Paul Masurel
Paul Masurel
Upon commit, how many fsync are happening? We sync explicitly in order to detect errors... But I think rust also sync_all upon drop. See #1221 in which the performance are...
It is probably possible to remove the `clippy::uninit_vec` thing in the LZ4 compression using MaybeUninit. Right now clippy complains because we end up creating a slice containing uninitialized data.
Depends on #11 The way `tantivy` works right now is it is a very naive algorithm. There is a lot of room to improve performance there. For instance, the must...
Avoid creating 8 segments if commit after 8 documents just because we use 8 threads. Maybe pack documents in blocks of 100 docs?
**Is your feature request related to a problem? Please describe.** People may want to index a field more than once. Typically, someone may want to index a field with and...
As reported by @ppodolsky . Right now block wand does not kick in when using boosting.
If the documents happen to be sorted or inversely sorted, we could improve the serialization speed. This can happen for some specific case like sorting by timestamp.
Currently we append all of the positng list data to a Vec that we sort. This is inefficient because we are sorting data that is already almost sorted. We could...
Tantivy requires that DocId never reach 2^31. Right now nothing prevents a merge operation to reach this value. This is a very unlikely thing to happen, but we should detect...
I don't understand the semantic difference between `append_entry_to_log` and `replicate_to_log`. Would it be correct to implement one using the other? Could we have a default implementation?