Looks good, though you may want to experiment a little bit with the base package and see how badly the installs are bloated
| - bokeh | ||
| - ipykernel | ||
| - ipywidgets | ||
| - matplotlib |
There was a problem hiding this comment.
yes, this sounds reasonable. OTOH, I'm not sure if the users of this tool would need mpl for their work that would be beyond running the notebook as is.
A compromise could be to use the base package (called matplotlib-base) that doesn't install optional dependencies.
There was a problem hiding this comment.
I think users can install that if they want it. I'd prefer to keep this file limited to what's needed by the notebook.
There was a problem hiding this comment.
sounds good for now.
The downside of leaving basic installs to the users is the bloated overall size -- e.g. if most of them install mpl then it would be more optimal to provide it centrally than having it take up space in most of the users directories.
The environment created by
conda-sphere_sdt.ymlfor the SPHEREx source discovery tool is about 1.2G. That's pretty large, which makes it more cumbersome to work with. I looked into whether we could reduce the size and found that matplotlib is being installed but not used. Removing it reduces the size of the environment by about a third, down to 770M.I tested this on fornax and see no difference in the source discovery tool notebook when running in an environment built by the current
conda-sphere_sdt.ymlvs the one in this PR that removes matplotlib. I'm checking with @afaisst to make sure I'm not missing something.