• Port pangu.js v10.4.0 plus unreleased master changes up to 8cd3a97d
  • Add spaces around -, *, =, and & between CJK and an opening bracket: CJK-(A) becomes CJK - (A)
  • Add spaces around - between brackets when a hyphen on the same line touches CJK: CJK-CJK[A]-(A) becomes CJK - CJK [A] - (A)
  • Keep bracket globs like *[0-9].log intact
  • Keep mo店+ attached as a name suffix
  • Convert only a lone, tight middle dot to ・; keep runs like •••• and middle dots with spaces around them
• Include LICENSE in the wheel and the sdist, and CHANGELOG.md in the sdist

• Implement Paranoid Text Spacing Algorithm v10 (ported from pangu.js v10.1.1)
  • Never add spaces around / next to CJK, except in file paths like /src/index.ts
  • Keep everything inside http:// and https:// URLs untouched
  • + between a word and CJK is a separator now, not a suffix: A+CJK becomes A + CJK
  • - between CJK and digits is a separator now, not a minus sign
  • Keep superscripts like ², ™, and ® attached to the text on their left
  • Add spaces between CJK and Letterlike Symbols like ℃ and №, and after © before digits
• Support Python 3.8+
• Rename spacing_text() to space_text() and spacing_file() to space_file()
• CLI -h shows the version at the start of the help text

• Implement Paranoid Text Spacing Algorithm v9 (ported from pangu.js v9.1.0)
• Support Python 3.11+
• Fully type-annotated (py.typed)
• Add a method: pangu.has_proper_spacing()
• Add a command-line tool: pangu-py (an alias of the original pangu, to avoid a name collision)
• Add a CLI option: -c/--check, which exits 0 if the text has proper spacing, 1 if not
• Add an encoding parameter to spacing_file(), which now reads files as UTF-8 by default
• Remove pangu.spacing(), use pangu.spacing_text()

Port of the pangu.js v9 text-spacing engine — same algorithm, same behavior, parity-locked by a 1:1 port of the pangu.js test suite.

Breaking changes:

• Half-width punctuation is no longer converted to full-width. The only conversions left: a colon in direct CJK contact right before parentheses becomes ：, and middle dots (· • ‧) normalize to ・. There is no flag to restore the old behavior
• pangu.spacing() is removed; use pangu.spacing_text()
• Output is no longer stripped of leading/trailing whitespace
• spacing_file() reads files as UTF-8 by default (new keyword-only encoding parameter) instead of the locale encoding
• The CLI reads stdin only when no argument is given; an explicit argument wins over piped stdin, and an explicit - argument always means stdin
• CLI usage errors exit 2 (argparse convention)
• Requires Python >= 3.11

New:

• HTML tag protection (attribute values still spaced), tag mentions, backtick protection, compound words, file paths, per-line slash/pipe/plus readings, affix readings
• pangu.has_proper_spacing(text)
• CLI -c/--check: exit 0 if the text has proper spacing, 1 if not (corrected text on stderr)

Internal:

• Packaging moved to pyproject.toml + uv_build; src/pangu/ package layout replaces the single-file module
• The engine (pangu._core) deliberately mirrors pangu.js src/shared/index.ts so upstream releases port as a mechanical diff (see docs/adr/0001)

• Implement Paranoid Text Spacing Algorithm v4
• Support Python 3.7
• Drop Python 2.7 support

• Support Python 3.6
• Add a method: pangu.spacing_file()
• Add a command-line tool: pangu

• Support Python 3.5
• Refactoring
• Rename text_spacing() to spacing_text()

• Add an alias to spacing(): text_spacing()
• Fix unicode issue in Python 2.x

• Fix setup.py

• Implement Paranoid Text Spacing Algorithm v2
• Synchronize version number with pangu.js

• Hello World