[FEATURE] Faster k-way merge of temp features
Closed this issue · 0 comments
msbarry commented
Currently we use a priority queue to do the k-way merge of temporary features at mbtiles write time, which is a single-threaded bottleneck during mbtiles writes:
It's possible to improve this by using parallel readers (#163) but there are also faster data structures available for the k-way merge. For example, a modified version of GraphHopper's MinHeapWithUpdate class.