termunicode 0.3.0

Unicode library to get character properties (width, emoji, printable, etc)

0.3.0

• feature: widthString gained a scalar-scan fast path — code-unit widths are summed straight from the table while every unit is a BMP scalar of non-zero width; the first surrogate or zero-width unit (combining marks, ZWJ, variation selectors, controls) falls back to the grapheme-cluster path. One ASCII char drops from ~225 to ~48 ns/call, a 40-char ASCII line from ~5.0 to ~0.3 µs, a 12-char CJK line from ~6.4 to ~0.5 µs; strings the scan cannot answer pay a few percent for the wasted scan. Divergence: a prepend-class cluster (e.g. U+0600 + digit) is summed per codepoint on the fast path, matching what terminals draw, while the cluster path counts only its first codepoint.
• chore: the widthChars cluster fold is a plain loop now; the outdated benchmark_width.dart was removed (benchmark_width_string.dart is the width benchmark).
• breaking change: emoji width policy narrowed — the table forces width 2 only for Emoji_Presentation codepoints. Text-presentation-default symbols such as U+2602 umbrella now measure 1 (previously 2); an emoji presentation selector (VS16) still makes them 2. Flags and other emoji-presentation characters are unchanged.
• feature: the text presentation selector VS15 (U+FE0E) is honored — a cluster carrying it measures 1. Terminals that ignore VS15 draw the glyph wider than measured; that is their deficiency.
• fix: a lone variation selector now measures 0 (a bare VS16 measured 2).
• fix: EmojiDataUCD.find binary-searched overlapping ranges and missed 46 covered codepoints (U+1F484 lipstick, U+1F48B kiss mark and the U+1F493-1F497 hearts among them), which shipped with the emoji bit unset; isEmojiCp now answers true for them. EmojiDataUCD stores rows per property and gained findProp(property, cp).
• chore: update unicode data to 17.0.0. Unicode 17 removed non-emoji symbols (e.g. U+2605 black star) from Extended_Pictographic, so isEmojiCp no longer reports them. The generated table grows ~2.7% (157.6 KB to 161.9 KB).

0.2.0

• fix: include RIS emojis as double width
• feature: refactor generator and added smoke tests
• feature: improve tests for ucd parsers
• chore: update unicode data to 16.0.0
• fix: correct width for some ambiguous characters
• fix: emoji-ambiguous characters widths
• fix: improve performance of width calculation
• Added new function widthChars that use the Characters package.
• Updated widthString string to use widthChars function.

0.1.0

• Initial version.