In this project, I evaluated the adoption of Rust (via PyO3) inside one of my personal projects, sysadmindb.
The idea was to replace the main regex parser from Python’s re, to Rust’s regex crate.
I benchmarked the current implementation and the proposed Rust version using pytest-benchmark.
The results showed that the “pure” Python implementation was faster than Rust’s (~1.4x).
After several trials and error, and further benchmarking, I concluded that for this specific regex (full of capturing groups)
Python’s re is faster that Rusts regex crate, hence, independent of the FFI boundary overhead, the python implementation would always be more performant.<\/p>\n\n\n\n
SysadminDB stores logs that can be queried using unix\/linux powertools such as grep, sed and awk.
The project is written in Python, and it contains a TCPServer to receive the logs, and an HTTPServer to query the logs.
The database is Sqlite.<\/p>\n\n\n\n
The main task of the TCPServer is to receive, parse and insert logs into the database.
The objective of this experiment was to improve the performance of this server by switching the regex parsing from Python to Rust.
This was chosen somewhat ad hoc, under the assumption that the migration would be relatively easy and provide a good ROI.<\/p>\n\n\n\n
This is the Python implementation:<\/p>\n\n\n\n