2021-04-16 23:21:06 +08:00
|
|
|
# ----------------------------------------------------------------------------
|
|
|
|
# pyglet
|
|
|
|
# Copyright (c) 2006-2008 Alex Holkner
|
2022-04-30 13:56:57 +08:00
|
|
|
# Copyright (c) 2008-2022 pyglet contributors
|
2021-04-16 23:21:06 +08:00
|
|
|
# All rights reserved.
|
|
|
|
#
|
|
|
|
# Redistribution and use in source and binary forms, with or without
|
|
|
|
# modification, are permitted provided that the following conditions
|
|
|
|
# are met:
|
|
|
|
#
|
|
|
|
# * Redistributions of source code must retain the above copyright
|
|
|
|
# notice, this list of conditions and the following disclaimer.
|
|
|
|
# * Redistributions in binary form must reproduce the above copyright
|
|
|
|
# notice, this list of conditions and the following disclaimer in
|
|
|
|
# the documentation and/or other materials provided with the
|
|
|
|
# distribution.
|
|
|
|
# * Neither the name of pyglet nor the names of its
|
|
|
|
# contributors may be used to endorse or promote products
|
|
|
|
# derived from this software without specific prior written
|
|
|
|
# permission.
|
|
|
|
#
|
|
|
|
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
|
|
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
|
|
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
|
|
# FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
|
|
# COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
|
|
# INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
|
|
# BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
|
|
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
|
|
|
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
|
|
|
# ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
|
|
# POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
# ----------------------------------------------------------------------------
|
|
|
|
|
|
|
|
"""
|
|
|
|
Implementation of the Truetype file format.
|
|
|
|
|
|
|
|
Typical applications will not need to use this module directly; look at
|
|
|
|
`pyglet.font` instead.
|
|
|
|
|
|
|
|
References:
|
|
|
|
* http://developer.apple.com/fonts/TTRefMan/RM06
|
|
|
|
* http://www.microsoft.com/typography/otspec
|
|
|
|
"""
|
|
|
|
|
|
|
|
import os
|
|
|
|
import mmap
|
|
|
|
import struct
|
|
|
|
import codecs
|
|
|
|
|
|
|
|
|
|
|
|
class TruetypeInfo:
|
|
|
|
"""Information about a single Truetype face.
|
|
|
|
|
|
|
|
The class memory-maps the font file to read the tables, so
|
|
|
|
it is vital that you call the `close` method to avoid large memory
|
|
|
|
leaks. Once closed, you cannot call any of the ``get_*`` methods.
|
|
|
|
|
|
|
|
Not all tables have been implemented yet (or likely ever will).
|
|
|
|
Currently only the name and metric tables are read; in particular
|
|
|
|
there is no glyph or hinting information.
|
|
|
|
"""
|
|
|
|
|
|
|
|
_name_id_lookup = {
|
|
|
|
'copyright': 0,
|
|
|
|
'family': 1,
|
|
|
|
'subfamily': 2,
|
|
|
|
'identifier': 3,
|
|
|
|
'name': 4,
|
|
|
|
'version': 5,
|
|
|
|
'postscript': 6,
|
|
|
|
'trademark': 7,
|
|
|
|
'manufacturer': 8,
|
|
|
|
'designer': 9,
|
|
|
|
'description': 10,
|
|
|
|
'vendor-url': 11,
|
|
|
|
'designer-url': 12,
|
|
|
|
'license': 13,
|
|
|
|
'license-url': 14,
|
|
|
|
'preferred-family': 16,
|
|
|
|
'preferred-subfamily': 17,
|
|
|
|
'compatible-name': 18,
|
|
|
|
'sample': 19,
|
|
|
|
}
|
|
|
|
|
|
|
|
_platform_id_lookup = {
|
|
|
|
'unicode': 0,
|
|
|
|
'macintosh': 1,
|
|
|
|
'iso': 2,
|
|
|
|
'microsoft': 3,
|
|
|
|
'custom': 4
|
|
|
|
}
|
|
|
|
|
|
|
|
_microsoft_encoding_lookup = {
|
|
|
|
1: 'utf_16_be',
|
|
|
|
2: 'shift_jis',
|
|
|
|
4: 'big5',
|
|
|
|
6: 'johab',
|
|
|
|
10: 'utf_16_be'
|
|
|
|
}
|
|
|
|
|
|
|
|
_macintosh_encoding_lookup = {
|
|
|
|
0: 'mac_roman'
|
|
|
|
}
|
|
|
|
|
|
|
|
def __init__(self, filename):
|
|
|
|
"""Read the given TrueType file.
|
|
|
|
|
|
|
|
:Parameters:
|
|
|
|
`filename`
|
|
|
|
The name of any Windows, OS2 or Macintosh Truetype file.
|
|
|
|
|
|
|
|
The object must be closed (see `close`) after use.
|
|
|
|
|
|
|
|
An exception will be raised if the file does not exist or cannot
|
|
|
|
be read.
|
|
|
|
"""
|
|
|
|
assert filename, "must provide a font file name"
|
|
|
|
length = os.stat(filename).st_size
|
|
|
|
self._fileno = os.open(filename, os.O_RDONLY)
|
|
|
|
if hasattr(mmap, 'MAP_SHARED'):
|
|
|
|
self._data = mmap.mmap(self._fileno, length, mmap.MAP_SHARED, mmap.PROT_READ)
|
|
|
|
else:
|
|
|
|
self._data = mmap.mmap(self._fileno, length, None, mmap.ACCESS_READ)
|
|
|
|
|
|
|
|
self._closed = False
|
|
|
|
|
|
|
|
offsets = _read_offset_table(self._data, 0)
|
|
|
|
self._tables = {}
|
|
|
|
for table in _read_table_directory_entry.array(self._data, offsets.size, offsets.num_tables):
|
|
|
|
self._tables[table.tag] = table
|
|
|
|
|
|
|
|
self._names = None
|
|
|
|
self._horizontal_metrics = None
|
|
|
|
self._character_advances = None
|
|
|
|
self._character_kernings = None
|
|
|
|
self._glyph_kernings = None
|
|
|
|
self._character_map = None
|
|
|
|
self._glyph_map = None
|
|
|
|
self._font_selection_flags = None
|
|
|
|
|
|
|
|
self.header = _read_head_table(self._data, self._tables['head'].offset)
|
|
|
|
self.horizontal_header = _read_horizontal_header(self._data, self._tables['hhea'].offset)
|
|
|
|
|
|
|
|
def get_font_selection_flags(self):
|
|
|
|
"""Return the font selection flags, as defined in OS/2 table"""
|
|
|
|
if not self._font_selection_flags:
|
|
|
|
OS2_table = _read_OS2_table(self._data, self._tables['OS/2'].offset)
|
|
|
|
self._font_selection_flags = OS2_table.fs_selection
|
|
|
|
return self._font_selection_flags
|
|
|
|
|
|
|
|
def is_bold(self):
|
|
|
|
"""Returns True iff the font describes itself as bold."""
|
|
|
|
return bool(self.get_font_selection_flags() & 0x20)
|
|
|
|
|
|
|
|
def is_italic(self):
|
|
|
|
"""Returns True iff the font describes itself as italic."""
|
|
|
|
return bool(self.get_font_selection_flags() & 0x1)
|
|
|
|
|
|
|
|
def get_names(self):
|
|
|
|
"""Returns a dictionary of names defined in the file.
|
|
|
|
|
|
|
|
The key of each item is a tuple of ``platform_id``, ``name_id``,
|
|
|
|
where each ID is the number as described in the Truetype format.
|
|
|
|
|
|
|
|
The value of each item is a tuple of
|
|
|
|
``encoding_id``, ``language_id``, ``value``, where ``value`` is
|
|
|
|
an encoded string.
|
|
|
|
"""
|
|
|
|
if self._names:
|
|
|
|
return self._names
|
|
|
|
naming_table = _read_naming_table(self._data, self._tables['name'].offset)
|
|
|
|
name_records = _read_name_record.array(
|
|
|
|
self._data, self._tables['name'].offset + naming_table.size, naming_table.count)
|
|
|
|
storage = naming_table.string_offset + self._tables['name'].offset
|
|
|
|
self._names = {}
|
|
|
|
for record in name_records:
|
|
|
|
value = self._data[record.offset + storage: record.offset + storage + record.length]
|
|
|
|
key = record.platform_id, record.name_id
|
|
|
|
value = (record.encoding_id, record.language_id, value)
|
|
|
|
if key not in self._names:
|
|
|
|
self._names[key] = []
|
|
|
|
self._names[key].append(value)
|
|
|
|
return self._names
|
|
|
|
|
|
|
|
def get_name(self, name, platform=None, languages=None):
|
|
|
|
"""Returns the value of the given name in this font.
|
|
|
|
|
|
|
|
:Parameters:
|
|
|
|
`name`
|
|
|
|
Either an integer, representing the name_id desired (see
|
|
|
|
font format); or a string describing it, see below for
|
|
|
|
valid names.
|
|
|
|
`platform`
|
|
|
|
Platform for the requested name. Can be the integer ID,
|
|
|
|
or a string describing it. By default, the Microsoft
|
|
|
|
platform is searched first, then Macintosh.
|
|
|
|
`languages`
|
|
|
|
A list of language IDs to search. The first language
|
|
|
|
which defines the requested name will be used. By default,
|
|
|
|
all English dialects are searched.
|
|
|
|
|
|
|
|
If the name is not found, ``None`` is returned. If the name
|
|
|
|
is found, the value will be decoded and returned as a unicode
|
|
|
|
string. Currently only some common encodings are supported.
|
|
|
|
|
|
|
|
Valid names to request are (supply as a string)::
|
|
|
|
|
|
|
|
'copyright'
|
|
|
|
'family'
|
|
|
|
'subfamily'
|
|
|
|
'identifier'
|
|
|
|
'name'
|
|
|
|
'version'
|
|
|
|
'postscript'
|
|
|
|
'trademark'
|
|
|
|
'manufacturer'
|
|
|
|
'designer'
|
|
|
|
'description'
|
|
|
|
'vendor-url'
|
|
|
|
'designer-url'
|
|
|
|
'license'
|
|
|
|
'license-url'
|
|
|
|
'preferred-family'
|
|
|
|
'preferred-subfamily'
|
|
|
|
'compatible-name'
|
|
|
|
'sample'
|
|
|
|
|
|
|
|
Valid platforms to request are (supply as a string)::
|
|
|
|
|
|
|
|
'unicode'
|
|
|
|
'macintosh'
|
|
|
|
'iso'
|
|
|
|
'microsoft'
|
|
|
|
'custom'
|
|
|
|
"""
|
|
|
|
|
|
|
|
names = self.get_names()
|
|
|
|
if type(name) == str:
|
|
|
|
name = self._name_id_lookup[name]
|
|
|
|
if not platform:
|
|
|
|
for platform in ('microsoft', 'macintosh'):
|
|
|
|
value = self.get_name(name, platform, languages)
|
|
|
|
if value:
|
|
|
|
return value
|
|
|
|
if type(platform) == str:
|
|
|
|
platform = self._platform_id_lookup[platform]
|
|
|
|
if not (platform, name) in names:
|
|
|
|
return None
|
|
|
|
|
|
|
|
if platform == 3: # setup for microsoft
|
|
|
|
encodings = self._microsoft_encoding_lookup
|
|
|
|
if not languages:
|
|
|
|
# Default to english languages for microsoft
|
|
|
|
languages = (0x409, 0x809, 0xc09, 0x1009, 0x1409, 0x1809)
|
|
|
|
elif platform == 1: # setup for macintosh
|
|
|
|
encodings = self.__macintosh_encoding_lookup
|
|
|
|
if not languages:
|
|
|
|
# Default to english for macintosh
|
|
|
|
languages = (0,)
|
|
|
|
|
|
|
|
for record in names[(platform, name)]:
|
|
|
|
if record[1] in languages and record[0] in encodings:
|
|
|
|
decoder = codecs.getdecoder(encodings[record[0]])
|
|
|
|
return decoder(record[2])[0]
|
|
|
|
return None
|
|
|
|
|
|
|
|
def get_horizontal_metrics(self):
|
|
|
|
"""Return all horizontal metric entries in table format."""
|
|
|
|
if not self._horizontal_metrics:
|
|
|
|
ar = _read_long_hor_metric.array(self._data,
|
|
|
|
self._tables['hmtx'].offset,
|
|
|
|
self.horizontal_header.number_of_h_metrics)
|
|
|
|
self._horizontal_metrics = ar
|
|
|
|
return self._horizontal_metrics
|
|
|
|
|
|
|
|
def get_character_advances(self):
|
|
|
|
"""Return a dictionary of character->advance.
|
|
|
|
|
|
|
|
They key of the dictionary is a unit-length unicode string,
|
|
|
|
and the value is a float giving the horizontal advance in
|
|
|
|
em.
|
|
|
|
"""
|
|
|
|
if self._character_advances:
|
|
|
|
return self._character_advances
|
|
|
|
ga = self.get_glyph_advances()
|
|
|
|
gmap = self.get_glyph_map()
|
|
|
|
self._character_advances = {}
|
|
|
|
for i in range(len(ga)):
|
|
|
|
if i in gmap and not gmap[i] in self._character_advances:
|
|
|
|
self._character_advances[gmap[i]] = ga[i]
|
|
|
|
return self._character_advances
|
|
|
|
|
|
|
|
def get_glyph_advances(self):
|
|
|
|
"""Return a dictionary of glyph->advance.
|
|
|
|
|
|
|
|
They key of the dictionary is the glyph index and the value is a float
|
|
|
|
giving the horizontal advance in em.
|
|
|
|
"""
|
|
|
|
hm = self.get_horizontal_metrics()
|
|
|
|
return [float(m.advance_width) / self.header.units_per_em for m in hm]
|
|
|
|
|
|
|
|
def get_character_kernings(self):
|
|
|
|
"""Return a dictionary of (left,right)->kerning
|
|
|
|
|
|
|
|
The key of the dictionary is a tuple of ``(left, right)``
|
|
|
|
where each element is a unit-length unicode string. The
|
|
|
|
value of the dictionary is the horizontal pairwise kerning
|
|
|
|
in em.
|
|
|
|
"""
|
|
|
|
if not self._character_kernings:
|
|
|
|
gmap = self.get_glyph_map()
|
|
|
|
kerns = self.get_glyph_kernings()
|
|
|
|
self._character_kernings = {}
|
|
|
|
for pair, value in kerns.items():
|
|
|
|
lglyph, rglyph = pair
|
|
|
|
lchar = lglyph in gmap and gmap[lglyph] or None
|
|
|
|
rchar = rglyph in gmap and gmap[rglyph] or None
|
|
|
|
if lchar and rchar:
|
|
|
|
self._character_kernings[(lchar, rchar)] = value
|
|
|
|
return self._character_kernings
|
|
|
|
|
|
|
|
def get_glyph_kernings(self):
|
|
|
|
"""Return a dictionary of (left,right)->kerning
|
|
|
|
|
|
|
|
The key of the dictionary is a tuple of ``(left, right)``
|
|
|
|
where each element is a glyph index. The value of the dictionary is
|
|
|
|
the horizontal pairwise kerning in em.
|
|
|
|
"""
|
|
|
|
if self._glyph_kernings:
|
|
|
|
return self._glyph_kernings
|
|
|
|
header = \
|
|
|
|
_read_kern_header_table(self._data, self._tables['kern'].offset)
|
|
|
|
offset = self._tables['kern'].offset + header.size
|
|
|
|
kernings = {}
|
|
|
|
for i in range(header.n_tables):
|
|
|
|
header = _read_kern_subtable_header(self._data, offset)
|
|
|
|
if header.coverage & header.horizontal_mask \
|
|
|
|
and not header.coverage & header.minimum_mask \
|
|
|
|
and not header.coverage & header.perpendicular_mask:
|
|
|
|
if header.coverage & header.format_mask == 0:
|
|
|
|
self._add_kernings_format0(kernings, offset + header.size)
|
|
|
|
offset += header.length
|
|
|
|
self._glyph_kernings = kernings
|
|
|
|
return kernings
|
|
|
|
|
|
|
|
def _add_kernings_format0(self, kernings, offset):
|
|
|
|
header = _read_kern_subtable_format0(self._data, offset)
|
|
|
|
kerning_pairs = _read_kern_subtable_format0Pair.array(self._data,
|
|
|
|
offset + header.size, header.n_pairs)
|
|
|
|
for pair in kerning_pairs:
|
|
|
|
if (pair.left, pair.right) in kernings:
|
|
|
|
kernings[(pair.left, pair.right)] += pair.value \
|
|
|
|
/ float(self.header.units_per_em)
|
|
|
|
else:
|
|
|
|
kernings[(pair.left, pair.right)] = pair.value \
|
|
|
|
/ float(self.header.units_per_em)
|
|
|
|
|
|
|
|
def get_glyph_map(self):
|
|
|
|
"""Calculate and return a reverse character map.
|
|
|
|
|
|
|
|
Returns a dictionary where the key is a glyph index and the
|
|
|
|
value is a unit-length unicode string.
|
|
|
|
"""
|
|
|
|
if self._glyph_map:
|
|
|
|
return self._glyph_map
|
|
|
|
cmap = self.get_character_map()
|
|
|
|
self._glyph_map = {}
|
|
|
|
for ch, glyph in cmap.items():
|
|
|
|
if not glyph in self._glyph_map:
|
|
|
|
self._glyph_map[glyph] = ch
|
|
|
|
return self._glyph_map
|
|
|
|
|
|
|
|
def get_character_map(self):
|
|
|
|
"""Return the character map.
|
|
|
|
|
|
|
|
Returns a dictionary where the key is a unit-length unicode
|
|
|
|
string and the value is a glyph index. Currently only
|
|
|
|
format 4 character maps are read.
|
|
|
|
"""
|
|
|
|
if self._character_map:
|
|
|
|
return self._character_map
|
|
|
|
cmap = _read_cmap_header(self._data, self._tables['cmap'].offset)
|
|
|
|
records = _read_cmap_encoding_record.array(self._data,
|
|
|
|
self._tables['cmap'].offset + cmap.size, cmap.num_tables)
|
|
|
|
self._character_map = {}
|
|
|
|
for record in records:
|
|
|
|
if record.platform_id == 3 and record.encoding_id == 1:
|
|
|
|
# Look at Windows Unicode charmaps only
|
|
|
|
offset = self._tables['cmap'].offset + record.offset
|
|
|
|
format_header = _read_cmap_format_header(self._data, offset)
|
|
|
|
if format_header.format == 4:
|
|
|
|
self._character_map = \
|
|
|
|
self._get_character_map_format4(offset)
|
|
|
|
break
|
|
|
|
return self._character_map
|
|
|
|
|
|
|
|
def _get_character_map_format4(self, offset):
|
|
|
|
# This is absolutely, without question, the *worst* file
|
|
|
|
# format ever. Whoever the fuckwit is that thought this up is
|
|
|
|
# a fuckwit.
|
|
|
|
header = _read_cmap_format4Header(self._data, offset)
|
|
|
|
seg_count = header.seg_count_x2 // 2
|
|
|
|
array_size = struct.calcsize('>%dH' % seg_count)
|
|
|
|
end_count = self._read_array('>%dH' % seg_count,
|
|
|
|
offset + header.size)
|
|
|
|
start_count = self._read_array('>%dH' % seg_count,
|
|
|
|
offset + header.size + array_size + 2)
|
|
|
|
id_delta = self._read_array('>%dh' % seg_count,
|
|
|
|
offset + header.size + array_size + 2 + array_size)
|
|
|
|
id_range_offset_address = \
|
|
|
|
offset + header.size + array_size + 2 + array_size + array_size
|
|
|
|
id_range_offset = self._read_array('>%dH' % seg_count,
|
|
|
|
id_range_offset_address)
|
|
|
|
character_map = {}
|
|
|
|
for i in range(0, seg_count):
|
|
|
|
if id_range_offset[i] != 0:
|
|
|
|
if id_range_offset[i] == 65535:
|
|
|
|
continue # Hack around a dodgy font (babelfish.ttf)
|
|
|
|
for c in range(start_count[i], end_count[i] + 1):
|
|
|
|
addr = id_range_offset[i] + 2 * (c - start_count[i]) + \
|
|
|
|
id_range_offset_address + 2 * i
|
|
|
|
g = struct.unpack('>H', self._data[addr:addr + 2])[0]
|
|
|
|
if g != 0:
|
|
|
|
character_map[chr(c)] = (g + id_delta[i]) % 65536
|
|
|
|
else:
|
|
|
|
for c in range(start_count[i], end_count[i] + 1):
|
|
|
|
g = (c + id_delta[i]) % 65536
|
|
|
|
if g != 0:
|
|
|
|
character_map[chr(c)] = g
|
|
|
|
return character_map
|
|
|
|
|
|
|
|
def _read_array(self, format, offset):
|
|
|
|
size = struct.calcsize(format)
|
|
|
|
return struct.unpack(format, self._data[offset:offset + size])
|
|
|
|
|
|
|
|
def close(self):
|
|
|
|
"""Close the font file.
|
|
|
|
|
|
|
|
This is a good idea, since the entire file is memory mapped in
|
|
|
|
until this method is called. After closing cannot rely on the
|
|
|
|
``get_*`` methods.
|
|
|
|
"""
|
|
|
|
|
|
|
|
self._data.close()
|
|
|
|
os.close(self._fileno)
|
|
|
|
self._closed = True
|
|
|
|
|
|
|
|
def __del__(self):
|
|
|
|
if not self._closed:
|
|
|
|
self.close()
|
|
|
|
|
|
|
|
|
|
|
|
def _read_table(*entries):
|
|
|
|
""" Generic table constructor used for table formats listed at
|
|
|
|
end of file."""
|
|
|
|
|
|
|
|
fmt = '>'
|
|
|
|
names = []
|
|
|
|
for entry in entries:
|
|
|
|
name, entry_type = entry.split(':')
|
|
|
|
names.append(name)
|
|
|
|
fmt += entry_type
|
|
|
|
|
|
|
|
class TableClass:
|
|
|
|
size = struct.calcsize(fmt)
|
|
|
|
|
|
|
|
def __init__(self, data, offset):
|
|
|
|
items = struct.unpack(fmt, data[offset:offset + self.size])
|
|
|
|
self.pairs = list(zip(names, items))
|
|
|
|
for pname, pvalue in self.pairs:
|
|
|
|
if isinstance(pvalue, bytes):
|
|
|
|
pvalue = pvalue.decode('utf-8')
|
|
|
|
setattr(self, pname, pvalue)
|
|
|
|
|
|
|
|
def __repr__(self):
|
|
|
|
return '{'+', '.join(['%s = %s' % (pname, pvalue) for pname, pvalue in self.pairs])+'}'
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def array(data, offset, count):
|
|
|
|
tables = []
|
|
|
|
for i in range(count):
|
|
|
|
tables.append(TableClass(data, offset))
|
|
|
|
offset += TableClass.size
|
|
|
|
return tables
|
|
|
|
|
|
|
|
return TableClass
|
|
|
|
|
|
|
|
|
|
|
|
# Table formats (see references)
|
|
|
|
|
|
|
|
_read_offset_table = _read_table('scalertype:I',
|
|
|
|
'num_tables:H',
|
|
|
|
'search_range:H',
|
|
|
|
'entry_selector:H',
|
|
|
|
'range_shift:H')
|
|
|
|
|
|
|
|
_read_table_directory_entry = _read_table('tag:4s',
|
|
|
|
'check_sum:I',
|
|
|
|
'offset:I',
|
|
|
|
'length:I')
|
|
|
|
|
|
|
|
_read_head_table = _read_table('version:i',
|
|
|
|
'font_revision:i',
|
|
|
|
'check_sum_adjustment:L',
|
|
|
|
'magic_number:L',
|
|
|
|
'flags:H',
|
|
|
|
'units_per_em:H',
|
|
|
|
'created:Q',
|
|
|
|
'modified:Q',
|
|
|
|
'x_min:h',
|
|
|
|
'y_min:h',
|
|
|
|
'x_max:h',
|
|
|
|
'y_max:h',
|
|
|
|
'mac_style:H',
|
|
|
|
'lowest_rec_p_pEM:H',
|
|
|
|
'font_direction_hint:h',
|
|
|
|
'index_to_loc_format:h',
|
|
|
|
'glyph_data_format:h')
|
|
|
|
|
|
|
|
_read_OS2_table = _read_table('version:H',
|
|
|
|
'x_avg_char_width:h',
|
|
|
|
'us_weight_class:H',
|
|
|
|
'us_width_class:H',
|
|
|
|
'fs_type:H',
|
|
|
|
'y_subscript_x_size:h',
|
|
|
|
'y_subscript_y_size:h',
|
|
|
|
'y_subscript_x_offset:h',
|
|
|
|
'y_subscript_y_offset:h',
|
|
|
|
'y_superscript_x_size:h',
|
|
|
|
'y_superscript_y_size:h',
|
|
|
|
'y_superscript_x_offset:h',
|
|
|
|
'y_superscript_y_offset:h',
|
|
|
|
'y_strikeout_size:h',
|
|
|
|
'y_strikeout_position:h',
|
|
|
|
's_family_class:h',
|
|
|
|
'panose1:B',
|
|
|
|
'panose2:B',
|
|
|
|
'panose3:B',
|
|
|
|
'panose4:B',
|
|
|
|
'panose5:B',
|
|
|
|
'panose6:B',
|
|
|
|
'panose7:B',
|
|
|
|
'panose8:B',
|
|
|
|
'panose9:B',
|
|
|
|
'panose10:B',
|
|
|
|
'ul_unicode_range1:L',
|
|
|
|
'ul_unicode_range2:L',
|
|
|
|
'ul_unicode_range3:L',
|
|
|
|
'ul_unicode_range4:L',
|
|
|
|
'ach_vend_id:I',
|
|
|
|
'fs_selection:H',
|
|
|
|
'us_first_char_index:H',
|
|
|
|
'us_last_char_index:H',
|
|
|
|
's_typo_ascender:h',
|
|
|
|
's_typo_descender:h',
|
|
|
|
's_typo_line_gap:h',
|
|
|
|
'us_win_ascent:H',
|
|
|
|
'us_win_descent:H',
|
|
|
|
'ul_code_page_range1:L',
|
|
|
|
'ul_code_page_range2:L',
|
|
|
|
'sx_height:h',
|
|
|
|
's_cap_height:h',
|
|
|
|
'us_default_char:H',
|
|
|
|
'us_break_char:H',
|
|
|
|
'us_max_context:H')
|
|
|
|
|
|
|
|
_read_kern_header_table = _read_table('version_num:H',
|
|
|
|
'n_tables:H')
|
|
|
|
|
|
|
|
_read_kern_subtable_header = _read_table('version:H',
|
|
|
|
'length:H',
|
|
|
|
'coverage:H')
|
|
|
|
|
|
|
|
_read_kern_subtable_header.horizontal_mask = 0x1
|
|
|
|
_read_kern_subtable_header.minimum_mask = 0x2
|
|
|
|
_read_kern_subtable_header.perpendicular_mask = 0x4
|
|
|
|
_read_kern_subtable_header.override_mask = 0x5
|
|
|
|
_read_kern_subtable_header.format_mask = 0xf0
|
|
|
|
|
|
|
|
_read_kern_subtable_format0 = _read_table('n_pairs:H',
|
|
|
|
'search_range:H',
|
|
|
|
'entry_selector:H',
|
|
|
|
'range_shift:H')
|
|
|
|
_read_kern_subtable_format0Pair = _read_table('left:H',
|
|
|
|
'right:H',
|
|
|
|
'value:h')
|
|
|
|
|
|
|
|
_read_cmap_header = _read_table('version:H',
|
|
|
|
'num_tables:H')
|
|
|
|
|
|
|
|
_read_cmap_encoding_record = _read_table('platform_id:H',
|
|
|
|
'encoding_id:H',
|
|
|
|
'offset:L')
|
|
|
|
|
|
|
|
_read_cmap_format_header = _read_table('format:H',
|
|
|
|
'length:H')
|
|
|
|
_read_cmap_format4Header = _read_table('format:H',
|
|
|
|
'length:H',
|
|
|
|
'language:H',
|
|
|
|
'seg_count_x2:H',
|
|
|
|
'search_range:H',
|
|
|
|
'entry_selector:H',
|
|
|
|
'range_shift:H')
|
|
|
|
|
|
|
|
_read_horizontal_header = _read_table('version:i',
|
|
|
|
'Advance:h',
|
|
|
|
'Descender:h',
|
|
|
|
'LineGap:h',
|
|
|
|
'advance_width_max:H',
|
|
|
|
'min_left_side_bearing:h',
|
|
|
|
'min_right_side_bearing:h',
|
|
|
|
'x_max_extent:h',
|
|
|
|
'caret_slope_rise:h',
|
|
|
|
'caret_slope_run:h',
|
|
|
|
'caret_offset:h',
|
|
|
|
'reserved1:h',
|
|
|
|
'reserved2:h',
|
|
|
|
'reserved3:h',
|
|
|
|
'reserved4:h',
|
|
|
|
'metric_data_format:h',
|
|
|
|
'number_of_h_metrics:H')
|
|
|
|
|
|
|
|
_read_long_hor_metric = _read_table('advance_width:H',
|
|
|
|
'lsb:h')
|
|
|
|
|
|
|
|
_read_naming_table = _read_table('format:H',
|
|
|
|
'count:H',
|
|
|
|
'string_offset:H')
|
|
|
|
|
|
|
|
_read_name_record = _read_table('platform_id:H',
|
|
|
|
'encoding_id:H',
|
|
|
|
'language_id:H',
|
|
|
|
'name_id:H',
|
|
|
|
'length:H',
|
|
|
|
'offset:H')
|