a2tm7a / exegr

Python port for rust's regex which is in turn a port of google's re2

Exegr

Trying to optimize code that relied on python's regex brought me to a point where I was still looking at ~400 (at times unicode rich) patterns and 10 sentences to search in. This search cost the algorithm over 100ms. It came to a point where it seemed there is no other way than to shed some code which meant a sacrifice of accuracy over speed.

I read about the problems with the implementation and better alternatives to regex libraries. I found re2 and its python wrapper the lack of features, change in APIs throughout the codebase meant having to keep this as a step to try later.

I decided to rewrite the particular function in Rust and port it back to Python using pyo3. The results have been dramatically huge.

This might not be helpful for general purpose usage due to poor abstraction for the sake of quick testing.

Installation

You can easily add exegr to your python projects like:

pip install exegr

For testing and development use the following:

cargo build --release
cp target/libexegr.so /path/to/py/project

On windows and linux, you can build normally with cargo build --release. On Mac Os, you need to set additional linker arguments. The simplest solution is to create a .cargo/config with the following content:

[target.x86_64-apple-darwin]
rustflags = [
  "-C", "link-arg=-undefined",
  "-C", "link-arg=dynamic_lookup",
]

Usage

from exegr import GroupRegexMatch

patterns: List[str] = []
sentences: List[str] = []
# load patterns
...
group_regex_match = GroupRegexMatch(patterns)

"""
The sentences passes to this function
are expected to be ASR output which provide
multiple options. If your input is a raw string
use sentences = [sentence] instead.
"""
group_regex_match(sentences)