"""ttLib/sfnt.py -- low-level module to deal with the sfnt file format. Defines two public classes: SFNTReader SFNTWriter (Normally you don't have to use these classes explicitly; they are used automatically by ttLib.TTFont.) The reading and writing of sfnt files is separated in two distinct classes, since whenever to number of tables changes or whenever a table's length chages you need to rewrite the whole file anyway. """ from io import BytesIO from types import SimpleNamespace from fontTools.misc.py23 import Tag from fontTools.misc import sstruct from fontTools.ttLib import TTLibError import struct from collections import OrderedDict import logging log = logging.getLogger(__name__) class SFNTReader(object): def __new__(cls, *args, **kwargs): """ Return an instance of the SFNTReader sub-class which is compatible with the input file type. """ if args and cls is SFNTReader: infile = args[0] infile.seek(0) sfntVersion = Tag(infile.read(4)) infile.seek(0) if sfntVersion == "wOF2": # return new WOFF2Reader object from fontTools.ttLib.woff2 import WOFF2Reader return object.__new__(WOFF2Reader) # return default object return object.__new__(cls) def __init__(self, file, checkChecksums=0, fontNumber=-1): self.file = file self.checkChecksums = checkChecksums self.flavor = None self.flavorData = None self.DirectoryEntry = SFNTDirectoryEntry self.file.seek(0) self.sfntVersion = self.file.read(4) self.file.seek(0) if self.sfntVersion == b"ttcf": header = readTTCHeader(self.file) numFonts = header.numFonts if not 0 <= fontNumber < numFonts: raise TTLibError("specify a font number between 0 and %d (inclusive)" % (numFonts - 1)) self.numFonts = numFonts self.file.seek(header.offsetTable[fontNumber]) data = self.file.read(sfntDirectorySize) if len(data) != sfntDirectorySize: raise TTLibError("Not a Font Collection (not enough data)") sstruct.unpack(sfntDirectoryFormat, data, self) elif self.sfntVersion == b"wOFF": self.flavor = "woff" self.DirectoryEntry = WOFFDirectoryEntry data = self.file.read(woffDirectorySize) if len(data) != woffDirectorySize: raise TTLibError("Not a WOFF font (not enough data)") sstruct.unpack(woffDirectoryFormat, data, self) else: data = self.file.read(sfntDirectorySize) if len(data) != sfntDirectorySize: raise TTLibError("Not a TrueType or OpenType font (not enough data)") sstruct.unpack(sfntDirectoryFormat, data, self) self.sfntVersion = Tag(self.sfntVersion) if self.sfntVersion not in ("\x00\x01\x00\x00", "OTTO", "true"): raise TTLibError("Not a TrueType or OpenType font (bad sfntVersion)") tables = {} for i in range(self.numTables): entry = self.DirectoryEntry() entry.fromFile(self.file) tag = Tag(entry.tag) tables[tag] = entry self.tables = OrderedDict(sorted(tables.items(), key=lambda i: i[1].offset)) # Load flavor data if any if self.flavor == "woff": self.flavorData = WOFFFlavorData(self) def has_key(self, tag): return tag in self.tables __contains__ = has_key def keys(self): return self.tables.keys() def __getitem__(self, tag): """Fetch the raw table data.""" entry = self.tables[Tag(tag)] data = entry.loadData (self.file) if self.checkChecksums: if tag == 'head': # Beh: we have to special-case the 'head' table. checksum = calcChecksum(data[:8] + b'\0\0\0\0' + data[12:]) else: checksum = calcChecksum(data) if self.checkChecksums > 1: # Be obnoxious, and barf when it's wrong assert checksum == entry.checkSum, "bad checksum for '%s' table" % tag elif checksum != entry.checkSum: # Be friendly, and just log a warning. log.warning("bad checksum for '%s' table", tag) return data def __delitem__(self, tag): del self.tables[Tag(tag)] def close(self): self.file.close() # We define custom __getstate__ and __setstate__ to make SFNTReader pickle-able # and deepcopy-able. When a TTFont is loaded as lazy=True, SFNTReader holds a # reference to an external file object which is not pickleable. So in __getstate__ # we store the file name and current position, and in __setstate__ we reopen the # same named file after unpickling. def __getstate__(self): if isinstance(self.file, BytesIO): # BytesIO is already pickleable, return the state unmodified return self.__dict__ # remove unpickleable file attribute, and only store its name and pos state = self.__dict__.copy() del state["file"] state["_filename"] = self.file.name state["_filepos"] = self.file.tell() return state def __setstate__(self, state): if "file" not in state: self.file = open(state.pop("_filename"), "rb") self.file.seek(state.pop("_filepos")) self.__dict__.update(state) # default compression level for WOFF 1.0 tables and metadata ZLIB_COMPRESSION_LEVEL = 6 # if set to True, use zopfli instead of zlib for compressing WOFF 1.0. # The Python bindings are available at https://pypi.python.org/pypi/zopfli USE_ZOPFLI = False # mapping between zlib's compression levels and zopfli's 'numiterations'. # Use lower values for files over several MB in size or it will be too slow ZOPFLI_LEVELS = { # 0: 0, # can't do 0 iterations... 1: 1, 2: 3, 3: 5, 4: 8, 5: 10, 6: 15, 7: 25, 8: 50, 9: 100, } def compress(data, level=ZLIB_COMPRESSION_LEVEL): """ Compress 'data' to Zlib format. If 'USE_ZOPFLI' variable is True, zopfli is used instead of the zlib module. The compression 'level' must be between 0 and 9. 1 gives best speed, 9 gives best compression (0 gives no compression at all). The default value is a compromise between speed and compression (6). """ if not (0 <= level <= 9): raise ValueError('Bad compression level: %s' % level) if not USE_ZOPFLI or level == 0: from zlib import compress return compress(data, level) else: from zopfli.zlib import compress return compress(data, numiterations=ZOPFLI_LEVELS[level]) class SFNTWriter(object): def __new__(cls, *args, **kwargs): """ Return an instance of the SFNTWriter sub-class which is compatible with the specified 'flavor'. """ flavor = None if kwargs and 'flavor' in kwargs: flavor = kwargs['flavor'] elif args and len(args) > 3: flavor = args[3] if cls is SFNTWriter: if flavor == "woff2": # return new WOFF2Writer object from fontTools.ttLib.woff2 import WOFF2Writer return object.__new__(WOFF2Writer) # return default object return object.__new__(cls) def __init__(self, file, numTables, sfntVersion="\000\001\000\000", flavor=None, flavorData=None): self.file = file self.numTables = numTables self.sfntVersion = Tag(sfntVersion) self.flavor = flavor self.flavorData = flavorData if self.flavor == "woff": self.directoryFormat = woffDirectoryFormat self.directorySize = woffDirectorySize self.DirectoryEntry = WOFFDirectoryEntry self.signature = "wOFF" # to calculate WOFF checksum adjustment, we also need the original SFNT offsets self.origNextTableOffset = sfntDirectorySize + numTables * sfntDirectoryEntrySize else: assert not self.flavor, "Unknown flavor '%s'" % self.flavor self.directoryFormat = sfntDirectoryFormat self.directorySize = sfntDirectorySize self.DirectoryEntry = SFNTDirectoryEntry from fontTools.ttLib import getSearchRange self.searchRange, self.entrySelector, self.rangeShift = getSearchRange(numTables, 16) self.directoryOffset = self.file.tell() self.nextTableOffset = self.directoryOffset + self.directorySize + numTables * self.DirectoryEntry.formatSize # clear out directory area self.file.seek(self.nextTableOffset) # make sure we're actually where we want to be. (old cStringIO bug) self.file.write(b'\0' * (self.nextTableOffset - self.file.tell())) self.tables = OrderedDict() def setEntry(self, tag, entry): if tag in self.tables: raise TTLibError("cannot rewrite '%s' table" % tag) self.tables[tag] = entry def __setitem__(self, tag, data): """Write raw table data to disk.""" if tag in self.tables: raise TTLibError("cannot rewrite '%s' table" % tag) entry = self.DirectoryEntry() entry.tag = tag entry.offset = self.nextTableOffset if tag == 'head': entry.checkSum = calcChecksum(data[:8] + b'\0\0\0\0' + data[12:]) self.headTable = data entry.uncompressed = True else: entry.checkSum = calcChecksum(data) entry.saveData(self.file, data) if self.flavor == "woff": entry.origOffset = self.origNextTableOffset self.origNextTableOffset += (entry.origLength + 3) & ~3 self.nextTableOffset = self.nextTableOffset + ((entry.length + 3) & ~3) # Add NUL bytes to pad the table data to a 4-byte boundary. # Don't depend on f.seek() as we need to add the padding even if no # subsequent write follows (seek is lazy), ie. after the final table # in the font. self.file.write(b'\0' * (self.nextTableOffset - self.file.tell())) assert self.nextTableOffset == self.file.tell() self.setEntry(tag, entry) def __getitem__(self, tag): return self.tables[tag] def close(self): """All tables must have been written to disk. Now write the directory. """ tables = sorted(self.tables.items()) if len(tables) != self.numTables: raise TTLibError("wrong number of tables; expected %d, found %d" % (self.numTables, len(tables))) if self.flavor == "woff": self.signature = b"wOFF" self.reserved = 0 self.totalSfntSize = 12 self.totalSfntSize += 16 * len(tables) for tag, entry in tables: self.totalSfntSize += (entry.origLength + 3) & ~3 data = self.flavorData if self.flavorData else WOFFFlavorData() if data.majorVersion is not None and data.minorVersion is not None: self.majorVersion = data.majorVersion self.minorVersion = data.minorVersion else: if hasattr(self, 'headTable'): self.majorVersion, self.minorVersion = struct.unpack(">HH", self.headTable[4:8]) else: self.majorVersion = self.minorVersion = 0 if data.metaData: self.metaOrigLength = len(data.metaData) self.file.seek(0,2) self.metaOffset = self.file.tell() compressedMetaData = compress(data.metaData) self.metaLength = len(compressedMetaData) self.file.write(compressedMetaData) else: self.metaOffset = self.metaLength = self.metaOrigLength = 0 if data.privData: self.file.seek(0,2) off = self.file.tell() paddedOff = (off + 3) & ~3 self.file.write('\0' * (paddedOff - off)) self.privOffset = self.file.tell() self.privLength = len(data.privData) self.file.write(data.privData) else: self.privOffset = self.privLength = 0 self.file.seek(0,2) self.length = self.file.tell() else: assert not self.flavor, "Unknown flavor '%s'" % self.flavor pass directory = sstruct.pack(self.directoryFormat, self) self.file.seek(self.directoryOffset + self.directorySize) seenHead = 0 for tag, entry in tables: if tag == "head": seenHead = 1 directory = directory + entry.toString() if seenHead: self.writeMasterChecksum(directory) self.file.seek(self.directoryOffset) self.file.write(directory) def _calcMasterChecksum(self, directory): # calculate checkSumAdjustment tags = list(self.tables.keys()) checksums = [] for i in range(len(tags)): checksums.append(self.tables[tags[i]].checkSum) if self.DirectoryEntry != SFNTDirectoryEntry: # Create a SFNT directory for checksum calculation purposes from fontTools.ttLib import getSearchRange self.searchRange, self.entrySelector, self.rangeShift = getSearchRange(self.numTables, 16) directory = sstruct.pack(sfntDirectoryFormat, self) tables = sorted(self.tables.items()) for tag, entry in tables: sfntEntry = SFNTDirectoryEntry() sfntEntry.tag = entry.tag sfntEntry.checkSum = entry.checkSum sfntEntry.offset = entry.origOffset sfntEntry.length = entry.origLength directory = directory + sfntEntry.toString() directory_end = sfntDirectorySize + len(self.tables) * sfntDirectoryEntrySize assert directory_end == len(directory) checksums.append(calcChecksum(directory)) checksum = sum(checksums) & 0xffffffff # BiboAfba! checksumadjustment = (0xB1B0AFBA - checksum) & 0xffffffff return checksumadjustment def writeMasterChecksum(self, directory): checksumadjustment = self._calcMasterChecksum(directory) # write the checksum to the file self.file.seek(self.tables['head'].offset + 8) self.file.write(struct.pack(">L", checksumadjustment)) def reordersTables(self): return False # -- sfnt directory helpers and cruft ttcHeaderFormat = """ > # big endian TTCTag: 4s # "ttcf" Version: L # 0x00010000 or 0x00020000 numFonts: L # number of fonts # OffsetTable[numFonts]: L # array with offsets from beginning of file # ulDsigTag: L # version 2.0 only # ulDsigLength: L # version 2.0 only # ulDsigOffset: L # version 2.0 only """ ttcHeaderSize = sstruct.calcsize(ttcHeaderFormat) sfntDirectoryFormat = """ > # big endian sfntVersion: 4s numTables: H # number of tables searchRange: H # (max2 <= numTables)*16 entrySelector: H # log2(max2 <= numTables) rangeShift: H # numTables*16-searchRange """ sfntDirectorySize = sstruct.calcsize(sfntDirectoryFormat) sfntDirectoryEntryFormat = """ > # big endian tag: 4s checkSum: L offset: L length: L """ sfntDirectoryEntrySize = sstruct.calcsize(sfntDirectoryEntryFormat) woffDirectoryFormat = """ > # big endian signature: 4s # "wOFF" sfntVersion: 4s length: L # total woff file size numTables: H # number of tables reserved: H # set to 0 totalSfntSize: L # uncompressed size majorVersion: H # major version of WOFF file minorVersion: H # minor version of WOFF file metaOffset: L # offset to metadata block metaLength: L # length of compressed metadata metaOrigLength: L # length of uncompressed metadata privOffset: L # offset to private data block privLength: L # length of private data block """ woffDirectorySize = sstruct.calcsize(woffDirectoryFormat) woffDirectoryEntryFormat = """ > # big endian tag: 4s offset: L length: L # compressed length origLength: L # original length checkSum: L # original checksum """ woffDirectoryEntrySize = sstruct.calcsize(woffDirectoryEntryFormat) class DirectoryEntry(object): def __init__(self): self.uncompressed = False # if True, always embed entry raw def fromFile(self, file): sstruct.unpack(self.format, file.read(self.formatSize), self) def fromString(self, str): sstruct.unpack(self.format, str, self) def toString(self): return sstruct.pack(self.format, self) def __repr__(self): if hasattr(self, "tag"): return "<%s '%s' at %x>" % (self.__class__.__name__, self.tag, id(self)) else: return "<%s at %x>" % (self.__class__.__name__, id(self)) def loadData(self, file): file.seek(self.offset) data = file.read(self.length) assert len(data) == self.length if hasattr(self.__class__, 'decodeData'): data = self.decodeData(data) return data def saveData(self, file, data): if hasattr(self.__class__, 'encodeData'): data = self.encodeData(data) self.length = len(data) file.seek(self.offset) file.write(data) def decodeData(self, rawData): return rawData def encodeData(self, data): return data class SFNTDirectoryEntry(DirectoryEntry): format = sfntDirectoryEntryFormat formatSize = sfntDirectoryEntrySize class WOFFDirectoryEntry(DirectoryEntry): format = woffDirectoryEntryFormat formatSize = woffDirectoryEntrySize def __init__(self): super(WOFFDirectoryEntry, self).__init__() # With fonttools<=3.1.2, the only way to set a different zlib # compression level for WOFF directory entries was to set the class # attribute 'zlibCompressionLevel'. This is now replaced by a globally # defined `ZLIB_COMPRESSION_LEVEL`, which is also applied when # compressing the metadata. For backward compatibility, we still # use the class attribute if it was already set. if not hasattr(WOFFDirectoryEntry, 'zlibCompressionLevel'): self.zlibCompressionLevel = ZLIB_COMPRESSION_LEVEL def decodeData(self, rawData): import zlib if self.length == self.origLength: data = rawData else: assert self.length < self.origLength data = zlib.decompress(rawData) assert len(data) == self.origLength return data def encodeData(self, data): self.origLength = len(data) if not self.uncompressed: compressedData = compress(data, self.zlibCompressionLevel) if self.uncompressed or len(compressedData) >= self.origLength: # Encode uncompressed rawData = data self.length = self.origLength else: rawData = compressedData self.length = len(rawData) return rawData class WOFFFlavorData(): Flavor = 'woff' def __init__(self, reader=None): self.majorVersion = None self.minorVersion = None self.metaData = None self.privData = None if reader: self.majorVersion = reader.majorVersion self.minorVersion = reader.minorVersion if reader.metaLength: reader.file.seek(reader.metaOffset) rawData = reader.file.read(reader.metaLength) assert len(rawData) == reader.metaLength data = self._decompress(rawData) assert len(data) == reader.metaOrigLength self.metaData = data if reader.privLength: reader.file.seek(reader.privOffset) data = reader.file.read(reader.privLength) assert len(data) == reader.privLength self.privData = data def _decompress(self, rawData): import zlib return zlib.decompress(rawData) def calcChecksum(data): """Calculate the checksum for an arbitrary block of data. Optionally takes a 'start' argument, which allows you to calculate a checksum in chunks by feeding it a previous result. If the data length is not a multiple of four, it assumes it is to be padded with null byte. >>> print(calcChecksum(b"abcd")) 1633837924 >>> print(calcChecksum(b"abcdxyz")) 3655064932 """ remainder = len(data) % 4 if remainder: data += b"\0" * (4 - remainder) value = 0 blockSize = 4096 assert blockSize % 4 == 0 for i in range(0, len(data), blockSize): block = data[i:i+blockSize] longs = struct.unpack(">%dL" % (len(block) // 4), block) value = (value + sum(longs)) & 0xffffffff return value def readTTCHeader(file): file.seek(0) data = file.read(ttcHeaderSize) if len(data) != ttcHeaderSize: raise TTLibError("Not a Font Collection (not enough data)") self = SimpleNamespace() sstruct.unpack(ttcHeaderFormat, data, self) if self.TTCTag != "ttcf": raise TTLibError("Not a Font Collection") assert self.Version == 0x00010000 or self.Version == 0x00020000, "unrecognized TTC version 0x%08x" % self.Version self.offsetTable = struct.unpack(">%dL" % self.numFonts, file.read(self.numFonts * 4)) if self.Version == 0x00020000: pass # ignoring version 2.0 signatures return self def writeTTCHeader(file, numFonts): self = SimpleNamespace() self.TTCTag = 'ttcf' self.Version = 0x00010000 self.numFonts = numFonts file.seek(0) file.write(sstruct.pack(ttcHeaderFormat, self)) offset = file.tell() file.write(struct.pack(">%dL" % self.numFonts, *([0] * self.numFonts))) return offset if __name__ == "__main__": import sys import doctest sys.exit(doctest.testmod().failed)