We have at times some portability issues as ScanCode depends on a few performance-critical libraries that use native C code. These require specific pre-built wheels or a toolchain that may not be available on a user's machine.

We could maximize portability by using fallback pure-python libraries for these libraries. For basic features in scancode-toolkit-mini, these would be pyahocorasick and intbitset. Both are libraries that we maintain.

We should have a degraded, not-as-fast but pure Python variants of these libraries so that we can install without these native dependencies, e.g. some scancode-toolkit variant that would be pure Python. This would help also towards running in the browser.

key libraries

• pyahocorasick: could expand the built-in simpler pure python implementation to implment the pyahocorasick APIs
• intbitset: could roll out a simple set-based fallback
• lxml (and other libs based on it such as xmldict): could use stdlib xml.etree instead

For lxml there are common idioms such as:

    try:
        from lxml import etree
    except ImportError:
        import xml.etree.ElementTree as etree