Broken Character Repair

Good to know

• This repairs one specific accident: text written as UTF-8 and then read as a Windows code page. The letters are all still there, just wearing the wrong clothes, so putting them back is exact rather than a guess.
• The opposite accident cannot be undone. Once characters show up as replacement characters the bytes behind them were thrown away at save time, and nothing can reconstruct them.
• Automatic detection tries each code page and keeps only a result that is genuinely less broken than what it started with, so text that was never damaged is returned untouched. Pick the code page yourself if you know it.
• Text that went through the accident twice — "ç" rather than "ç" — is repaired in as many passes as it takes, up to three.