Module scoring 

Surprisal-based line scoring for deterministic compression.

Each line receives an information density score based on:

• Character-level entropy (Shannon)
• Structural markers (paths, numbers, identifiers)
• Repetition detection (overlap with previous lines)

Structs

LineScore

Score for a single line — higher means more informative.

Functions

score_lines

Scores all lines in the input text for information density.