icupy

Python bindings for ICU4C using pybind11.

Changes from ICU4C

• Naming Conventions

  Renamed functions, methods, and enums to conform to PEP 8.

  • Function Names: use lower_case_with_underscores style.
  • Method Names: use lower_case_with_underscores style. Also, use one leading underscore only for protected methods.
  • C++ Enum Member Names: use UPPER_CASE_WITH_UNDERSCORES style without a leading "k". (e.g., kDateOffset → DATE_OFFSET)
  • APIs that match Python reserved words: e.g.,
    • with() → with_()

• Error Handling

  • ICU C/C++ API errors are raised as icupy.icu.ICUError exceptions. The underlying UErrorCode can be retrieved from the error_code attribute of the exception.

    For example:

    from icupy import icu
    try:
        pass  # Call ICU API here...
    except icu.ICUError as e:
        print(e.error_code)  # → icu.ErrorCode
        print(e.error_code.get())  # → icu.UErrorCode
    

Examples

• icu::UnicodeString with predefined error callback function

  # from Unicode to codepage
  from icupy import icu
  cnv = icu.ucnv_open("iso8859-1")
  context = icu.UserContext(icu.UCNV_ESCAPE_C)  # \uXXXX
  action = icu.UConverterFromUCallback(icu.UCNV_FROM_U_CALLBACK_ESCAPE, context)
  old_action = icu.ucnv_set_from_u_call_back(cnv, action)
  s = icu.UnicodeString("A€B")
  s.extract(cnv)  # → b'A\\u20ACB'
  

  # from codepage to Unicode
  from icupy import icu
  cnv = icu.ucnv_open("Shift-JIS")
  context = icu.UserContext(icu.UCNV_ESCAPE_XML_HEX)  # &#xXXXX;
  action = icu.UConverterToUCallback(icu.UCNV_TO_U_CALLBACK_ESCAPE, context)
  old_action = icu.ucnv_set_to_u_call_back(cnv, action)
  src = b"\x61\xeb\x40\x62"  # 0xeb 0x40: UNASSIGNED SEQUENCE
  s = icu.UnicodeString(src, -1, cnv)
  str(s)  # → 'aë@b'
  

• icu::UnicodeString with custom error callback function

  # from Unicode to codepage
  from icupy import icu
  from icupy.utils import gc
  def from_unicode_cb(
      options: object,
      args: icu.UConverterFromUnicodeArgs,
      code_units: str,
      length: int,
      code_point: int,
      reason: icu.UConverterCallbackReason,
      error_code: icu.ErrorCode,
  ) -> None:
      _ = options, length, code_point  # unused
      if reason in [icu.UCNV_UNASSIGNED, icu.UCNV_ILLEGAL, icu.UCNV_IRREGULAR]:
          error_code.set(icu.U_ZERO_ERROR)
          source = "".join(f"\\u{ord(c):04x}" for c in code_units)
          icu.ucnv_cb_from_u_write_bytes(args, source, len(source), 0)
  
  with gc(icu.ucnv_open("iso8859-1"), icu.ucnv_close) as cnv:
      action = icu.UConverterFromUCallback(from_unicode_cb)
      old_action = icu.ucnv_set_from_u_call_back(cnv, action)
      s = icu.UnicodeString("A€B")
      s.extract(cnv)  # → b'A\\u20acB'
  

  # from codepage to Unicode
  from icupy import icu
  from icupy.utils import gc
  def to_unicode_cb(
      options: object,
      args: icu.UConverterToUnicodeArgs,
      code_units: bytes,
      length: int,
      reason: icu.UConverterCallbackReason,
      error_code: icu.ErrorCode,
  ) -> None:
      _ = options, length  # unused
      if reason in [icu.UCNV_UNASSIGNED, icu.UCNV_ILLEGAL, icu.UCNV_IRREGULAR]:
          error_code.set(icu.U_ZERO_ERROR)
          source = "".join(f"%{b:02X}" for b in code_units)
          icu.ucnv_cb_to_u_write_uchars(args, source, len(source), 0)
  
  with gc(icu.ucnv_open("Shift-JIS"), icu.ucnv_close) as cnv:
      action = icu.UConverterToUCallback(to_unicode_cb)
      old_action = icu.ucnv_set_to_u_call_back(cnv, action)
      src = b"\x61\xeb\x40\x62"  # 0xeb 0x40: UNASSIGNED SEQUENCE
      s = icu.UnicodeString(src, -1, cnv)
      str(s)  # → 'a%EB%40b'
  

• icu::BreakIterator for word-breaks

  from icupy import icu
  bi = icu.BreakIterator.create_word_instance("en_US")
  src = icu.UnicodeString("Alice was beginning to get very tired of sitting by her sister on the bank.")
  bi.set_text(src)
  result = []
  start = bi.first()
  while (end := bi.next()) != icu.UBRK_DONE:
      if bi.get_rule_status() != icu.UBRK_WORD_NONE:
          result.append(src[start:end])
      start = end
  
  # result: ['Alice', 'was', 'beginning', 'to', 'get', 'very', 'tired', 'of', 'sitting', 'by', 'her', 'sister', 'on', 'the', 'bank']
  

• Natural sort (human-friendly sorting)

  from icupy import icu
  coll = icu.Collator.create_instance("en_US")
  coll.set_attribute(icu.UCOL_NUMERIC_COLLATION, icu.UCOL_ON)
  data = ["file1.txt", "file10.txt", "file2.txt", "file20.txt", "file3.txt"]
  sorted(data, key=coll.get_sort_key)
  # ['file1.txt', 'file2.txt', 'file3.txt', 'file10.txt', 'file20.txt']
  

• icu::IDNA (UTS #46)

  from icupy import icu
  uts46 = icu.IDNA.create_uts46_instance(icu.UIDNA_DEFAULT | icu.UIDNA_CHECK_BIDI | icu.UIDNA_CHECK_CONTEXTJ)
  dest = icu.UnicodeString()
  info = icu.IDNAInfo()
  # a + ZERO WIDTH NON-JOINER + b.com
  uts46.name_to_ascii("a\u200cb.com", dest, info)  # → 'xn--ab-j1t.com'
  bool(info.get_errors() & icu.UIDNA_ERROR_BIDI)  # → False
  bool(info.get_errors() & icu.UIDNA_ERROR_CONTEXTJ)  # → True
  

• icu::number::NumberFormatter (ICU 60+)

  from icupy import icu
  from icupy.icu import number
  template = (
      number.NumberFormatter.with_()
      .notation(number.Notation.compact_short())
      .unit(icu.CurrencyUnit("EUR"))
      .precision(number.Precision.max_significant_digits(2))
  )
  template.locale("en_US").format_int(1234).to_string()  # "€1.2K" in en-US
  

• icu::RegexMatcher::find with custom callback function

  from icupy import icu
  src = icu.UnicodeString("aaaaaaaaaaaaaaaaaaab")
  matcher = icu.RegexMatcher("((.)\\2)x", src, 0)
  def progress_callback(options: dict[str, int], match_index: int) -> bool:
      if not isinstance(options, dict):
          return False
      calls = options.get("numCalls", 0) + 1
      options["numCalls"] = calls
      options["lastIndex"] = match_index
      max_calls = options.get("maxCalls", -1)
      return True if max_calls < 0 else calls < max_calls
  
  info = {}
  context = icu.UserContext(info)
  callback = icu.URegexFindProgressCallback(progress_callback, context)
  matcher.set_find_progress_callback(callback)
  matcher.find(0)  # → False
  # info: {'numCalls': 18, 'lastIndex': 18}
  info.clear()
  info["maxCalls"] = 5
  matcher.find(0)  # → ICUError: U_REGEX_STOPPED_BY_CALLER
  # info: {'maxCalls': 5, 'numCalls': 5, 'lastIndex': 5}
  

• icu::number::SimpleNumberFormatter (ICU 73+)

  from icupy import icu
  from icupy.icu import number
  fmt = number.SimpleNumberFormatter.for_locale_and_grouping_strategy("de-CH", icu.UNUM_GROUPING_ON_ALIGNED)
  fmtval = fmt.format_int64(1234567)
  fmtval.to_string()  # → "1'234'567"
  

• Subclassing icu::Transliterator

  # Uppercase letters while skipping text enclosed in backticks
  from icupy import icu
  class TestTransliterator(icu.Transliterator):
      def __init__(self, filter_set: icu.UnicodeSet | None = None) -> None:
          icu.Transliterator.__init__(self, "Any-UpperWithoutCode", filter_set)
      def _handle_transliterate(
          self,
          text: icu.Replaceable,
          pos: icu.UTransPosition,
          incremental: bool,
      ) -> None:
          # Implement the transliteration algorithm here.
          cursor = pos.start
          in_backtick = False
          while cursor < pos.limit:
              c = text.char32_at(cursor)
              char_len = icu.u16_length(c)
              if c == 0x60:
                  in_backtick = not in_backtick
                  cursor += char_len
                  continue
              if not in_backtick and icu.u_isalpha(c) and icu.u_islower(c):
                  upper = icu.u_toupper(c)
                  if upper != c:
                      text.handle_replace_between(cursor, cursor + char_len, chr(upper))
                      char_len = icu.u16_length(upper)
              cursor += char_len
          pos.start = pos.limit
  
  tl = TestTransliterator()
  text = icu.UnicodeString("Subclasses must implement `_handle_transliterate()`, which defines their own transliteration algorithm.")
  tl.transliterate(text)
  # text: "SUBCLASSES MUST IMPLEMENT `_handle_transliterate()`, WHICH DEFINES THEIR OWN TRANSLITERATION ALGORITHM."
  

Installation

Prerequisites

• Python >=3.10
• ICU4C (ICU - The International Components for Unicode) (>=70 recommended)
• C++17 compatible compiler (see Supported Compilers)
• CMake >=3.15

Installing prerequisites

• Windows:

  Install the following dependencies:

  • Python >=3.10
  • Pre-built ICU4C binary package (>=70 recommended)
  • C++17 compatible compiler. Visual Studio 2022 or newer recommended
  • CMake >=3.15
    • Note: Add CMake to the system PATH.

• Linux:

  To install dependencies, run the following command:

  • Ubuntu/Debian:

    sudo apt install g++ cmake libicu-dev python3-dev python3-pip
    

  • Fedora:

    sudo dnf install gcc-c++ cmake icu libicu-devel python3-devel
    

  Note: If your system's ICU is out of date, consider building ICU4C from source or installing pre-built ICU4C binary package.

Installing icupy

1. Configuring environment variables

   • Windows:

     • Set the ICU_ROOT environment variable to the root of the ICU installation.

       For example, if the ICU is located in C:\icu4c:

       in PowerShell:

       $env:ICU_ROOT = "C:\icu4c"
       

       or in Command Prompt:

       set ICU_ROOT=C:\icu4c
       

     • To verify settings using icuinfo (64-bit):

       in PowerShell:

       & $env:ICU_ROOT\bin64\icuinfo
       

       or in Command Prompt:

       %ICU_ROOT%\bin64\icuinfo
       

   • Linux:

     • If the ICU is located in a non-regular place, set the PKG_CONFIG_PATH and LD_LIBRARY_PATH environment variables.

       For example, if the ICU is located in /usr/local:

       export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig:$PKG_CONFIG_PATH
       export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
       

     • To verify settings using pkg-config:

       pkg-config --cflags --libs icu-uc
       # -I/usr/local/include -L/usr/local/lib -licuuc
       

2. Installing from PyPI

   pip install icupy
   

   Optionally, CMake environment variables are available. For example, using the Ninja build system and Clang:

   CMAKE_GENERATOR=Ninja CXX=clang++ pip install icupy
   

   Alternatively, installing development version from the git repository:

   pip install git+https://github.com/miute/icupy.git
   

Usage

1. Configuring environment variables

   • Windows:

     • Set the ICU_ROOT environment variable to the root of the ICU installation (default is C:\icu).

       For example, if the ICU is located in C:\icu4c:

       in PowerShell:

       $env:ICU_ROOT = "C:\icu4c"
       

       or in Command Prompt:

       set ICU_ROOT=C:\icu4c
       

   • Linux:

     • If the ICU is located in a non-regular place, set the LD_LIBRARY_PATH environment variables.

       For example, if the ICU is located in /usr/local:

       export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
       

2. Using icupy

   import icupy.icu as icu
   # or
   from icupy import icu
   

License

This project is licensed under the MIT License.