Line-Aware Editing

Text editors are line-oriented at the user level: cursor moves up/down by line, line numbers display, selections often span lines.

Track lines:

• For each line: byte offset where it starts.
• Update on insert/delete.

On insert: increment all line_starts after pos by len(inserted), and add new line_starts for newlines in inserted text.

On delete: similar, decrement and remove crossed newlines.

Modern editors maintain this incrementally for O(log N) updates (e.g., piece-table extends with line break index per piece).

Tab handling:

• Tabs expand to spaces visually (e.g., tab = 8 cols, or 4 cols).
• Some editors store as actual tab char; visually align.
• Configuration nightmare.

Unicode awareness:

• Byte position vs character position vs grapheme cluster.
• Emoji (👨‍👩‍👧‍👦 = 7 codepoints + ZWJ joiners) are one visible character.
• Backspace deletes the whole grapheme cluster.
• Cursor moves visually, not by codepoint.

UTF-8 encoded:

• ASCII = 1 byte.
• Latin = 2 bytes.
• BMP = 3 bytes.
• Emoji + supplementary = 4 bytes.

For internal storage, byte position is fine. For display, need to know grapheme boundaries.

Bidirectional text (Arabic, Hebrew):

• Mixed LTR + RTL.
• Visual order ≠ logical order.
• Unicode bidi algorithm handles.
• Most editors get this approximately right; edge cases break.