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 "***************************************************************"
30 # ---------------------------------------------------------------------------
31 # Generation of single fields
32 # ---------------------------------------------------------------------------
37 # Values are tuple (c type, pb ltype)
38 FieldD = descriptor.FieldDescriptorProto
40 FieldD.TYPE_BOOL: ('bool', 'PB_LTYPE_VARINT'),
41 FieldD.TYPE_DOUBLE: ('double', 'PB_LTYPE_FIXED64'),
42 FieldD.TYPE_FIXED32: ('uint32_t', 'PB_LTYPE_FIXED32'),
43 FieldD.TYPE_FIXED64: ('uint64_t', 'PB_LTYPE_FIXED64'),
44 FieldD.TYPE_FLOAT: ('float', 'PB_LTYPE_FIXED32'),
45 FieldD.TYPE_INT32: ('int32_t', 'PB_LTYPE_VARINT'),
46 FieldD.TYPE_INT64: ('int64_t', 'PB_LTYPE_VARINT'),
47 FieldD.TYPE_SFIXED32: ('int32_t', 'PB_LTYPE_FIXED32'),
48 FieldD.TYPE_SFIXED64: ('int64_t', 'PB_LTYPE_FIXED64'),
49 FieldD.TYPE_SINT32: ('int32_t', 'PB_LTYPE_SVARINT'),
50 FieldD.TYPE_SINT64: ('int64_t', 'PB_LTYPE_SVARINT'),
51 FieldD.TYPE_UINT32: ('uint32_t', 'PB_LTYPE_VARINT'),
52 FieldD.TYPE_UINT64: ('uint64_t', 'PB_LTYPE_VARINT')
56 '''Keeps a set of nested names and formats them to C identifier.
57 You can subclass this with your own implementation.
59 def __init__(self, parts = ()):
60 if isinstance(parts, Names):
62 self.parts = tuple(parts)
65 return '_'.join(self.parts)
67 def __add__(self, other):
68 if isinstance(other, (str, unicode)):
69 return Names(self.parts + (other,))
70 elif isinstance(other, tuple):
71 return Names(self.parts + other)
73 raise ValueError("Name parts should be of type str")
75 def names_from_type_name(type_name):
76 '''Parse Names() from FieldDescriptorProto type_name'''
77 if type_name[0] != '.':
78 raise NotImplementedError("Lookup of non-absolute type names is not supported")
79 return Names(type_name[1:].split('.'))
82 def __init__(self, names, desc):
83 '''desc is EnumDescriptorProto'''
84 self.names = names + desc.name
85 self.values = [(self.names + x.name, x.number) for x in desc.value]
88 result = 'typedef enum _%s {\n' % self.names
89 result += ',\n'.join([" %s = %d" % x for x in self.values])
90 result += '\n} %s;' % self.names
94 def __init__(self, struct_name, desc):
95 '''desc is FieldDescriptorProto'''
96 self.tag = desc.number
97 self.struct_name = struct_name
101 self.max_count = None
104 # Parse nanopb-specific field options
105 if desc.options.HasExtension(nanopb_pb2.nanopb):
106 ext = desc.options.Extensions[nanopb_pb2.nanopb]
107 if ext.HasField("max_size"):
108 self.max_size = ext.max_size
109 if ext.HasField("max_count"):
110 self.max_count = ext.max_count
112 if desc.HasField('default_value'):
113 self.default = desc.default_value
116 # HTYPE is the high-order nibble of nanopb field description,
117 # defining whether value is required/optional/repeated.
119 if desc.label == FieldD.LABEL_REQUIRED:
120 self.htype = 'PB_HTYPE_REQUIRED'
121 elif desc.label == FieldD.LABEL_OPTIONAL:
122 self.htype = 'PB_HTYPE_OPTIONAL'
123 elif desc.label == FieldD.LABEL_REPEATED:
124 if self.max_count is None:
127 self.htype = 'PB_HTYPE_ARRAY'
128 self.array_decl = '[%d]' % self.max_count
130 raise NotImplementedError(desc.label)
132 # Decide LTYPE and CTYPE
133 # LTYPE is the low-order nibble of nanopb field description,
134 # defining how to decode an individual value.
135 # CTYPE is the name of the c type to use in the struct.
136 if datatypes.has_key(desc.type):
137 self.ctype, self.ltype = datatypes[desc.type]
138 elif desc.type == FieldD.TYPE_ENUM:
139 self.ltype = 'PB_LTYPE_VARINT'
140 self.ctype = names_from_type_name(desc.type_name)
141 if self.default is not None:
142 self.default = self.ctype + self.default
143 elif desc.type == FieldD.TYPE_STRING:
144 self.ltype = 'PB_LTYPE_STRING'
145 if self.max_size is None:
149 self.array_decl += '[%d]' % self.max_size
150 elif desc.type == FieldD.TYPE_BYTES:
151 self.ltype = 'PB_LTYPE_BYTES'
152 if self.max_size is None:
155 self.ctype = self.struct_name + self.name + 't'
156 elif desc.type == FieldD.TYPE_MESSAGE:
157 self.ltype = 'PB_LTYPE_SUBMESSAGE'
158 self.ctype = self.submsgname = names_from_type_name(desc.type_name)
160 raise NotImplementedError(desc.type)
163 self.htype = 'PB_HTYPE_CALLBACK'
164 self.ctype = 'pb_callback_t'
167 def __cmp__(self, other):
168 return cmp(self.tag, other.tag)
171 if self.htype == 'PB_HTYPE_OPTIONAL':
172 result = ' bool has_' + self.name + ';\n'
173 elif self.htype == 'PB_HTYPE_ARRAY':
174 result = ' size_t ' + self.name + '_count;\n'
177 result += ' %s %s%s;' % (self.ctype, self.name, self.array_decl)
181 '''Return definitions for any special types this field might need.'''
182 if self.ltype == 'PB_LTYPE_BYTES' and self.max_size is not None:
183 result = 'typedef struct {\n'
184 result += ' size_t size;\n'
185 result += ' uint8_t bytes[%d];\n' % self.max_size
186 result += '} %s;\n' % self.ctype
191 def default_decl(self, declaration_only = False):
192 '''Return definition for this field's default value.'''
193 if self.default is None:
196 if self.ltype == 'PB_LTYPE_STRING':
198 if self.max_size is None:
199 return None # Not implemented
201 array_decl = '[%d]' % (self.max_size + 1)
202 default = str(self.default).encode('string_escape')
203 default = default.replace('"', '\\"')
204 default = '"' + default + '"'
205 elif self.ltype == 'PB_LTYPE_BYTES':
206 data = self.default.decode('string_escape')
207 data = ['0x%02x' % ord(c) for c in data]
209 if self.max_size is None:
210 return None # Not implemented
214 default = '{%d, {%s}}' % (len(data), ','.join(data))
217 ctype, default = self.ctype, self.default
221 return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
223 return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
225 def pb_field_t(self, prev_field_name):
226 '''Return the pb_field_t initializer to use in the constant array.
227 prev_field_name is the name of the previous field or None.
229 result = ' {%d, ' % self.tag
230 result += '(pb_type_t) ((int) ' + self.htype
231 if self.ltype is not None:
232 result += ' | (int) ' + self.ltype
235 if prev_field_name is None:
236 result += ' offsetof(%s, %s),' % (self.struct_name, self.name)
238 result += ' pb_delta_end(%s, %s, %s),' % (self.struct_name, self.name, prev_field_name)
240 if self.htype == 'PB_HTYPE_OPTIONAL':
241 result += '\n pb_delta(%s, has_%s, %s),' % (self.struct_name, self.name, self.name)
242 elif self.htype == 'PB_HTYPE_ARRAY':
243 result += '\n pb_delta(%s, %s_count, %s),' % (self.struct_name, self.name, self.name)
248 if self.htype == 'PB_HTYPE_ARRAY':
249 result += '\n pb_membersize(%s, %s[0]),' % (self.struct_name, self.name)
250 result += ('\n pb_membersize(%s, %s) / pb_membersize(%s, %s[0]),'
251 % (self.struct_name, self.name, self.struct_name, self.name))
253 result += '\n pb_membersize(%s, %s),' % (self.struct_name, self.name)
256 if self.ltype == 'PB_LTYPE_SUBMESSAGE':
257 result += '\n &%s_fields}' % self.submsgname
258 elif self.default is None or self.htype == 'PB_HTYPE_CALLBACK':
261 result += '\n &%s_default}' % (self.struct_name + self.name)
265 def largest_field_value(self):
266 '''Determine if this field needs 16bit or 32bit pb_field_t structure to compile properly.
267 Returns numeric value or a C-expression for assert.'''
268 if self.ltype == 'PB_LTYPE_SUBMESSAGE':
269 if self.htype == 'PB_HTYPE_ARRAY':
270 return 'pb_membersize(%s, %s[0])' % (self.struct_name, self.name)
272 return 'pb_membersize(%s, %s)' % (self.struct_name, self.name)
274 return max(self.tag, self.max_size, self.max_count)
281 # ---------------------------------------------------------------------------
282 # Generation of messages (structures)
283 # ---------------------------------------------------------------------------
287 def __init__(self, names, desc):
289 self.fields = [Field(self.name, f) for f in desc.field]
290 self.ordered_fields = self.fields[:]
291 self.ordered_fields.sort()
293 def get_dependencies(self):
294 '''Get list of type names that this structure refers to.'''
295 return [str(field.ctype) for field in self.fields]
298 result = 'typedef struct _%s {\n' % self.name
299 result += '\n'.join([str(f) for f in self.ordered_fields])
300 result += '\n} %s;' % self.name
305 for field in self.fields:
306 types = field.types()
307 if types is not None:
308 result += types + '\n'
311 def default_decl(self, declaration_only = False):
313 for field in self.fields:
314 default = field.default_decl(declaration_only)
315 if default is not None:
316 result += default + '\n'
319 def fields_declaration(self):
320 result = 'extern const pb_field_t %s_fields[%d];' % (self.name, len(self.fields) + 1)
323 def fields_definition(self):
324 result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, len(self.fields) + 1)
327 for field in self.ordered_fields:
328 result += field.pb_field_t(prev)
332 result += ' PB_LAST_FIELD\n};'
340 # ---------------------------------------------------------------------------
341 # Processing of entire .proto files
342 # ---------------------------------------------------------------------------
345 def iterate_messages(desc, names = Names()):
346 '''Recursively find all messages. For each, yield name, DescriptorProto.'''
347 if hasattr(desc, 'message_type'):
348 submsgs = desc.message_type
350 submsgs = desc.nested_type
352 for submsg in submsgs:
353 sub_names = names + submsg.name
354 yield sub_names, submsg
356 for x in iterate_messages(submsg, sub_names):
359 def parse_file(fdesc):
360 '''Takes a FileDescriptorProto and returns tuple (enum, messages).'''
366 base_name = Names(fdesc.package.split('.'))
370 for enum in fdesc.enum_type:
371 enums.append(Enum(base_name, enum))
373 for names, message in iterate_messages(fdesc, base_name):
374 messages.append(Message(names, message))
375 for enum in message.enum_type:
376 enums.append(Enum(names, enum))
378 return enums, messages
382 From http://code.activestate.com/recipes/577413-topological-sort/
383 This function is under the MIT license.
385 for k, v in data.items():
386 v.discard(k) # Ignore self dependencies
387 extra_items_in_deps = reduce(set.union, data.values(), set()) - set(data.keys())
388 data.update(dict([(item, set()) for item in extra_items_in_deps]))
390 ordered = set(item for item,dep in data.items() if not dep)
393 for item in sorted(ordered):
395 data = dict([(item, (dep - ordered)) for item,dep in data.items()
396 if item not in ordered])
397 assert not data, "A cyclic dependency exists amongst %r" % data
399 def sort_dependencies(messages):
400 '''Sort a list of Messages based on dependencies.'''
403 for message in messages:
404 dependencies[str(message.name)] = set(message.get_dependencies())
405 message_by_name[str(message.name)] = message
407 for msgname in toposort2(dependencies):
408 if msgname in message_by_name:
409 yield message_by_name[msgname]
411 def generate_header(dependencies, headername, enums, messages):
412 '''Generate content for a header file.
413 Generates strings, which should be concatenated and stored to file.
416 yield '/* Automatically generated nanopb header */\n'
418 symbol = headername.replace('.', '_').upper()
419 yield '#ifndef _PB_%s_\n' % symbol
420 yield '#define _PB_%s_\n' % symbol
421 yield '#include <pb.h>\n\n'
423 for dependency in dependencies:
424 noext = os.path.splitext(dependency)[0]
425 yield '#include "%s.pb.h"\n' % noext
428 yield '/* Enum definitions */\n'
430 yield str(enum) + '\n\n'
432 yield '/* Struct definitions */\n'
433 for msg in sort_dependencies(messages):
435 yield str(msg) + '\n\n'
437 yield '/* Default values for struct fields */\n'
439 yield msg.default_decl(True)
442 yield '/* Struct field encoding specification for nanopb */\n'
444 yield msg.fields_declaration() + '\n'
447 count_required_fields = lambda m: len([f for f in msg.fields if f.htype == 'PB_HTYPE_REQUIRED'])
448 largest_msg = max(messages, key = count_required_fields)
449 largest_count = count_required_fields(largest_msg)
450 if largest_count > 64:
451 yield '\n/* Check that missing required fields will be properly detected */\n'
452 yield '#if PB_MAX_REQUIRED_FIELDS < %d\n' % largest_count
453 yield '#error Properly detecting missing required fields in %s requires \\\n' % largest_msg.name
454 yield ' setting PB_MAX_REQUIRED_FIELDS to %d or more.\n' % largest_count
461 for field in msg.fields:
462 status = field.largest_field_value()
463 if isinstance(status, (str, unicode)):
464 checks.append(status)
467 worst_field = str(field.struct_name) + '.' + str(field.name)
469 if worst > 255 or checks:
470 yield '\n/* Check that field information fits in pb_field_t */\n'
473 yield '#if !defined(PB_FIELD_16BIT) && !defined(PB_FIELD_32BIT)\n'
475 yield '#error Field descriptor for %s is too large. Define PB_FIELD_16BIT to fix this.\n' % worst_field
477 assertion = ' && '.join(str(c) + ' < 256' for c in checks)
478 yield 'STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_16BIT)\n' % assertion
481 if worst > 65535 or checks:
482 yield '#if !defined(PB_FIELD_32BIT)\n'
484 yield '#error Field descriptor for %s is too large. Define PB_FIELD_32BIT to fix this.\n' % worst_field
486 assertion = ' && '.join(str(c) + ' < 65536' for c in checks)
487 yield 'STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_32BIT)\n' % assertion
493 def generate_source(headername, enums, messages):
494 '''Generate content for a source file.'''
496 yield '/* Automatically generated nanopb constant definitions */\n'
497 yield '#include "%s"\n\n' % headername
500 yield msg.default_decl(False)
505 yield msg.fields_definition() + '\n\n'
509 # ---------------------------------------------------------------------------
510 # Command line interface
511 # ---------------------------------------------------------------------------
515 from optparse import OptionParser
517 optparser = OptionParser(
518 usage = "Usage: nanopb_generator.py [options] file.pb ...",
519 epilog = "Compile file.pb from file.proto by: 'protoc -ofile.pb file.proto'. " +
520 "Output will be written to file.pb.h and file.pb.c.")
521 optparser.add_option("-x", dest="exclude", metavar="FILE", action="append", default=[],
522 help="Exclude file from generated #include list.")
523 optparser.add_option("-q", "--quiet", dest="quiet", action="store_true", default=False,
524 help="Don't print anything except errors.")
526 def process(filenames, options):
527 '''Process the files given on the command line.'''
530 optparser.print_help()
533 for filename in filenames:
534 data = open(filename, 'rb').read()
535 fdesc = descriptor.FileDescriptorSet.FromString(data)
536 enums, messages = parse_file(fdesc.file[0])
538 noext = os.path.splitext(filename)[0]
539 headername = noext + '.pb.h'
540 sourcename = noext + '.pb.c'
541 headerbasename = os.path.basename(headername)
543 if not options.quiet:
544 print "Writing to " + headername + " and " + sourcename
546 # List of .proto files that should not be included in the C header file
547 # even if they are mentioned in the source .proto.
548 excludes = ['nanopb.proto', 'google/protobuf/descriptor.proto']
549 dependencies = [d for d in fdesc.file[0].dependency if d not in excludes]
551 header = open(headername, 'w')
552 for part in generate_header(dependencies, headerbasename, enums, messages):
555 source = open(sourcename, 'w')
556 for part in generate_source(headerbasename, enums, messages):
561 if __name__ == '__main__':
562 options, filenames = optparser.parse_args()
563 status = process(filenames, options)