Open alexanderguzhva opened 5 months ago
@alexanderguzhva Agreed, I would prefer to use RE2. How much work is it?
@xiaofan-luan in the best case scenario, the work is to brainlessly replace std::regex
with re2 facilities and make sure that all unit tests pass.
Also, it seems that Milvus uses the Rust library called 'tantivy', which also has some regex facilities, but I'm not familiar with the library.
@xiaofan-luan in the best case scenario, the work is to brainlessly replace
std::regex
with re2 facilities and make sure that all unit tests pass. Also, it seems that Milvus uses the Rust library called 'tantivy', which also has some regex facilities, but I'm not familiar with the library.
So currently if user build tantivy index, we will tantivy to accelerate. Otherwise it is bruteforce search and we use re2.
Is there an existing issue for this?
Is your feature request related to a problem? Please describe.
std::regex
is notoriously slow and is a part of STL because it is a reference stable engine, not because of its performance.Describe the solution you'd like.
Consider replacing
std::regex
with an alternative regex engine, such asboost::regex
,Intel Hyperscan
orRE2
, if appropriate.Check the benchmark numbers, for example https://github.com/HFTrader/regex-performance
Describe an alternate solution.
No response
Anything else? (Additional Context)
No response