Lib.unicodedata.py.orig Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of jython-standalone Show documentation
Show all versions of jython-standalone Show documentation
Jython is an implementation of the high-level, dynamic, object-oriented
language Python written in 100% Pure Java, and seamlessly integrated with
the Java platform. It thus allows you to run Python on any Java platform.
import java.lang.Character
try:
# import from jarjar-ed version
from org.python.icu.text import Normalizer
from org.python.icu.lang import UCharacter, UProperty
from org.python.icu.util import VersionInfo
from org.python.icu.lang.UCharacter import EastAsianWidth, DecompositionType
from org.python.icu.lang.UCharacterEnums import ECharacterCategory, ECharacterDirection
except ImportError:
# development version of Jython, so use extlibs
from com.ibm.icu.text import Normalizer
from com.ibm.icu.lang import UCharacter, UProperty
from com.ibm.icu.util import VersionInfo
from com.ibm.icu.lang.UCharacter import EastAsianWidth, DecompositionType
from com.ibm.icu.lang.UCharacterEnums import ECharacterCategory, ECharacterDirection
__all__ = (
"bidirectional", "category", "combining", "decimal", "decomposition", "digit", "east_asian_width",
"lookup", "mirrored", "name", "normalize", "numeric", "unidata_version")
_forms = {
'NFC': Normalizer.NFC,
'NFKC': Normalizer.NFKC,
'NFD': Normalizer.NFD,
'NFKD': Normalizer.NFKD
}
Nonesuch = object() # to distinguish from None, which is a valid return value for some functions
def _validate_unichr(unichr):
if not(isinstance(unichr, unicode)):
raise TypeError("must be unicode, not {}".format(type(unichr).__name__))
if len(unichr) > 1 or len(unichr) == 0:
raise TypeError("need a single Unicode character as parameter")
def _get_codepoint(unichr):
_validate_unichr(unichr)
return ord(unichr)
def name(unichr, default=Nonesuch):
# handle None
n = UCharacter.getName(_get_codepoint(unichr))
if n is None:
if default is not Nonesuch:
return default
else:
raise ValueError("no such name")
return n
def lookup(name):
codepoint = UCharacter.getCharFromName(name)
if codepoint == -1:
raise KeyError("undefined character name '{}".format(name))
return unichr(codepoint)
def digit(unichr, default=Nonesuch):
d = UCharacter.digit(_get_codepoint(unichr))
if d == -1:
if default is not Nonesuch:
return default
else:
raise ValueError("not a digit")
return d
def decimal(unichr, default=Nonesuch):
d = UCharacter.getNumericValue(_get_codepoint(unichr))
if d < 0 or d > 9:
if default is not Nonesuch:
return default
else:
raise ValueError("not a decimal")
return d
def numeric(unichr, default=Nonesuch):
n = UCharacter.getUnicodeNumericValue(_get_codepoint(unichr))
if n == UCharacter.NO_NUMERIC_VALUE:
if default is not Nonesuch:
return default
else:
raise ValueError("not a numeric")
return n
_decomp = {
DecompositionType.CANONICAL: "canonical",
DecompositionType.CIRCLE: "circle",
DecompositionType.COMPAT: "compat",
DecompositionType.FINAL: "final",
DecompositionType.FONT: "font",
DecompositionType.FRACTION: "fraction",
DecompositionType.INITIAL: "initial",
DecompositionType.ISOLATED: "isolated",
DecompositionType.MEDIAL: "medial",
DecompositionType.NARROW: "narrow",
DecompositionType.NOBREAK: "nobreak",
DecompositionType.NONE: None,
DecompositionType.SMALL: "small",
DecompositionType.SQUARE: "square",
DecompositionType.SUB: "sub",
DecompositionType.SUPER: "super",
DecompositionType.VERTICAL: "vertical",
DecompositionType.WIDE: "wide"
}
def _get_decomp_type(unichr):
if unichr == u"\u2044": # FRACTION SLASH
# special case this for CPython compatibility even though this returns as not being combining, eg, see
# http://www.fileformat.info/info/unicode/char/2044/index.htm
return "fraction"
else:
return _decomp[UCharacter.getIntPropertyValue(ord(unichr), UProperty.DECOMPOSITION_TYPE)]
def decomposition(unichr):
_validate_unichr(unichr)
d = Normalizer.decompose(unichr, True)
decomp_type = None
if len(d) == 1:
decomp_type = _get_decomp_type(unichr)
else:
for c in d:
decomp_type = _get_decomp_type(c)
# print "Got a decomp_type %r %r %r" % (c, d, decomp_type)
if decomp_type is not None:
break
hexed = " ".join(("{0:04X}".format(ord(c)) for c in d))
if decomp_type:
return "<{}> {}".format(decomp_type, hexed)
elif len(d) == 1:
return ""
else:
return hexed
# To map from ICU4J enumerations for category, bidirection, and
# east_asian_width to the underlying property values that Python uses
# from UnicodeData.txt required a manual mapping between the following
# two files:
#
# http://icu-project.org/apiref/icu4j/constant-values.html
# http://www.unicode.org/Public/6.3.0/ucd/PropertyValueAliases.txt
_cat = {
ECharacterCategory.COMBINING_SPACING_MARK: "Mc",
ECharacterCategory.CONNECTOR_PUNCTUATION: "Pc",
ECharacterCategory.CONTROL: "Cc",
ECharacterCategory.CURRENCY_SYMBOL: "Sc",
ECharacterCategory.DASH_PUNCTUATION: "Pd",
ECharacterCategory.DECIMAL_DIGIT_NUMBER: "Nd",
ECharacterCategory.ENCLOSING_MARK: "Me",
ECharacterCategory.END_PUNCTUATION: "Pe",
ECharacterCategory.FINAL_PUNCTUATION: "Pf",
ECharacterCategory.FORMAT: "Cf",
# per http://icu-project.org/apiref/icu4j/com/ibm/icu/lang/UCharacterEnums.ECharacterCategory.html#GENERAL_OTHER_TYPES
# - no characters in [UnicodeData.txt] have this property
ECharacterCategory.GENERAL_OTHER_TYPES: "Cn Not Assigned",
ECharacterCategory.INITIAL_PUNCTUATION: "Pi",
ECharacterCategory.LETTER_NUMBER: "Nl",
ECharacterCategory.LINE_SEPARATOR: "Zl",
ECharacterCategory.LOWERCASE_LETTER: "Ll",
ECharacterCategory.MATH_SYMBOL: "Sm",
ECharacterCategory.MODIFIER_LETTER: "Lm",
ECharacterCategory.MODIFIER_SYMBOL: "Sk",
ECharacterCategory.NON_SPACING_MARK: "Mn",
ECharacterCategory.OTHER_LETTER: "Lo",
ECharacterCategory.OTHER_NUMBER: "No",
ECharacterCategory.OTHER_PUNCTUATION: "Po",
ECharacterCategory.OTHER_SYMBOL: "So",
ECharacterCategory.PARAGRAPH_SEPARATOR: "Zp",
ECharacterCategory.PRIVATE_USE: "Co",
ECharacterCategory.SPACE_SEPARATOR: "Zs",
ECharacterCategory.START_PUNCTUATION: "Ps",
ECharacterCategory.SURROGATE: "Cs",
ECharacterCategory.TITLECASE_LETTER: "Lt",
ECharacterCategory.UNASSIGNED: "Cn",
ECharacterCategory.UPPERCASE_LETTER: "Lu",
}
def category(unichr):
return _cat[UCharacter.getType(_get_codepoint(unichr))]
_dir = {
ECharacterDirection.ARABIC_NUMBER: "An",
ECharacterDirection.BLOCK_SEPARATOR: "B",
ECharacterDirection.BOUNDARY_NEUTRAL: "BN",
ECharacterDirection.COMMON_NUMBER_SEPARATOR: "CS",
ECharacterDirection.DIR_NON_SPACING_MARK: "NSM",
ECharacterDirection.EUROPEAN_NUMBER: "EN",
ECharacterDirection.EUROPEAN_NUMBER_SEPARATOR: "ES",
ECharacterDirection.EUROPEAN_NUMBER_TERMINATOR: "ET",
ECharacterDirection.FIRST_STRONG_ISOLATE: "FSI",
ECharacterDirection.LEFT_TO_RIGHT: "L",
ECharacterDirection.LEFT_TO_RIGHT_EMBEDDING: "LRE",
ECharacterDirection.LEFT_TO_RIGHT_ISOLATE: "LRI",
ECharacterDirection.LEFT_TO_RIGHT_OVERRIDE: "LRO",
ECharacterDirection.OTHER_NEUTRAL: "ON",
ECharacterDirection.POP_DIRECTIONAL_FORMAT: "PDF",
ECharacterDirection.POP_DIRECTIONAL_ISOLATE: "PDI",
ECharacterDirection.RIGHT_TO_LEFT: "R",
ECharacterDirection.RIGHT_TO_LEFT_ARABIC: "AL",
ECharacterDirection.RIGHT_TO_LEFT_EMBEDDING: "RLE",
ECharacterDirection.RIGHT_TO_LEFT_ISOLATE: "RLI",
ECharacterDirection.RIGHT_TO_LEFT_OVERRIDE: "RLO",
ECharacterDirection.SEGMENT_SEPARATOR: "S",
ECharacterDirection.WHITE_SPACE_NEUTRAL: "WS"
}
def bidirectional(unichr):
return _dir[UCharacter.getDirection(_get_codepoint(unichr))]
def combining(unichr):
return UCharacter.getCombiningClass(_get_codepoint(unichr))
def mirrored(unichr):
return UCharacter.isMirrored(_get_codepoint(unichr))
_eaw = {
# http://www.unicode.org/reports/tr11/
EastAsianWidth.AMBIGUOUS : "A",
EastAsianWidth.COUNT : "?", # apparently not used, see above TR
EastAsianWidth.FULLWIDTH : "F",
EastAsianWidth.HALFWIDTH : "H",
EastAsianWidth.NARROW : "Na",
EastAsianWidth.NEUTRAL : "N",
EastAsianWidth.WIDE : "W"
}
def east_asian_width(unichr):
return _eaw[UCharacter.getIntPropertyValue(_get_codepoint(unichr), UProperty.EAST_ASIAN_WIDTH)]
def normalize(form, unistr):
"""
Return the normal form 'form' for the Unicode string unistr. Valid
values for form are 'NFC', 'NFKC', 'NFD', and 'NFKD'.
"""
try:
normalizer_form = _forms[form]
except KeyError:
raise ValueError('invalid normalization form')
return Normalizer.normalize(unistr, normalizer_form)
def get_icu_version():
versions = []
for k in VersionInfo.__dict__.iterkeys():
if k.startswith("UNICODE_"):
v = getattr(VersionInfo, k)
versions.append((v.getMajor(), v.getMinor(), v.getMilli()))
return ".".join(str(x) for x in max(versions))
unidata_version = get_icu_version()