1 '''Generate header file for nanopb from a ProtoBuf FileDescriptorSet.'''
4 import google.protobuf.descriptor_pb2 as descriptor
7 print "*************************************************************"
8 print "*** Could not import the Google protobuf Python libraries ***"
9 print "*** Try installing package 'python-protobuf' or similar. ***"
10 print "*************************************************************"
18 print "***************************************************************"
19 print "*** Could not import the precompiled nanopb_pb2.py. ***"
20 print "*** Run 'make' in the 'generator' folder to update the file.***"
21 print "***************************************************************"
27 # Values are tuple (c type, pb ltype)
28 FieldD = descriptor.FieldDescriptorProto
30 FieldD.TYPE_BOOL: ('bool', 'PB_LTYPE_VARINT'),
31 FieldD.TYPE_DOUBLE: ('double', 'PB_LTYPE_FIXED64'),
32 FieldD.TYPE_FIXED32: ('uint32_t', 'PB_LTYPE_FIXED32'),
33 FieldD.TYPE_FIXED64: ('uint64_t', 'PB_LTYPE_FIXED64'),
34 FieldD.TYPE_FLOAT: ('float', 'PB_LTYPE_FIXED32'),
35 FieldD.TYPE_INT32: ('int32_t', 'PB_LTYPE_VARINT'),
36 FieldD.TYPE_INT64: ('int64_t', 'PB_LTYPE_VARINT'),
37 FieldD.TYPE_SFIXED32: ('int32_t', 'PB_LTYPE_FIXED32'),
38 FieldD.TYPE_SFIXED64: ('int64_t', 'PB_LTYPE_FIXED64'),
39 FieldD.TYPE_SINT32: ('int32_t', 'PB_LTYPE_SVARINT'),
40 FieldD.TYPE_SINT64: ('int64_t', 'PB_LTYPE_SVARINT'),
41 FieldD.TYPE_UINT32: ('uint32_t', 'PB_LTYPE_VARINT'),
42 FieldD.TYPE_UINT64: ('uint64_t', 'PB_LTYPE_VARINT')
46 '''Keeps a set of nested names and formats them to C identifier.
47 You can subclass this with your own implementation.
49 def __init__(self, parts = ()):
50 if isinstance(parts, Names):
52 self.parts = tuple(parts)
55 return '_'.join(self.parts)
57 def __add__(self, other):
58 if isinstance(other, (str, unicode)):
59 return Names(self.parts + (other,))
60 elif isinstance(other, tuple):
61 return Names(self.parts + other)
63 raise ValueError("Name parts should be of type str")
65 def names_from_type_name(type_name):
66 '''Parse Names() from FieldDescriptorProto type_name'''
67 if type_name[0] != '.':
68 raise NotImplementedError("Lookup of non-absolute type names is not supported")
69 return Names(type_name[1:].split('.'))
72 def __init__(self, names, desc):
73 '''desc is EnumDescriptorProto'''
74 self.names = names + desc.name
75 self.values = [(self.names + x.name, x.number) for x in desc.value]
78 result = 'typedef enum {\n'
79 result += ',\n'.join([" %s = %d" % x for x in self.values])
80 result += '\n} %s;' % self.names
84 def __init__(self, struct_name, desc):
85 '''desc is FieldDescriptorProto'''
86 self.tag = desc.number
87 self.struct_name = struct_name
94 # Parse nanopb-specific field options
95 if desc.options.HasExtension(nanopb_pb2.nanopb):
96 ext = desc.options.Extensions[nanopb_pb2.nanopb]
97 if ext.HasField("max_size"):
98 self.max_size = ext.max_size
99 if ext.HasField("max_count"):
100 self.max_count = ext.max_count
102 if desc.HasField('default_value'):
103 self.default = desc.default_value
106 # HTYPE is the high-order nibble of nanopb field description,
107 # defining whether value is required/optional/repeated.
109 if desc.label == FieldD.LABEL_REQUIRED:
110 self.htype = 'PB_HTYPE_REQUIRED'
111 elif desc.label == FieldD.LABEL_OPTIONAL:
112 self.htype = 'PB_HTYPE_OPTIONAL'
113 elif desc.label == FieldD.LABEL_REPEATED:
114 if self.max_count is None:
117 self.htype = 'PB_HTYPE_ARRAY'
118 self.array_decl = '[%d]' % self.max_count
120 raise NotImplementedError(desc.label)
122 # Decide LTYPE and CTYPE
123 # LTYPE is the low-order nibble of nanopb field description,
124 # defining how to decode an individual value.
125 # CTYPE is the name of the c type to use in the struct.
126 if datatypes.has_key(desc.type):
127 self.ctype, self.ltype = datatypes[desc.type]
128 elif desc.type == FieldD.TYPE_ENUM:
129 self.ltype = 'PB_LTYPE_VARINT'
130 self.ctype = names_from_type_name(desc.type_name)
131 if self.default is not None:
132 self.default = self.ctype + self.default
133 elif desc.type == FieldD.TYPE_STRING:
134 self.ltype = 'PB_LTYPE_STRING'
135 if self.max_size is None:
139 self.array_decl += '[%d]' % self.max_size
140 elif desc.type == FieldD.TYPE_BYTES:
141 self.ltype = 'PB_LTYPE_BYTES'
142 if self.max_size is None:
145 self.ctype = self.struct_name + self.name + 't'
146 elif desc.type == FieldD.TYPE_MESSAGE:
147 self.ltype = 'PB_LTYPE_SUBMESSAGE'
148 self.ctype = self.submsgname = names_from_type_name(desc.type_name)
150 raise NotImplementedError(desc.type)
153 self.htype = 'PB_HTYPE_CALLBACK'
154 self.ctype = 'pb_callback_t'
157 def __cmp__(self, other):
158 return cmp(self.tag, other.tag)
161 if self.htype == 'PB_HTYPE_OPTIONAL':
162 result = ' bool has_' + self.name + ';\n'
163 elif self.htype == 'PB_HTYPE_ARRAY':
164 result = ' size_t ' + self.name + '_count;\n'
167 result += ' %s %s%s;' % (self.ctype, self.name, self.array_decl)
171 '''Return definitions for any special types this field might need.'''
172 if self.ltype == 'PB_LTYPE_BYTES' and self.max_size is not None:
173 result = 'typedef struct {\n'
174 result += ' size_t size;\n'
175 result += ' uint8_t bytes[%d];\n' % self.max_size
176 result += '} %s;\n' % self.ctype
181 def default_decl(self, declaration_only = False):
182 '''Return definition for this field's default value.'''
183 if self.default is None:
186 if self.ltype == 'PB_LTYPE_STRING':
188 if self.max_size is None:
189 return None # Not implemented
191 array_decl = '[%d]' % (self.max_size + 1)
192 default = str(self.default).encode('string_escape')
193 default = default.replace('"', '\\"')
194 default = '"' + default + '"'
195 elif self.ltype == 'PB_LTYPE_BYTES':
196 data = self.default.decode('string_escape')
197 data = ['0x%02x' % ord(c) for c in data]
199 if self.max_size is None:
200 return None # Not implemented
204 default = '{%d, {%s}}' % (len(data), ','.join(data))
207 ctype, default = self.ctype, self.default
211 return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
213 return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
215 def pb_field_t(self, prev_field_name):
216 '''Return the pb_field_t initializer to use in the constant array.
217 prev_field_name is the name of the previous field or None.
219 result = ' {%d, ' % self.tag
220 result += '(pb_type_t) ((int) ' + self.htype
221 if self.ltype is not None:
222 result += ' | (int) ' + self.ltype
225 if prev_field_name is None:
226 result += ' offsetof(%s, %s),' % (self.struct_name, self.name)
228 result += ' pb_delta_end(%s, %s, %s),' % (self.struct_name, self.name, prev_field_name)
230 if self.htype == 'PB_HTYPE_OPTIONAL':
231 result += '\n pb_delta(%s, has_%s, %s),' % (self.struct_name, self.name, self.name)
232 elif self.htype == 'PB_HTYPE_ARRAY':
233 result += '\n pb_delta(%s, %s_count, %s),' % (self.struct_name, self.name, self.name)
238 if self.htype == 'PB_HTYPE_ARRAY':
239 result += '\n pb_membersize(%s, %s[0]),' % (self.struct_name, self.name)
240 result += ('\n pb_membersize(%s, %s) / pb_membersize(%s, %s[0]),'
241 % (self.struct_name, self.name, self.struct_name, self.name))
243 result += '\n pb_membersize(%s, %s),' % (self.struct_name, self.name)
246 if self.ltype == 'PB_LTYPE_SUBMESSAGE':
247 result += '\n &%s_fields}' % self.submsgname
248 elif self.default is None or self.htype == 'PB_HTYPE_CALLBACK':
251 result += '\n &%s_default}' % (self.struct_name + self.name)
255 def largest_field_value(self):
256 '''Determine if this field needs 16bit or 32bit pb_field_t structure to compile properly.
257 Returns numeric value or a C-expression for assert.'''
258 if self.ltype == 'PB_LTYPE_SUBMESSAGE':
259 if self.htype == 'PB_HTYPE_ARRAY':
260 return 'pb_membersize(%s, %s[0])' % (self.struct_name, self.name)
262 return 'pb_membersize(%s, %s)' % (self.struct_name, self.name)
264 return max(self.tag, self.max_size, self.max_count)
268 def __init__(self, names, desc):
270 self.fields = [Field(self.name, f) for f in desc.field]
271 self.ordered_fields = self.fields[:]
272 self.ordered_fields.sort()
274 def get_dependencies(self):
275 '''Get list of type names that this structure refers to.'''
276 return [str(field.ctype) for field in self.fields]
279 result = 'typedef struct {\n'
280 result += '\n'.join([str(f) for f in self.ordered_fields])
281 result += '\n} %s;' % self.name
286 for field in self.fields:
287 types = field.types()
288 if types is not None:
289 result += types + '\n'
292 def default_decl(self, declaration_only = False):
294 for field in self.fields:
295 default = field.default_decl(declaration_only)
296 if default is not None:
297 result += default + '\n'
300 def fields_declaration(self):
301 result = 'extern const pb_field_t %s_fields[%d];' % (self.name, len(self.fields) + 1)
304 def fields_definition(self):
305 result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, len(self.fields) + 1)
308 for field in self.ordered_fields:
309 result += field.pb_field_t(prev)
313 result += ' PB_LAST_FIELD\n};'
316 def iterate_messages(desc, names = Names()):
317 '''Recursively find all messages. For each, yield name, DescriptorProto.'''
318 if hasattr(desc, 'message_type'):
319 submsgs = desc.message_type
321 submsgs = desc.nested_type
323 for submsg in submsgs:
324 sub_names = names + submsg.name
325 yield sub_names, submsg
327 for x in iterate_messages(submsg, sub_names):
330 def parse_file(fdesc):
331 '''Takes a FileDescriptorProto and returns tuple (enum, messages).'''
337 base_name = Names(fdesc.package.split('.'))
341 for enum in fdesc.enum_type:
342 enums.append(Enum(base_name, enum))
344 for names, message in iterate_messages(fdesc, base_name):
345 messages.append(Message(names, message))
346 for enum in message.enum_type:
347 enums.append(Enum(names, enum))
349 return enums, messages
353 From http://code.activestate.com/recipes/577413-topological-sort/
354 This function is under the MIT license.
356 for k, v in data.items():
357 v.discard(k) # Ignore self dependencies
358 extra_items_in_deps = reduce(set.union, data.values(), set()) - set(data.keys())
359 data.update(dict([(item, set()) for item in extra_items_in_deps]))
361 ordered = set(item for item,dep in data.items() if not dep)
364 for item in sorted(ordered):
366 data = dict([(item, (dep - ordered)) for item,dep in data.items()
367 if item not in ordered])
368 assert not data, "A cyclic dependency exists amongst %r" % data
370 def sort_dependencies(messages):
371 '''Sort a list of Messages based on dependencies.'''
374 for message in messages:
375 dependencies[str(message.name)] = set(message.get_dependencies())
376 message_by_name[str(message.name)] = message
378 for msgname in toposort2(dependencies):
379 if msgname in message_by_name:
380 yield message_by_name[msgname]
382 def generate_header(dependencies, headername, enums, messages):
383 '''Generate content for a header file.
384 Generates strings, which should be concatenated and stored to file.
387 yield '/* Automatically generated nanopb header */\n'
389 symbol = headername.replace('.', '_').upper()
390 yield '#ifndef _PB_%s_\n' % symbol
391 yield '#define _PB_%s_\n' % symbol
392 yield '#include <pb.h>\n\n'
394 for dependency in dependencies:
395 noext = os.path.splitext(dependency)[0]
396 yield '#include "%s.pb.h"\n' % noext
399 yield '/* Enum definitions */\n'
401 yield str(enum) + '\n\n'
403 yield '/* Struct definitions */\n'
404 for msg in sort_dependencies(messages):
406 yield str(msg) + '\n\n'
408 yield '/* Default values for struct fields */\n'
410 yield msg.default_decl(True)
413 yield '/* Struct field encoding specification for nanopb */\n'
415 yield msg.fields_declaration() + '\n'
417 count_required_fields = lambda m: len([f for f in msg.fields if f.htype == 'PB_HTYPE_REQUIRED'])
418 largest_msg = max(messages, key = count_required_fields)
419 largest_count = count_required_fields(largest_msg)
420 if largest_count > 64:
421 yield '\n/* Check that missing required fields will be properly detected */\n'
422 yield '#if PB_MAX_REQUIRED_FIELDS < %d\n' % largest_count
423 yield '#error Properly detecting missing required fields in %s requires \\\n' % largest_msg.name
424 yield ' setting PB_MAX_REQUIRED_FIELDS to %d or more.\n' % largest_count
431 for field in msg.fields:
432 status = field.largest_field_value()
433 if isinstance(status, (str, unicode)):
434 checks.append(status)
437 worst_field = str(field.struct_name) + '.' + str(field.name)
439 if worst > 255 or checks:
440 yield '\n/* Check that field information fits in pb_field_t */\n'
443 yield '#if !defined(PB_FIELD_16BIT) && !defined(PB_FIELD_32BIT)\n'
445 yield '#error Field descriptor for %s is too large. Define PB_FIELD_16BIT to fix this.\n' % worst_field
447 assertion = ' && '.join(str(c) + ' < 256' for c in checks)
448 yield 'STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_16BIT)\n' % assertion
451 if worst > 65535 or checks:
452 yield '#if !defined(PB_FIELD_32BIT)\n'
454 yield '#error Field descriptor for %s is too large. Define PB_FIELD_32BIT to fix this.\n' % worst_field
456 assertion = ' && '.join(str(c) + ' < 65536' for c in checks)
457 yield 'STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_32BIT)\n' % assertion
463 def generate_source(headername, enums, messages):
464 '''Generate content for a source file.'''
466 yield '/* Automatically generated nanopb constant definitions */\n'
467 yield '#include "%s"\n\n' % headername
470 yield msg.default_decl(False)
475 yield msg.fields_definition() + '\n\n'
477 if __name__ == '__main__':
481 if len(sys.argv) != 2:
482 print "Usage: " + sys.argv[0] + " file.pb"
483 print "where file.pb has been compiled from .proto by:"
484 print "protoc -ofile.pb file.proto"
485 print "Output fill be written to file.pb.h and file.pb.c"
488 data = open(sys.argv[1], 'rb').read()
489 fdesc = descriptor.FileDescriptorSet.FromString(data)
490 enums, messages = parse_file(fdesc.file[0])
492 noext = os.path.splitext(sys.argv[1])[0]
493 headername = noext + '.pb.h'
494 sourcename = noext + '.pb.c'
495 headerbasename = os.path.basename(headername)
497 print "Writing to " + headername + " and " + sourcename
499 # List of .proto files that should not be included in the C header file
500 # even if they are mentioned in the source .proto.
501 excludes = ['nanopb.proto', 'google/protobuf/descriptor.proto']
502 dependencies = [d for d in fdesc.file[0].dependency if d not in excludes]
504 header = open(headername, 'w')
505 for part in generate_header(dependencies, headerbasename, enums, messages):
508 source = open(sourcename, 'w')
509 for part in generate_source(headerbasename, enums, messages):