In the last PR, I forgot to include compression of the edge indices, and this PR adds that.
One of them has a size as large as the total number of edges, so this could decent (10-20%) additional compression.
As far as I can tell, using .save() in the builder compresses all of the datasets in node and edge h5 files.
In the last PR, I forgot to include compression of the edge indices, and this PR adds that. One of them has a size as large as the total number of edges, so this could decent (10-20%) additional compression. As far as I can tell, using .save() in the builder compresses all of the datasets in node and edge h5 files.