Closed martindurant closed 5 years ago
Actually, the executable dot
is in the bin/ directory of the environment.
Depending on whether I try from a terminal directly, where dot
is definitely installed, getting
Warning: Could not load "/opt/conda/envs/dask-scipy/lib/graphviz/libgvplugin_pango.so.6" - file not found
Warning: Could not load "/opt/conda/envs/dask-scipy/lib/graphviz/libgvplugin_pango.so.6" - file not found
Warning: Could not load "/opt/conda/envs/dask-scipy/lib/graphviz/libgvplugin_pango.so.6" - file not found
Format: "png" not recognized. Use one of: bmp canon cmap cmapx cmapx_np dot eps fig gv ico imap imap_np ismap jpe jpeg jpg pdf pic plain plain-extpng pov ps ps2 svg svgz tif tiff tk vml vmlz xdot xdot1.2 xdot1.4
Have you installed the yum
requirements with your system package manager?
I don't understand - the whole point of the conda install is not to have to do such things, else I would just install graphviz with yum too.
So I'm open to having this discussion, but think we need to shift forum to the webpage repo. Otherwise it will be buried in a feedstock issue and hard to find/link to in the future. Could I ask you to open this issue?
The issue is specific to installing graphviz, so this seems like exactly the right place to discuss it, or in the graphviz-feedstock repo. As you know, graphviz has proven a particularly annoying thorn to dask visualisation, and would love to have an in-one conda solution.
To be clear, I'm saying that a discussion about how to eliminate dependencies on system package managers (e.g. yum
) is an org wide discussion and needs to be moved to the webpage repo for higher visibility and to actually coordinate a solution (as I agree one is needed). Certainly graphviz is one case, but it is not the only one.
I agree that this could benefit from broader discussion
This case should be fixed now as PR ( https://github.com/conda-forge/graphviz-feedstock/pull/23 ) switched to using conda-forge's X11 packages, which negates the need for yum
in the case of graphviz
.
Installing with the linked environment, dask's visualize method (run from a jupyter-lab notebook) produces