3 ms·
Pretty cool. Did you bake the viz yourself? How does it work.. Can you explain a bit how you handle the sheer scale of it. And how the nodes are positioned - I
by bluishgreen 6y ago
Pretty cool. Did you bake the viz yourself? How does it work.. Can you explain a bit how you handle the sheer scale of it. And how the nodes are positioned - I can't imagine it is some sort of graph node attraction/repulsion algorithm as it will kill the server? Do you precompute?
- 7373737373 6y agoI wrote down some things here: https://github.com/void4/notes/issues/48 https://github.com/void4/notes/issues/48 The code is here: https://github.com/void4/wikiscape https://github.com/void4/wikiscape tl;dr: wrote scripts to parse wikipedia dump, used LargeVis (https://github.com/lferry007/LargeVis https://github.com/lferry007/LargeVis) to layout the graph, wrote further scripts to generate tiles, tiles in layers up to a specific zoom level are saved as files and served directly by nginx, closer ones are generated dynamically by flask, which also handles querying for the closest node to the mouse position efficiently with https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.cKDTree.html https://docs.scipy.org/doc/scipy/reference/generated/scipy.s...