git-svn-id: svn://svn.code.sf.net/p/fonttools/code/trunk@26 4cde692c-a291-49d1-8350-778aa11640f8
461 lines
15 KiB
Python
461 lines
15 KiB
Python
"""cffLib.py -- read/write tools for Adobe CFF fonts."""
|
|
|
|
#
|
|
# $Id: cffLib.py,v 1.3 1999-12-20 22:02:10 Just Exp $"
|
|
#
|
|
|
|
import struct, sstruct
|
|
import string
|
|
import types
|
|
import psCharStrings
|
|
|
|
|
|
cffHeaderFormat = """
|
|
major: B
|
|
minor: B
|
|
hdrSize: B
|
|
offSize: B
|
|
"""
|
|
|
|
class CFFFontSet:
|
|
|
|
def __init__(self):
|
|
self.fonts = {}
|
|
|
|
def decompile(self, data):
|
|
sstruct.unpack(cffHeaderFormat, data[:4], self)
|
|
assert self.major == 1 and self.minor == 0, \
|
|
"unknown CFF format: %d.%d" % (self.major, self.minor)
|
|
restdata = data[self.hdrSize:]
|
|
|
|
self.fontNames, restdata = readINDEX(restdata)
|
|
topDicts, restdata = readINDEX(restdata)
|
|
strings, restdata = readINDEX(restdata)
|
|
strings = IndexedStrings(strings)
|
|
globalSubrs, restdata = readINDEX(restdata)
|
|
self.GlobalSubrs = map(psCharStrings.T2CharString, globalSubrs)
|
|
|
|
for i in range(len(topDicts)):
|
|
font = self.fonts[self.fontNames[i]] = CFFFont()
|
|
font.GlobalSubrs = self.GlobalSubrs # Hmm.
|
|
font.decompile(data, topDicts[i], strings, self) # maybe only 'on demand'?
|
|
|
|
|
|
def compile(self):
|
|
strings = IndexedStrings()
|
|
XXXX
|
|
|
|
def toXML(self, xmlWriter, progress=None):
|
|
xmlWriter.newline()
|
|
for fontName in self.fontNames:
|
|
xmlWriter.begintag("CFFFont", name=fontName)
|
|
xmlWriter.newline()
|
|
font = self.fonts[fontName]
|
|
font.toXML(xmlWriter, progress)
|
|
xmlWriter.endtag("CFFFont")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
xmlWriter.begintag("GlobalSubrs")
|
|
xmlWriter.newline()
|
|
for i in range(len(self.GlobalSubrs)):
|
|
xmlWriter.newline()
|
|
xmlWriter.begintag("CharString", id=i)
|
|
xmlWriter.newline()
|
|
self.GlobalSubrs[i].toXML(xmlWriter)
|
|
xmlWriter.endtag("CharString")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
xmlWriter.endtag("GlobalSubrs")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
|
|
def fromXML(self, (name, attrs, content)):
|
|
xxx
|
|
|
|
|
|
class IndexedStrings:
|
|
|
|
def __init__(self, strings=None):
|
|
if strings is None:
|
|
strings = []
|
|
self.strings = strings
|
|
|
|
def __getitem__(self, SID):
|
|
if SID < cffStandardStringCount:
|
|
return cffStandardStrings[SID]
|
|
else:
|
|
return self.strings[SID - cffStandardStringCount]
|
|
|
|
def getSID(self, s):
|
|
if not hasattr(self, "stringMapping"):
|
|
self.buildStringMapping()
|
|
if cffStandardStringMapping.has_key(s):
|
|
SID = cffStandardStringMapping[s]
|
|
if self.stringMapping.has_key(s):
|
|
SID = self.stringMapping[s]
|
|
else:
|
|
SID = len(self.strings) + cffStandardStringCount
|
|
self.strings.append(s)
|
|
self.stringMapping[s] = SID
|
|
return SID
|
|
|
|
def getStrings(self):
|
|
return self.strings
|
|
|
|
def buildStringMapping(self):
|
|
self.stringMapping = {}
|
|
for index in range(len(self.strings)):
|
|
self.stringMapping[self.strings[index]] = index + cffStandardStringCount
|
|
|
|
|
|
class CFFFont:
|
|
|
|
defaults = psCharStrings.topDictDefaults
|
|
|
|
def __init__(self):
|
|
pass
|
|
|
|
def __getattr__(self, attr):
|
|
if not self.defaults.has_key(attr):
|
|
raise AttributeError, attr
|
|
return self.defaults[attr]
|
|
|
|
def fromDict(self, dict):
|
|
self.__dict__.update(dict)
|
|
|
|
def decompile(self, data, topDictData, strings, fontSet):
|
|
top = psCharStrings.TopDictDecompiler(strings)
|
|
top.decompile(topDictData)
|
|
self.fromDict(top.getDict())
|
|
|
|
# get private dict
|
|
size, offset = self.Private
|
|
#print "YYY Private (size, offset):", size, offset
|
|
privateData = data[offset:offset+size]
|
|
self.Private = PrivateDict()
|
|
self.Private.decompile(data[offset:], privateData, strings)
|
|
|
|
# get raw charstrings
|
|
#print "YYYY CharStrings offset:", self.CharStrings
|
|
rawCharStrings, restdata = readINDEX(data[self.CharStrings:])
|
|
nGlyphs = len(rawCharStrings)
|
|
|
|
# get charset (or rather: get glyphNames)
|
|
charsetOffset = self.charset
|
|
if charsetOffset == 0:
|
|
xxx # standard charset
|
|
else:
|
|
#print "YYYYY charsetOffset:", charsetOffset
|
|
format = ord(data[charsetOffset])
|
|
if format == 0:
|
|
xxx
|
|
elif format == 1:
|
|
charSet = parseCharsetFormat1(nGlyphs,
|
|
data[charsetOffset+1:], strings)
|
|
elif format == 2:
|
|
charSet = parseCharsetFormat2(nGlyphs,
|
|
data[charsetOffset+1:], strings)
|
|
elif format == 3:
|
|
xxx
|
|
else:
|
|
xxx
|
|
self.charset = charSet
|
|
|
|
assert len(charSet) == nGlyphs
|
|
self.CharStrings = charStrings = {}
|
|
if self.CharstringType == 2:
|
|
# Type 2 CharStrings
|
|
charStringClass = psCharStrings.T2CharString
|
|
else:
|
|
# Type 1 CharStrings
|
|
charStringClass = psCharStrings.T1CharString
|
|
for i in range(nGlyphs):
|
|
charStrings[charSet[i]] = charStringClass(rawCharStrings[i])
|
|
assert len(charStrings) == nGlyphs
|
|
|
|
# XXX Encoding!
|
|
encoding = self.Encoding
|
|
if encoding not in (0, 1):
|
|
# encoding is an _offset_ from the beginning of 'data' to an encoding subtable
|
|
XXX
|
|
self.Encoding = encoding
|
|
|
|
def getGlyphOrder(self):
|
|
return self.charset
|
|
|
|
def setGlyphOrder(self, glyphOrder):
|
|
self.charset = glyphOrder
|
|
|
|
def decompileAllCharStrings(self):
|
|
if self.CharstringType == 2:
|
|
# Type 2 CharStrings
|
|
decompiler = psCharStrings.SimpleT2Decompiler(self.Private.Subrs, self.GlobalSubrs)
|
|
for charString in self.CharStrings.values():
|
|
if charString.needsDecompilation():
|
|
decompiler.reset()
|
|
decompiler.execute(charString)
|
|
else:
|
|
# Type 1 CharStrings
|
|
for charString in self.CharStrings.values():
|
|
charString.decompile()
|
|
|
|
def toXML(self, xmlWriter, progress=None):
|
|
xmlWriter.newline()
|
|
# first dump the simple values
|
|
self.toXMLSimpleValues(xmlWriter)
|
|
|
|
# dump charset
|
|
# XXX
|
|
|
|
# decompile all charstrings
|
|
if progress:
|
|
progress.setlabel("Decompiling CharStrings...")
|
|
self.decompileAllCharStrings()
|
|
|
|
# dump private dict
|
|
xmlWriter.begintag("Private")
|
|
xmlWriter.newline()
|
|
self.Private.toXML(xmlWriter)
|
|
xmlWriter.endtag("Private")
|
|
xmlWriter.newline()
|
|
|
|
self.toXMLCharStrings(xmlWriter, progress)
|
|
|
|
def toXMLSimpleValues(self, xmlWriter):
|
|
keys = self.__dict__.keys()
|
|
keys.remove("CharStrings")
|
|
keys.remove("Private")
|
|
keys.remove("charset")
|
|
keys.remove("GlobalSubrs")
|
|
keys.sort()
|
|
for key in keys:
|
|
value = getattr(self, key)
|
|
if key == "Encoding":
|
|
if value == 0:
|
|
# encoding is (Adobe) Standard Encoding
|
|
value = "StandardEncoding"
|
|
elif value == 1:
|
|
# encoding is Expert Encoding
|
|
value = "ExpertEncoding"
|
|
if type(value) == types.ListType:
|
|
value = string.join(map(str, value), " ")
|
|
else:
|
|
value = str(value)
|
|
xmlWriter.begintag(key)
|
|
if hasattr(value, "toXML"):
|
|
xmlWriter.newline()
|
|
value.toXML(xmlWriter)
|
|
xmlWriter.newline()
|
|
else:
|
|
xmlWriter.write(value)
|
|
xmlWriter.endtag(key)
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
|
|
def toXMLCharStrings(self, xmlWriter, progress=None):
|
|
charStrings = self.CharStrings
|
|
xmlWriter.newline()
|
|
xmlWriter.begintag("CharStrings")
|
|
xmlWriter.newline()
|
|
glyphNames = charStrings.keys()
|
|
glyphNames.sort()
|
|
for glyphName in glyphNames:
|
|
if progress:
|
|
progress.setlabel("Dumping 'CFF ' table... (%s)" % glyphName)
|
|
progress.increment()
|
|
xmlWriter.newline()
|
|
charString = charStrings[glyphName]
|
|
xmlWriter.begintag("CharString", name=glyphName)
|
|
xmlWriter.newline()
|
|
charString.toXML(xmlWriter)
|
|
xmlWriter.endtag("CharString")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
xmlWriter.endtag("CharStrings")
|
|
xmlWriter.newline()
|
|
|
|
|
|
class PrivateDict:
|
|
|
|
defaults = psCharStrings.privateDictDefaults
|
|
|
|
def __init__(self):
|
|
pass
|
|
|
|
def decompile(self, data, privateData, strings):
|
|
p = psCharStrings.PrivateDictDecompiler(strings)
|
|
p.decompile(privateData)
|
|
self.fromDict(p.getDict())
|
|
|
|
# get local subrs
|
|
#print "YYY Private.Subrs:", self.Subrs
|
|
chunk = data[self.Subrs:]
|
|
localSubrs, restdata = readINDEX(chunk)
|
|
self.Subrs = map(psCharStrings.T2CharString, localSubrs)
|
|
|
|
def toXML(self, xmlWriter):
|
|
xmlWriter.newline()
|
|
keys = self.__dict__.keys()
|
|
keys.remove("Subrs")
|
|
for key in keys:
|
|
value = getattr(self, key)
|
|
if type(value) == types.ListType:
|
|
value = string.join(map(str, value), " ")
|
|
else:
|
|
value = str(value)
|
|
xmlWriter.begintag(key)
|
|
xmlWriter.write(value)
|
|
xmlWriter.endtag(key)
|
|
xmlWriter.newline()
|
|
# write subroutines
|
|
xmlWriter.newline()
|
|
xmlWriter.begintag("Subrs")
|
|
xmlWriter.newline()
|
|
for i in range(len(self.Subrs)):
|
|
xmlWriter.newline()
|
|
xmlWriter.begintag("CharString", id=i)
|
|
xmlWriter.newline()
|
|
self.Subrs[i].toXML(xmlWriter)
|
|
xmlWriter.endtag("CharString")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
xmlWriter.endtag("Subrs")
|
|
xmlWriter.newline()
|
|
xmlWriter.newline()
|
|
|
|
def __getattr__(self, attr):
|
|
if not self.defaults.has_key(attr):
|
|
raise AttributeError, attr
|
|
return self.defaults[attr]
|
|
|
|
def fromDict(self, dict):
|
|
self.__dict__.update(dict)
|
|
|
|
|
|
def readINDEX(data):
|
|
count, = struct.unpack(">H", data[:2])
|
|
count = int(count)
|
|
offSize = ord(data[2])
|
|
data = data[3:]
|
|
offsets = []
|
|
for index in range(count+1):
|
|
chunk = data[index * offSize: (index+1) * offSize]
|
|
chunk = '\0' * (4 - offSize) + chunk
|
|
offset, = struct.unpack(">L", chunk)
|
|
offset = int(offset)
|
|
offsets.append(offset)
|
|
data = data[(count+1) * offSize:]
|
|
prev = offsets[0]
|
|
stuff = []
|
|
for next in offsets[1:]:
|
|
chunk = data[prev-1:next-1]
|
|
assert len(chunk) == next - prev
|
|
stuff.append(chunk)
|
|
prev = next
|
|
data = data[next-1:]
|
|
return stuff, data
|
|
|
|
|
|
def parseCharsetFormat1(nGlyphs, data, strings):
|
|
charSet = ['.notdef']
|
|
count = 1
|
|
while count < nGlyphs:
|
|
first = int(struct.unpack(">H", data[:2])[0])
|
|
nLeft = ord(data[2])
|
|
data = data[3:]
|
|
for SID in range(first, first+nLeft+1):
|
|
charSet.append(strings[SID])
|
|
count = count + nLeft + 1
|
|
return charSet
|
|
|
|
|
|
def parseCharsetFormat2(nGlyphs, data, strings):
|
|
charSet = ['.notdef']
|
|
count = 1
|
|
while count < nGlyphs:
|
|
first = int(struct.unpack(">H", data[:2])[0])
|
|
nLeft = int(struct.unpack(">H", data[2:4])[0])
|
|
data = data[4:]
|
|
for SID in range(first, first+nLeft+1):
|
|
charSet.append(strings[SID])
|
|
count = count + nLeft + 1
|
|
return charSet
|
|
|
|
|
|
# The 391 Standard Strings as used in the CFF format.
|
|
# from Adobe Technical None #5176, version 1.0, 18 March 1998
|
|
|
|
cffStandardStrings = ['.notdef', 'space', 'exclam', 'quotedbl', 'numbersign',
|
|
'dollar', 'percent', 'ampersand', 'quoteright', 'parenleft', 'parenright',
|
|
'asterisk', 'plus', 'comma', 'hyphen', 'period', 'slash', 'zero', 'one',
|
|
'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine', 'colon',
|
|
'semicolon', 'less', 'equal', 'greater', 'question', 'at', 'A', 'B', 'C',
|
|
'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R',
|
|
'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'bracketleft', 'backslash',
|
|
'bracketright', 'asciicircum', 'underscore', 'quoteleft', 'a', 'b', 'c',
|
|
'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r',
|
|
's', 't', 'u', 'v', 'w', 'x', 'y', 'z', 'braceleft', 'bar', 'braceright',
|
|
'asciitilde', 'exclamdown', 'cent', 'sterling', 'fraction', 'yen', 'florin',
|
|
'section', 'currency', 'quotesingle', 'quotedblleft', 'guillemotleft',
|
|
'guilsinglleft', 'guilsinglright', 'fi', 'fl', 'endash', 'dagger',
|
|
'daggerdbl', 'periodcentered', 'paragraph', 'bullet', 'quotesinglbase',
|
|
'quotedblbase', 'quotedblright', 'guillemotright', 'ellipsis', 'perthousand',
|
|
'questiondown', 'grave', 'acute', 'circumflex', 'tilde', 'macron', 'breve',
|
|
'dotaccent', 'dieresis', 'ring', 'cedilla', 'hungarumlaut', 'ogonek', 'caron',
|
|
'emdash', 'AE', 'ordfeminine', 'Lslash', 'Oslash', 'OE', 'ordmasculine', 'ae',
|
|
'dotlessi', 'lslash', 'oslash', 'oe', 'germandbls', 'onesuperior',
|
|
'logicalnot', 'mu', 'trademark', 'Eth', 'onehalf', 'plusminus', 'Thorn',
|
|
'onequarter', 'divide', 'brokenbar', 'degree', 'thorn', 'threequarters',
|
|
'twosuperior', 'registered', 'minus', 'eth', 'multiply', 'threesuperior',
|
|
'copyright', 'Aacute', 'Acircumflex', 'Adieresis', 'Agrave', 'Aring',
|
|
'Atilde', 'Ccedilla', 'Eacute', 'Ecircumflex', 'Edieresis', 'Egrave',
|
|
'Iacute', 'Icircumflex', 'Idieresis', 'Igrave', 'Ntilde', 'Oacute',
|
|
'Ocircumflex', 'Odieresis', 'Ograve', 'Otilde', 'Scaron', 'Uacute',
|
|
'Ucircumflex', 'Udieresis', 'Ugrave', 'Yacute', 'Ydieresis', 'Zcaron',
|
|
'aacute', 'acircumflex', 'adieresis', 'agrave', 'aring', 'atilde', 'ccedilla',
|
|
'eacute', 'ecircumflex', 'edieresis', 'egrave', 'iacute', 'icircumflex',
|
|
'idieresis', 'igrave', 'ntilde', 'oacute', 'ocircumflex', 'odieresis',
|
|
'ograve', 'otilde', 'scaron', 'uacute', 'ucircumflex', 'udieresis', 'ugrave',
|
|
'yacute', 'ydieresis', 'zcaron', 'exclamsmall', 'Hungarumlautsmall',
|
|
'dollaroldstyle', 'dollarsuperior', 'ampersandsmall', 'Acutesmall',
|
|
'parenleftsuperior', 'parenrightsuperior', 'twodotenleader', 'onedotenleader',
|
|
'zerooldstyle', 'oneoldstyle', 'twooldstyle', 'threeoldstyle', 'fouroldstyle',
|
|
'fiveoldstyle', 'sixoldstyle', 'sevenoldstyle', 'eightoldstyle',
|
|
'nineoldstyle', 'commasuperior', 'threequartersemdash', 'periodsuperior',
|
|
'questionsmall', 'asuperior', 'bsuperior', 'centsuperior', 'dsuperior',
|
|
'esuperior', 'isuperior', 'lsuperior', 'msuperior', 'nsuperior', 'osuperior',
|
|
'rsuperior', 'ssuperior', 'tsuperior', 'ff', 'ffi', 'ffl', 'parenleftinferior',
|
|
'parenrightinferior', 'Circumflexsmall', 'hyphensuperior', 'Gravesmall',
|
|
'Asmall', 'Bsmall', 'Csmall', 'Dsmall', 'Esmall', 'Fsmall', 'Gsmall', 'Hsmall',
|
|
'Ismall', 'Jsmall', 'Ksmall', 'Lsmall', 'Msmall', 'Nsmall', 'Osmall', 'Psmall',
|
|
'Qsmall', 'Rsmall', 'Ssmall', 'Tsmall', 'Usmall', 'Vsmall', 'Wsmall', 'Xsmall',
|
|
'Ysmall', 'Zsmall', 'colonmonetary', 'onefitted', 'rupiah', 'Tildesmall',
|
|
'exclamdownsmall', 'centoldstyle', 'Lslashsmall', 'Scaronsmall', 'Zcaronsmall',
|
|
'Dieresissmall', 'Brevesmall', 'Caronsmall', 'Dotaccentsmall', 'Macronsmall',
|
|
'figuredash', 'hypheninferior', 'Ogoneksmall', 'Ringsmall', 'Cedillasmall',
|
|
'questiondownsmall', 'oneeighth', 'threeeighths', 'fiveeighths', 'seveneighths',
|
|
'onethird', 'twothirds', 'zerosuperior', 'foursuperior', 'fivesuperior',
|
|
'sixsuperior', 'sevensuperior', 'eightsuperior', 'ninesuperior', 'zeroinferior',
|
|
'oneinferior', 'twoinferior', 'threeinferior', 'fourinferior', 'fiveinferior',
|
|
'sixinferior', 'seveninferior', 'eightinferior', 'nineinferior', 'centinferior',
|
|
'dollarinferior', 'periodinferior', 'commainferior', 'Agravesmall',
|
|
'Aacutesmall', 'Acircumflexsmall', 'Atildesmall', 'Adieresissmall', 'Aringsmall',
|
|
'AEsmall', 'Ccedillasmall', 'Egravesmall', 'Eacutesmall', 'Ecircumflexsmall',
|
|
'Edieresissmall', 'Igravesmall', 'Iacutesmall', 'Icircumflexsmall',
|
|
'Idieresissmall', 'Ethsmall', 'Ntildesmall', 'Ogravesmall', 'Oacutesmall',
|
|
'Ocircumflexsmall', 'Otildesmall', 'Odieresissmall', 'OEsmall', 'Oslashsmall',
|
|
'Ugravesmall', 'Uacutesmall', 'Ucircumflexsmall', 'Udieresissmall',
|
|
'Yacutesmall', 'Thornsmall', 'Ydieresissmall', '001.000', '001.001', '001.002',
|
|
'001.003', 'Black', 'Bold', 'Book', 'Light', 'Medium', 'Regular', 'Roman',
|
|
'Semibold'
|
|
]
|
|
|
|
cffStandardStringCount = 391
|
|
assert len(cffStandardStrings) == cffStandardStringCount
|
|
# build reverse mapping
|
|
cffStandardStringMapping = {}
|
|
for _i in range(cffStandardStringCount):
|
|
cffStandardStringMapping[cffStandardStrings[_i]] = _i
|
|
|
|
|