Encoding & ANSEL

A GEDCOM file is plain text, but “plain text” hides a choice: which encoding maps the file’s bytes to actual characters. Get it wrong and Müller becomes Müller or worse. Older genealogy files — especially from legacy software — often use ANSEL, an encoding you’ll meet almost nowhere else.

You don’t have to think about any of this. When you import a file, Linea Codex inspects the raw bytes, works out the encoding, and decodes your names and places correctly — ANSEL included. This page is the detail behind that promise: how detection runs, what’s supported, what ANSEL is, and how to take manual control when you want it.

How detection runs

Detection is a pipeline of ordered stages. Each looks at the file’s bytes; the first stage that recognises the encoding wins, and later stages aren’t consulted. The stages, in order:

#	Stage	What it looks for	Result
1	Byte-order mark (BOM)	A leading EF BB BF (UTF-8) or FF FE / FE FF (UTF-16)	UTF-8 or UTF-16 (UNICODE), authoritatively
2	UTF-16 without a BOM	A regular pattern of 00 (NUL) bytes in alternating positions	UTF-16 LE or BE
3	ANSEL heuristic	The byte patterns characteristic of ANSEL (see below)	ANSEL
4	Statistical analysis	Byte-frequency analysis to identify the most likely encoding	Whatever scores highest (e.g. a Windows code page)
5	CHAR header scan	The 1 CHAR <value> line in the header, read as ASCII	The encoding the file declares
6	Unknown	Nothing matched with confidence	Flagged for you to choose

A few things worth knowing about this order:

• A BOM is trusted above everything, including the CHAR header — the bytes on disk are the ground truth, and a BOM is unambiguous.
• The CHAR header is consulted late, not first. A declared 1 CHAR value can be wrong (software that writes one encoding but labels it another is common), so the byte-level evidence is weighed first. The header is the tie-breaker, not the oracle.
• The CHAR scan understands more than the four standard GEDCOM values. Well-known non-standard labels such as ANSI (Windows-1252), IBMPC (CP437), and MACINTOSH are recognised, and any value the browser’s decoder accepts is honoured as a fallback.
• If no stage is confident, the encoding is reported as undetermined rather than guessed wrongly — and the import UI asks you to pick (see Overriding the detected encoding).

Supported encodings

Linea Codex reads far more than the encodings GEDCOM officially blesses, because real files use them. The four standard GEDCOM encodings are handled natively; everything else is decoded through the browser’s built-in text decoder.

Encoding	Notes
UTF-8	The modern default; the only encoding GEDCOM 7.0 permits.
UNICODE	In GEDCOM, UNICODE means UTF-16 (little- or big-endian).
ASCII	7-bit; a subset of both UTF-8 and ANSEL.
ANSEL	The legacy genealogy encoding — see the next section.

Beyond these, a broad set of legacy code pages is available for import and export, including the Windows code pages (1250–1258), the ISO 8859 family, CP437 / CP850 / CP866 (DOS), classic Macintosh, and KOI8-R / KOI8-U (Cyrillic). These are the encodings offered when you override detection or choose an output encoding.

Which encodings each GEDCOM version allows

What a file may declare depends on its version. This is the encoding side of the versions & conversion matrix:

Version	Permitted encodings (HEAD.CHAR)	Default
5.5	ANSEL, UNICODE, ASCII	ANSEL
5.5.1	ANSEL, UTF-8, UNICODE, ASCII	ANSEL
7.0	UTF-8 only	— (UTF-8; an optional BOM may lead the file)

GEDCOM 7.0 dropped the CHAR header entirely: every 7.0 file is UTF-8, full stop. This is why converting to 7.0 always lands on a UTF-8-compatible encoding — there’s no other option to land on.

Linea Codex stays permissive

These are the spec-compliant encodings. Linea Codex will still read a 5.5 file that declares, say, ANSI, or a UTF-8 file with no CHAR line at all — it decodes the bytes and lets you work. It warns where a file departs from the standard; it doesn’t refuse to open it.

What ANSEL is

ANSEL (the American National Standard for Extended Latin) was the default character set for GEDCOM 5.5 and 5.5.1, so a large share of older genealogy files are encoded in it. It’s a single-byte encoding: bytes 0x00–0x7F are plain ASCII, and the upper range carries the accented and special letters genealogy needs.

ANSEL has one trait that makes it unusual, and it’s the reason a generic text decoder mangles it:

• Combining diacritics come before the base letter — the opposite of Unicode, where the accent follows the letter it modifies.
• A handful of upper-range bytes are precomposed special characters in their own right — Ł, Ø, Đ, Æ, the British pound sign, the euro sign, and so on.
• The bytes 0x80–0x9F (and a few others) are simply unused — their presence is a strong hint the file isn’t ANSEL, which is exactly what the detection heuristic keys on.

Mapping ANSEL to Unicode

Decoding ANSEL isn’t a byte-for-byte table lookup — it’s a reorder. When Linea Codex reads an ANSEL file it:

1. Reads the leading combining mark(s),
2. reorders them to follow the base letter (Unicode order), and
3. precomposes the result into a single character where one exists (so é is one code point, not a bare e plus a floating accent).

Writing ANSEL reverses this: it decomposes the precomposed character and emits the diacritic first.

A worked example — é (e with acute):

ANSEL bytes:    E2 65          (acute-accent byte, then the letter "e")
Decoded order:  e + ◌́          (base letter, then combining acute — U+0065 U+0301)
Precomposed:    é              (U+00E9)

Notice the byte order: E2 (the acute) precedes 65 (the e). A decoder that assumed Unicode order would read it as an accent floating on whatever came before, which is how ANSEL files turn to mojibake in tools that don’t know the convention. Linea Codex’s ANSEL codec handles the reorder and precomposition both ways, so a round-trip through the app preserves your accented names exactly.

Not every character survives a downgrade

ANSEL is a fixed repertoire. A name written in, say, Cyrillic or Greek has no ANSEL representation, so writing it out as ANSEL can’t be faithful. That’s why downgrading toward an ANSEL-era version is a best-effort operation, and why the export and conversion dialogs check every character first and tell you which ones can’t be represented. See Converting GEDCOM versions.

Overriding the detected encoding

Detection is right the overwhelming majority of the time, but you stay in control. On the start page’s import dialog, the detected encoding is shown inline (Detected: …) so you can confirm it at a glance. If you disagree — or if detection couldn’t decide and shows Could not detect file encoding. Please select manually. — open the Encoding dropdown and pick the right one before importing. There’s no extra confirmation step: your choice takes effect when you import.

When you start a new project instead of importing, you choose the encoding up front alongside the GEDCOM version. Picking 7.0 locks the encoding to UTF-8 (GEDCOM 7.0 requires UTF-8 encoding.), since that’s the only encoding the version allows.

Choosing an output encoding on export

The encoding a file was imported with is fixed inside the project. When you export, though, you can write the file in a different encoding to suit wherever it’s headed — without touching the project’s internal encoding.

File → Export As… opens a picker for the target encoding and validates the whole document against it. It tells you either All characters are compatible with … or exactly which characters can’t be represented and how often each occurs, with a nudge to Select an encoding that supports all characters in the document, such as UTF-8. The full walkthrough — including GEDZIP and ZIP archives — is in Saving & exporting.

The same character-compatibility check runs automatically when you convert to a version whose encodings can’t hold your data, or when you save a file whose imported encoding isn’t a standard GEDCOM one — Linea Codex surfaces the mismatch and offers a compatible encoding rather than silently dropping characters.

---

See also: Import a file · Saving & exporting · Converting GEDCOM versions · Versions & conversion matrix.