mirror of
https://github.com/quickwit-oss/tantivy.git
synced 2026-05-20 10:10:42 +00:00
7183ac6cbc
`MergeOptimizedInvertedIndexReader` was added in #32 in order to avoid making small reads to our underlying `FileHandle`. It did so by reading the entire content of the posting lists and positions at open time. As that PR says: > A likely downside to this approach is that now pg_search will be, indirectly, holding onto a lot of heap-allocated memory that was read from its block storage. Perhaps in the (near) future we can further optimize the new `MergeOptimizedInvertedIndexReader` such that it pages in blocks of a few megabytes at a time, on demand, rather than the whole file. This PR makes that change. But it additionally removes code that was later added in #47 to borrow individual entries rather than creating `OwnedBytes` for them. I believe that this code was added due to a misunderstanding: `OwnedBytes` is a total misnomer: the bytes are not "owned": they are immutably borrowed and reference counted. An `OwnedBytes` object can be created for any type which derefs to a slice of bytes, and can be cheaply cloned and sliced. So there is no need to actually borrow _or_ copy the buffer under the `OwnedBytes`. Removing the code that was doing so allows us to safely recreate our buffer without worrying about the lifetimes of buffers that we've handed out.