High-performance regex operations for Python using Rust backends with parallel processing support.

⚠️ Warning: This is a raw development project and may be unstable. Use at your own risk.

• Python 3.12+
• Rust toolchain
• uv package manager

git clone https://github.com/Aljutor/yurki.git cd yurki uv sync maturin develop

For performance testing, use release build:

maturin develop --release

• Custom Python types: yurki.List (immutable) and yurki.String match the Python 3.12 object layout but use a Rust-side allocator, avoiding the CPython heap.
• SIMD Unicode reader: vectorised path that converts Python text to Rust &str.
• Bump allocator: thread-local arena for short-lived allocations; resets automatically, minimising locking and fragmentation.
• Parallel processing: Rayon work pool distributes work across available cores.

Find Operations:

• 4 jobs: 0.72s vs Python: 12.49s (17.4x speedup)
• 1 job: 2.30s vs Python: 12.49s (5.4x speedup)

Match Operations:

• 4 jobs: 0.33s vs Python: 11.67s (35.2x speedup)
• 1 job: 1.27s vs Python: 11.67s (9.2x speedup)

Capture Operations:

• 4 jobs: 2.83s vs Python: 16.97s (6.0x speedup)
• 1 job: 6.58s vs Python: 16.97s (2.6x speedup)

Replace Operations:

• 4 jobs: 0.64s vs Python: 3.73s (5.9x speedup)
• 1 job: 1.76s vs Python: 3.73s (2.1x speedup)

Split Operations:

• 4 jobs: 1.33s vs Python: 7.15s (5.4x speedup)
• 1 job: 3.34s vs Python: 7.15s (2.1x speedup)

Note: Performance varies by operation type and dataset characteristics. Threading overhead may impact performance on small datasets.

import yurki.regexp as regexp data = ['hello world', 'test 123', 'no match here'] # Find first regex match in each string # Returns list of matched strings (empty string if no match) regexp.find(data, pattern, case=False, jobs=1, inplace=False) regexp.find(data, r'\d+') # ['', '123', ''] # Check if each string matches pattern  # Returns list of booleans regexp.is_match(data, pattern, case=False, jobs=1, inplace=False) regexp.is_match(data, r'\d+') # [False, True, False] # Capture regex groups # Returns list of lists: [full_match, group1, group2, ...] regexp.capture(data, pattern, case=False, jobs=1, inplace=False) regexp.capture(data, r'(\w+) (\d+)') # [[], ['test 123', 'test', '123'], []] # Split strings by regex delimiter # Returns list of lists regexp.split(data, pattern, case=False, jobs=1, inplace=False) regexp.split(['a,b;c', 'x,y'], r'[,;]') # [['a', 'b', 'c'], ['x', 'y']] # Replace regex matches  # Use count=0 for all matches. Supports backreferences ($1, $2) regexp.replace(data, pattern, replacement, count=1, case=False, jobs=1, inplace=False) regexp.replace(data, r'\d+', 'NUM') # ['hello world', 'test NUM', 'no match here'] # Parallel processing for large datasets regexp.find(large_data, pattern, jobs=4) # In-place operations for memory efficiency regexp.replace(data, pattern, replacement, inplace=True)

Parameters:

• data: List of strings to process
• pattern: Regex pattern string
• case: Case-insensitive matching when True
• jobs: Number of parallel workers
• inplace: Modify original list when True

MIT