geforkt von Mirrors/Paper
1378ec9381
The metadata system generates unique keys for metadata entries based on the subject metadata is being applied to and the name of the metadata being applied. It was assumed this would be an expensive operation so a cache was put in place to ensure this was done as little as possible. In reality this cache only has a benefit when you have a hit rate above ~90% and is otherwise much slower. As the implementation of the cache is a hashmap of hashmaps it also uses a significant amount of memory which is not worth it even for the performance increase with a high hit rate. This commit simply removes the cache which results in speedups for most cases and large memory savings. By: crast <contact@jamescrasta.com> |
||
---|---|---|
paper-api |