Implement fix for mojibake

[audiomuze]

import polars as pl
import polars.selectors as cs

def repair_all_mojibake(df: pl.DataFrame) -> pl.DataFrame:
    """
    Vectorized cleanup for every string column in a Polars DataFrame.
    Fixes Greek, Cyrillic, Central European, and global UTF-8 encoding corruptions.
    """
    
    # Target all columns containing text strings
    string_cols = cs.string()
    
    return df.with_columns(
        pl.coalesce(
            # 1. Try treating as Global UTF-8 Mojibake (e.g., Café -> Café)
            pl.col(string_cols).str.encode("cp1252").str.decode("utf-8", strict=False),
            
            # 2. Try treating as Legacy Greek Mojibake (e.g., Ñüæá -> Ρόζα)
            pl.col(string_cols).str.encode("cp1252").str.decode("cp1253", strict=False),
            
            # 3. Try treating as Legacy Cyrillic Mojibake (e.g., Çåìôèñà -> Земфира)
            pl.col(string_cols).str.encode("cp1252").str.decode("cp1251", strict=False),
            
            # 4. Try treating as Legacy Central European (e.g., Krakã³w -> Kraków)
            pl.col(string_cols).str.encode("cp1252").str.decode("cp1250", strict=False),
            
            # 5. Fallback: If no decoding strategy worked, keep the original raw text untouched
            pl.col(string_cols)
        ).alias(cs.expand()) # Keeps original column names intact
    )

# =====================================================================
# Demonstration & Verification
# =====================================================================

# Simulated messy multi-column music database
dirty_df = pl.DataFrame({
    "song_title": ["Ñüæá", "Café Blue", "Çåìôèñà Track", "Clean English Song"],
    "artist_name": ["Íôáëüñáò", "Niño Pã©rez", "Unknown", "Stamatis Kokotas"],
    "album_id": [101, 102, 103, 104] # Integers are automatically skipped by the selector
})

clean_df = repair_all_mojibake(dirty_df)
print(clean_df)