Add initial graph layout algorithm #398
Conversation
datashader/layout.py
Outdated
|
|
||
| # This comes from the sparse FR layout in NetworkX | ||
| A = nx.to_scipy_sparse_matrix(graph, dtype='f') | ||
| nnodes, _ = A.shape |
There was a problem hiding this comment.
Is this just a temporary implementation for testing? We can't depend on networkx in general.
There was a problem hiding this comment.
This was meant as temporary until I could switch over to using dataframes for input. The original implementation relied on NetworkX.
datashader/layout.py
Outdated
| nodes. | ||
| dim : int | ||
| Coordinate dimensions of each node. | ||
|
|
There was a problem hiding this comment.
If this is a ParameterizedFunction, it should have parameters, i.e. declare iterations = param.Integer(10,doc="Number of iterations") at the class level, with __call__ set up as other ParameterizedFunctions are (no explicit named arguments like that) and no docstring like this.
There was a problem hiding this comment.
Thanks. I updated the class.
datashader/layout.py
Outdated
| # this is the largest step allowed in the dynamics. | ||
| t = 0.1 | ||
|
|
||
| # simple cooling scheme. |
There was a problem hiding this comment.
Should the meat of this function be in a separate Numba-ized function?
We removed the NetworkX requirement and can now take two dataframes (nodes and edges) as input, like the bundling module.
|
@jbednar Thanks for the feedback. The initial algorithm is now fully parameterized, uses only dataframes for both input and output, and uses Numba as appropriate. I'm going to merge this so that I can work on a simple demo and then create a new dev release. |
|
Looks good, thanks! |
No description provided.