MatchingWords
structure uses too much memory
#3115
Labels
milli
Related to the milli workspace
performance
Related to the performance in term of search/indexation speed or RAM/CPU/Disk consumption
v1.0.0
PRs/issues solved in v1.0.0 released on 2023-02-06
Milestone
Certain search queries create very large DFAs (deterministic finite automata) in
MatchingWords
. They are the single biggest consumer of RAM within the execution a search query. This is because the DFAs:Simply deduplicating the DFAs and avoiding creating them for long words would already go a long way towards mitigating the problem. I wrote a proof of concept that deduplicates the DFAs on the branch
cache_matching_words
of milli. It does indeed dramatically reduce memory usage and also speed up some particular search requests significantly.TODO
main
The text was updated successfully, but these errors were encountered: