• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1"""Compiles/decompiles SVG table.
2
3https://docs.microsoft.com/en-us/typography/opentype/spec/svg
4
5The XML format is:
6
7.. code-block:: xml
8
9	<SVG>
10		<svgDoc endGlyphID="1" startGlyphID="1">
11			<![CDATA[ <complete SVG doc> ]]
12		</svgDoc>
13	...
14		<svgDoc endGlyphID="n" startGlyphID="m">
15			<![CDATA[ <complete SVG doc> ]]
16		</svgDoc>
17	</SVG>
18"""
19
20from fontTools.misc.textTools import bytesjoin, safeEval, strjoin, tobytes, tostr
21from fontTools.misc import sstruct
22from . import DefaultTable
23from collections.abc import Sequence
24from dataclasses import dataclass, astuple
25from io import BytesIO
26import struct
27import logging
28
29
30log = logging.getLogger(__name__)
31
32
33SVG_format_0 = """
34	>   # big endian
35	version:                  H
36	offsetToSVGDocIndex:      L
37	reserved:                 L
38"""
39
40SVG_format_0Size = sstruct.calcsize(SVG_format_0)
41
42doc_index_entry_format_0 = """
43	>   # big endian
44	startGlyphID:             H
45	endGlyphID:               H
46	svgDocOffset:             L
47	svgDocLength:             L
48"""
49
50doc_index_entry_format_0Size = sstruct.calcsize(doc_index_entry_format_0)
51
52
53
54class table_S_V_G_(DefaultTable.DefaultTable):
55
56	def decompile(self, data, ttFont):
57		self.docList = []
58		# Version 0 is the standardized version of the table; and current.
59		# https://www.microsoft.com/typography/otspec/svg.htm
60		sstruct.unpack(SVG_format_0, data[:SVG_format_0Size], self)
61		if self.version != 0:
62			log.warning(
63				"Unknown SVG table version '%s'. Decompiling as version 0.", self.version)
64		# read in SVG Documents Index
65		# data starts with the first entry of the entry list.
66		pos = subTableStart = self.offsetToSVGDocIndex
67		self.numEntries = struct.unpack(">H", data[pos:pos+2])[0]
68		pos += 2
69		if self.numEntries > 0:
70			data2 = data[pos:]
71			entries = []
72			for i in range(self.numEntries):
73				docIndexEntry, data2 = sstruct.unpack2(doc_index_entry_format_0, data2, DocumentIndexEntry())
74				entries.append(docIndexEntry)
75
76			for entry in entries:
77				start = entry.svgDocOffset + subTableStart
78				end = start + entry.svgDocLength
79				doc = data[start:end]
80				compressed = False
81				if doc.startswith(b"\x1f\x8b"):
82					import gzip
83					bytesIO = BytesIO(doc)
84					with gzip.GzipFile(None, "r", fileobj=bytesIO) as gunzipper:
85						doc = gunzipper.read()
86					del bytesIO
87					compressed = True
88				doc = tostr(doc, "utf_8")
89				self.docList.append(
90					SVGDocument(doc, entry.startGlyphID, entry.endGlyphID, compressed)
91				)
92
93	def compile(self, ttFont):
94		version = 0
95		offsetToSVGDocIndex = SVG_format_0Size # I start the SVGDocIndex right after the header.
96		# get SGVDoc info.
97		docList = []
98		entryList = []
99		numEntries = len(self.docList)
100		datum = struct.pack(">H",numEntries)
101		entryList.append(datum)
102		curOffset = len(datum) + doc_index_entry_format_0Size*numEntries
103		seenDocs = {}
104		allCompressed = getattr(self, "compressed", False)
105		for i, doc in enumerate(self.docList):
106			if isinstance(doc, (list, tuple)):
107				doc = SVGDocument(*doc)
108				self.docList[i] = doc
109			docBytes = tobytes(doc.data, encoding="utf_8")
110			if (allCompressed or doc.compressed) and not docBytes.startswith(b"\x1f\x8b"):
111				import gzip
112				bytesIO = BytesIO()
113				# mtime=0 strips the useless timestamp and makes gzip output reproducible;
114				# equivalent to `gzip -n`
115				with gzip.GzipFile(None, "w", fileobj=bytesIO, mtime=0) as gzipper:
116					gzipper.write(docBytes)
117				gzipped = bytesIO.getvalue()
118				if len(gzipped) < len(docBytes):
119					docBytes = gzipped
120				del gzipped, bytesIO
121			docLength = len(docBytes)
122			if docBytes in seenDocs:
123				docOffset = seenDocs[docBytes]
124			else:
125				docOffset = curOffset
126				curOffset += docLength
127				seenDocs[docBytes] = docOffset
128				docList.append(docBytes)
129			entry = struct.pack(">HHLL", doc.startGlyphID, doc.endGlyphID, docOffset, docLength)
130			entryList.append(entry)
131		entryList.extend(docList)
132		svgDocData = bytesjoin(entryList)
133
134		reserved = 0
135		header = struct.pack(">HLL", version, offsetToSVGDocIndex, reserved)
136		data = [header, svgDocData]
137		data = bytesjoin(data)
138		return data
139
140	def toXML(self, writer, ttFont):
141		for i, doc in enumerate(self.docList):
142			if isinstance(doc, (list, tuple)):
143				doc = SVGDocument(*doc)
144				self.docList[i] = doc
145			attrs = {"startGlyphID": doc.startGlyphID, "endGlyphID": doc.endGlyphID}
146			if doc.compressed:
147				attrs["compressed"] = 1
148			writer.begintag("svgDoc", **attrs)
149			writer.newline()
150			writer.writecdata(doc.data)
151			writer.newline()
152			writer.endtag("svgDoc")
153			writer.newline()
154
155	def fromXML(self, name, attrs, content, ttFont):
156		if name == "svgDoc":
157			if not hasattr(self, "docList"):
158				self.docList = []
159			doc = strjoin(content)
160			doc = doc.strip()
161			startGID = int(attrs["startGlyphID"])
162			endGID = int(attrs["endGlyphID"])
163			compressed = bool(safeEval(attrs.get("compressed", "0")))
164			self.docList.append(SVGDocument(doc, startGID, endGID, compressed))
165		else:
166			log.warning("Unknown %s %s", name, content)
167
168
169class DocumentIndexEntry(object):
170	def __init__(self):
171		self.startGlyphID = None # USHORT
172		self.endGlyphID = None # USHORT
173		self.svgDocOffset = None # ULONG
174		self.svgDocLength = None # ULONG
175
176	def __repr__(self):
177		return "startGlyphID: %s, endGlyphID: %s, svgDocOffset: %s, svgDocLength: %s" % (self.startGlyphID, self.endGlyphID, self.svgDocOffset, self.svgDocLength)
178
179
180@dataclass
181class SVGDocument(Sequence):
182	data: str
183	startGlyphID: int
184	endGlyphID: int
185	compressed: bool = False
186
187	# Previously, the SVG table's docList attribute contained a lists of 3 items:
188	# [doc, startGlyphID, endGlyphID]; later, we added a `compressed` attribute.
189	# For backward compatibility with code that depends of them being sequences of
190	# fixed length=3, we subclass the Sequence abstract base class and pretend only
191	# the first three items are present. 'compressed' is only accessible via named
192	# attribute lookup like regular dataclasses: i.e. `doc.compressed`, not `doc[3]`
193	def __getitem__(self, index):
194		return astuple(self)[:3][index]
195
196	def __len__(self):
197		return 3
198