ClickHouse Release 24.6

In this blog post, we introduced space-filling curves and their application to optimizing queries in a database. We discussed how Hilbert curves can be used to encode multiple columns as a single value while preserving locality, which can improve the performance of range queries and nearest neighbor searches. We also provided an example using NOAA Global Historical Climatology Network data to demonstrate the effectiveness of this technique. We showed that by ordering the table based on Hilbert encoding of latitude and longitude values, we could significantly reduce the number of rows read during a query, resulting in faster performance. This improvement is due to the preservation of spatial locality provided by the Hilbert curve, which ensures that points close in high-dimensional space are part of the same granules. In summary, using space-filling curves like Hilbert encoding can be an effective way to optimize queries for certain types of data, particularly when columns have similar properties and a high number of distinct values. However, it is essential to consider factors such as correlation between columns and the specific requirements of your query before deciding whether this technique is suitable for your use case.