A fast, parallel, command-line fixer for Delphi source file encodings. It scans .pas, .dpr, …; detects UTF-8/ASCII vs. legacy single-byte encodings; repairs mixed lines (Windows-1250/1252/ANSI), and writes clean UTF-8 with an optional BOM — without breaking your line endings or folder structure.

• Old codebases often mix ANSI and UTF-8 (sometimes within a single file).
• Delphi 12+ works best when sources are in UTF-8 (usually with BOM).
• Manual converting is slow and risky. This tool automates it safely and fast.

EncodingFixTool [params]

Examples:

:: Dry run, current folder, defaults (.pas,.dpr, recursive, BOM=yes)
EncodingFixTool dry

:: Fix a project tree verbosely and back it up before changes
EncodingFixTool path=C:\Projs\MyApp v bkp-dir=C:\backup\myapp

:: Only scan .pas files in ./src (not recursive), remove BOM if present
EncodingFixTool path=.\src recursive=n ext=pas utf8-bom=n

:: Multiple ext forms are OK; quoted lists work
EncodingFixTool ext="*.pas,*.dpr, .dfm"

Tip: Parameters accept key=value or key:value. You can prefix flags with - (and / on Windows), e.g. -v, -dry.

If path=C:\tmp\ and bkp-dir=C:\bkp and a processed file is C:\tmp\src\foo\bar\Main.pas → backup is written to:

C:\bkp\src\foo\bar\Main.pas

1. Gather files Walk the path (recursively by default), matching the configured extensions.

2. Process in parallel Uses TParallel.For to utilize multiple cores. Console output is synchronized, and the “files changed” counter is atomic.

3. Detect encoding per file (fast pre-check)

   • Uses System.WideStrUtils.DetectUTF8Encoding on the raw bytes.
   • If US-ASCII → file is left as-is (never add a BOM).
   • If UTF-8 → ensure BOM matches utf8-bom option; adjust if needed (optionally backing up first).
   • Else (treated as ANSI / unknown) → go to step 4.

4. Repair mixed/legacy encodings per line

   • Split the original bytes by raw CR/LF/CRLF (no decoding yet).

   • For each line, try:

     • UTF-8 (strict) — if it round-trips, use it.

     • Windows-1250 (Central Europe/PL), Windows-1252 (Western/DE), and ANSI — decode and score:

       • +2 for valid Polish/German diacritics,
       • +1 for typical source characters (letters/digits/whitespace/common punctuation),
       • −2 for control chars (except tab),
       • −1 for U+FFFD replacements.

     • Pick the highest-scoring decode for that line.

   • Reassemble the file:

     • Preserve the dominant original EOL style (CRLF/LF/CR).
     • Preserve whether the file ended with a trailing EOL.

   • Save as UTF-8 (BOM per utf8-bom), optionally to backup first.

5. Summary At the end prints elapsed time and number of files changed.

Dry run:

Would fix: src\Utils\StrTools.pas (mixed bytes; would save UTF-8 (BOM=Y, EOL=CRLF) (dry-run))
OK   : src\Main.dpr (UTF-8 OK)
Done in 00:12.384. Files changed: 0

Actual run:

Fixed: src\Utils\StrTools.pas (detected Windows-1250; saved UTF-8 (BOM=Y, EOL=CRLF))
Fixed: src\Forms\About.pas (detected UTF-8; Added UTF-8 BOM)
Done in 00:08.972. Files changed: 2

When silent is enabled, only the summary is suppressed too. In dry-run mode, change messages are phrased as “Would fix”.

• BOM policy

  • Default is BOM = yes (Delphi IDE/compiler are happiest this way).
  • US-ASCII files are never given a BOM, even if utf8-bom=y.
  • Existing UTF-8 files are re-saved only if the BOM policy differs.

• Line endings Dominant EOL (CRLF/LF/CR) is detected from raw bytes and preserved; trailing newline presence is preserved.

• Relative reporting Paths in logs are shown relative to the scanned path, for readability.

• 0 – completed (even if some files failed; failures are printed).
• 2 – the path argument didn’t exist.