4 ms·

Does anyone happen to know of any geospatial indexing solutions that can also index simultaneously by other dimensions. Temporal indexing for example (not only

• geophile 5y ago
  Z-order does this easily, just index on (x, y, t). I wouldn't recommend it for more than maybe 4-6 dimensions though.
• rurabe 5y ago
  H3 cell indexes are just integers so you can easily make a compound index of (cell_index,timestamp). This would be easy in SQL and I'd have to imagine just about anything else that supports a compound index.
• jandrewrogers 5y ago
  Yes, with caveats and reasons you don't often see it. As a practical matter, you want to fit the indexing structure to the properties of your data model as closely as possible. Increasing the generality of high-dimensionality spatial indexes comes at a high cognitive cost, so most complex high-dimensionality indexes are bespoke designs to limit generality. Things become pretty messy when you mix dimension types that are interval-like (e.g. polygons) and monotonic-like (e.g. temporal) so almost no one does it. You can build, say, a general 8-dimensional index that can handle (some) distributions of interval and monotonic types simultaneously, in addition to the usual boring data types, that has excellent performance and scalability characteristics. It would probably only require something like 1000 lines of C++, so not too onerous. However, the code logic would be nearly impenetrable to read, never mind write, which matters for practical engineering.
  • eerikkivistik 5y ago
    Thank you for the thorough answer.