3 '''Generate header file for nanopb from a ProtoBuf FileDescriptorSet.'''
4 nanopb_version = "nanopb-0.3.4-dev"
10 # Add some dummy imports to keep packaging tools happy.
11 import google, distutils.util # bbfreeze seems to need these
12 import pkg_resources # pyinstaller / protobuf 2.5 seem to need these
14 # Don't care, we will error out later if it is actually important.
18 import google.protobuf.text_format as text_format
19 import google.protobuf.descriptor_pb2 as descriptor
22 *************************************************************
23 *** Could not import the Google protobuf Python libraries ***
24 *** Try installing package 'python-protobuf' or similar. ***
25 *************************************************************
30 import proto.nanopb_pb2 as nanopb_pb2
31 import proto.plugin_pb2 as plugin_pb2
34 ********************************************************************
35 *** Failed to import the protocol definitions for generator. ***
36 *** You have to run 'make' in the nanopb/generator/proto folder. ***
37 ********************************************************************
41 # ---------------------------------------------------------------------------
42 # Generation of single fields
43 # ---------------------------------------------------------------------------
48 # Values are tuple (c type, pb type, encoded size, int_size_allowed)
49 FieldD = descriptor.FieldDescriptorProto
51 FieldD.TYPE_BOOL: ('bool', 'BOOL', 1, False),
52 FieldD.TYPE_DOUBLE: ('double', 'DOUBLE', 8, False),
53 FieldD.TYPE_FIXED32: ('uint32_t', 'FIXED32', 4, False),
54 FieldD.TYPE_FIXED64: ('uint64_t', 'FIXED64', 8, False),
55 FieldD.TYPE_FLOAT: ('float', 'FLOAT', 4, False),
56 FieldD.TYPE_INT32: ('int32_t', 'INT32', 10, True),
57 FieldD.TYPE_INT64: ('int64_t', 'INT64', 10, True),
58 FieldD.TYPE_SFIXED32: ('int32_t', 'SFIXED32', 4, False),
59 FieldD.TYPE_SFIXED64: ('int64_t', 'SFIXED64', 8, False),
60 FieldD.TYPE_SINT32: ('int32_t', 'SINT32', 5, True),
61 FieldD.TYPE_SINT64: ('int64_t', 'SINT64', 10, True),
62 FieldD.TYPE_UINT32: ('uint32_t', 'UINT32', 5, True),
63 FieldD.TYPE_UINT64: ('uint64_t', 'UINT64', 10, True)
66 # Integer size overrides (from .proto settings)
68 nanopb_pb2.IS_8: 'int8_t',
69 nanopb_pb2.IS_16: 'int16_t',
70 nanopb_pb2.IS_32: 'int32_t',
71 nanopb_pb2.IS_64: 'int64_t',
75 '''Keeps a set of nested names and formats them to C identifier.'''
76 def __init__(self, parts = ()):
77 if isinstance(parts, Names):
79 self.parts = tuple(parts)
82 return '_'.join(self.parts)
84 def __add__(self, other):
85 if isinstance(other, (str, unicode)):
86 return Names(self.parts + (other,))
87 elif isinstance(other, tuple):
88 return Names(self.parts + other)
90 raise ValueError("Name parts should be of type str")
92 def __eq__(self, other):
93 return isinstance(other, Names) and self.parts == other.parts
95 def names_from_type_name(type_name):
96 '''Parse Names() from FieldDescriptorProto type_name'''
97 if type_name[0] != '.':
98 raise NotImplementedError("Lookup of non-absolute type names is not supported")
99 return Names(type_name[1:].split('.'))
101 def varint_max_size(max_value):
102 '''Returns the maximum number of bytes a varint can take when encoded.'''
103 for i in range(1, 11):
104 if (max_value >> (i * 7)) == 0:
106 raise ValueError("Value too large for varint: " + str(max_value))
108 assert varint_max_size(0) == 1
109 assert varint_max_size(127) == 1
110 assert varint_max_size(128) == 2
113 '''Class used to represent the encoded size of a field or a message.
114 Consists of a combination of symbolic sizes and integer sizes.'''
115 def __init__(self, value = 0, symbols = []):
116 if isinstance(value, (str, Names)):
117 symbols = [str(value)]
120 self.symbols = symbols
122 def __add__(self, other):
123 if isinstance(other, (int, long)):
124 return EncodedSize(self.value + other, self.symbols)
125 elif isinstance(other, (str, Names)):
126 return EncodedSize(self.value, self.symbols + [str(other)])
127 elif isinstance(other, EncodedSize):
128 return EncodedSize(self.value + other.value, self.symbols + other.symbols)
130 raise ValueError("Cannot add size: " + repr(other))
132 def __mul__(self, other):
133 if isinstance(other, (int, long)):
134 return EncodedSize(self.value * other, [str(other) + '*' + s for s in self.symbols])
136 raise ValueError("Cannot multiply size: " + repr(other))
140 return str(self.value)
142 return '(' + str(self.value) + ' + ' + ' + '.join(self.symbols) + ')'
144 def upperlimit(self):
151 def __init__(self, names, desc, enum_options):
152 '''desc is EnumDescriptorProto'''
154 self.options = enum_options
155 self.names = names + desc.name
157 if enum_options.long_names:
158 self.values = [(self.names + x.name, x.number) for x in desc.value]
160 self.values = [(names + x.name, x.number) for x in desc.value]
162 self.value_longnames = [self.names + x.name for x in desc.value]
163 self.packed = enum_options.packed_enum
166 result = 'typedef enum _%s {\n' % self.names
167 result += ',\n'.join([" %s = %d" % x for x in self.values])
171 result += ' pb_packed'
173 result += ' %s;' % self.names
175 if not self.options.long_names:
176 # Define the long names always so that enum value references
177 # from other files work properly.
178 for i, x in enumerate(self.values):
179 result += '\n#define %s %s' % (self.value_longnames[i], x[0])
184 def __init__(self, struct_name, desc, field_options):
185 '''desc is FieldDescriptorProto'''
186 self.tag = desc.number
187 self.struct_name = struct_name
188 self.union_name = None
189 self.name = desc.name
192 self.max_count = None
197 # Parse field options
198 if field_options.HasField("max_size"):
199 self.max_size = field_options.max_size
201 if field_options.HasField("max_count"):
202 self.max_count = field_options.max_count
204 if desc.HasField('default_value'):
205 self.default = desc.default_value
207 # Check field rules, i.e. required/optional/repeated.
209 if desc.label == FieldD.LABEL_REQUIRED:
210 self.rules = 'REQUIRED'
211 elif desc.label == FieldD.LABEL_OPTIONAL:
212 self.rules = 'OPTIONAL'
213 elif desc.label == FieldD.LABEL_REPEATED:
214 self.rules = 'REPEATED'
215 if self.max_count is None:
216 can_be_static = False
218 self.array_decl = '[%d]' % self.max_count
220 raise NotImplementedError(desc.label)
222 # Check if the field can be implemented with static allocation
223 # i.e. whether the data size is known.
224 if desc.type == FieldD.TYPE_STRING and self.max_size is None:
225 can_be_static = False
227 if desc.type == FieldD.TYPE_BYTES and self.max_size is None:
228 can_be_static = False
230 # Decide how the field data will be allocated
231 if field_options.type == nanopb_pb2.FT_DEFAULT:
233 field_options.type = nanopb_pb2.FT_STATIC
235 field_options.type = nanopb_pb2.FT_CALLBACK
237 if field_options.type == nanopb_pb2.FT_STATIC and not can_be_static:
238 raise Exception("Field %s is defined as static, but max_size or "
239 "max_count is not given." % self.name)
241 if field_options.type == nanopb_pb2.FT_STATIC:
242 self.allocation = 'STATIC'
243 elif field_options.type == nanopb_pb2.FT_POINTER:
244 self.allocation = 'POINTER'
245 elif field_options.type == nanopb_pb2.FT_CALLBACK:
246 self.allocation = 'CALLBACK'
248 raise NotImplementedError(field_options.type)
250 # Decide the C data type to use in the struct.
251 if datatypes.has_key(desc.type):
252 self.ctype, self.pbtype, self.enc_size, isa = datatypes[desc.type]
254 # Override the field size if user wants to use smaller integers
255 if isa and field_options.int_size != nanopb_pb2.IS_DEFAULT:
256 self.ctype = intsizes[field_options.int_size]
257 if desc.type == FieldD.TYPE_UINT32 or desc.type == FieldD.TYPE_UINT64:
258 self.ctype = 'u' + self.ctype;
259 elif desc.type == FieldD.TYPE_ENUM:
261 self.ctype = names_from_type_name(desc.type_name)
262 if self.default is not None:
263 self.default = self.ctype + self.default
264 self.enc_size = 5 # protoc rejects enum values > 32 bits
265 elif desc.type == FieldD.TYPE_STRING:
266 self.pbtype = 'STRING'
268 if self.allocation == 'STATIC':
270 self.array_decl += '[%d]' % self.max_size
271 self.enc_size = varint_max_size(self.max_size) + self.max_size
272 elif desc.type == FieldD.TYPE_BYTES:
273 self.pbtype = 'BYTES'
274 if self.allocation == 'STATIC':
275 self.ctype = self.struct_name + self.name + 't'
276 self.enc_size = varint_max_size(self.max_size) + self.max_size
277 elif self.allocation == 'POINTER':
278 self.ctype = 'pb_bytes_array_t'
279 elif desc.type == FieldD.TYPE_MESSAGE:
280 self.pbtype = 'MESSAGE'
281 self.ctype = self.submsgname = names_from_type_name(desc.type_name)
282 self.enc_size = None # Needs to be filled in after the message type is available
284 raise NotImplementedError(desc.type)
286 def __cmp__(self, other):
287 return cmp(self.tag, other.tag)
291 if self.allocation == 'POINTER':
292 if self.rules == 'REPEATED':
293 result += ' pb_size_t ' + self.name + '_count;\n'
295 if self.pbtype == 'MESSAGE':
296 # Use struct definition, so recursive submessages are possible
297 result += ' struct _%s *%s;' % (self.ctype, self.name)
298 elif self.rules == 'REPEATED' and self.pbtype in ['STRING', 'BYTES']:
299 # String/bytes arrays need to be defined as pointers to pointers
300 result += ' %s **%s;' % (self.ctype, self.name)
302 result += ' %s *%s;' % (self.ctype, self.name)
303 elif self.allocation == 'CALLBACK':
304 result += ' pb_callback_t %s;' % self.name
306 if self.rules == 'OPTIONAL' and self.allocation == 'STATIC':
307 result += ' bool has_' + self.name + ';\n'
308 elif self.rules == 'REPEATED' and self.allocation == 'STATIC':
309 result += ' pb_size_t ' + self.name + '_count;\n'
310 result += ' %s %s%s;' % (self.ctype, self.name, self.array_decl)
314 '''Return definitions for any special types this field might need.'''
315 if self.pbtype == 'BYTES' and self.allocation == 'STATIC':
316 result = 'typedef PB_BYTES_ARRAY_T(%d) %s;\n' % (self.max_size, self.ctype)
321 def get_dependencies(self):
322 '''Get list of type names used by this field.'''
323 if self.allocation == 'STATIC':
324 return [str(self.ctype)]
328 def get_initializer(self, null_init, inner_init_only = False):
329 '''Return literal expression for this field's default value.
330 null_init: If True, initialize to a 0 value instead of default from .proto
331 inner_init_only: If True, exclude initialization for any count/has fields
335 if self.pbtype == 'MESSAGE':
337 inner_init = '%s_init_zero' % self.ctype
339 inner_init = '%s_init_default' % self.ctype
340 elif self.default is None or null_init:
341 if self.pbtype == 'STRING':
343 elif self.pbtype == 'BYTES':
344 inner_init = '{0, {0}}'
345 elif self.pbtype == 'ENUM':
346 inner_init = '(%s)0' % self.ctype
350 if self.pbtype == 'STRING':
351 inner_init = self.default.encode('utf-8').encode('string_escape')
352 inner_init = inner_init.replace('"', '\\"')
353 inner_init = '"' + inner_init + '"'
354 elif self.pbtype == 'BYTES':
355 data = str(self.default).decode('string_escape')
356 data = ['0x%02x' % ord(c) for c in data]
358 inner_init = '{0, {0}}'
360 inner_init = '{%d, {%s}}' % (len(data), ','.join(data))
361 elif self.pbtype in ['FIXED32', 'UINT32']:
362 inner_init = str(self.default) + 'u'
363 elif self.pbtype in ['FIXED64', 'UINT64']:
364 inner_init = str(self.default) + 'ull'
365 elif self.pbtype in ['SFIXED64', 'INT64']:
366 inner_init = str(self.default) + 'll'
368 inner_init = str(self.default)
374 if self.allocation == 'STATIC':
375 if self.rules == 'REPEATED':
377 outer_init += ', '.join([inner_init] * self.max_count)
379 elif self.rules == 'OPTIONAL':
380 outer_init = 'false, ' + inner_init
382 outer_init = inner_init
383 elif self.allocation == 'POINTER':
384 if self.rules == 'REPEATED':
385 outer_init = '0, NULL'
388 elif self.allocation == 'CALLBACK':
389 if self.pbtype == 'EXTENSION':
392 outer_init = '{{NULL}, NULL}'
396 def default_decl(self, declaration_only = False):
397 '''Return definition for this field's default value.'''
398 if self.default is None:
402 default = self.get_initializer(False, True)
405 if self.pbtype == 'STRING':
406 if self.allocation != 'STATIC':
407 return None # Not implemented
408 array_decl = '[%d]' % self.max_size
409 elif self.pbtype == 'BYTES':
410 if self.allocation != 'STATIC':
411 return None # Not implemented
414 return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
416 return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
419 '''Return the #define for the tag number of this field.'''
420 identifier = '%s_%s_tag' % (self.struct_name, self.name)
421 return '#define %-40s %d\n' % (identifier, self.tag)
423 def pb_field_t(self, prev_field_name):
424 '''Return the pb_field_t initializer to use in the constant array.
425 prev_field_name is the name of the previous field or None.
428 if self.rules == 'ONEOF':
429 result = ' PB_ONEOF_FIELD(%s, ' % self.union_name
431 result = ' PB_FIELD('
433 result += '%3d, ' % self.tag
434 result += '%-8s, ' % self.pbtype
435 result += '%s, ' % self.rules
436 result += '%-8s, ' % self.allocation
437 result += '%s, ' % ("FIRST" if not prev_field_name else "OTHER")
438 result += '%s, ' % self.struct_name
439 result += '%s, ' % self.name
440 result += '%s, ' % (prev_field_name or self.name)
442 if self.pbtype == 'MESSAGE':
443 result += '&%s_fields)' % self.submsgname
444 elif self.default is None:
446 elif self.pbtype in ['BYTES', 'STRING'] and self.allocation != 'STATIC':
447 result += '0)' # Arbitrary size default values not implemented
448 elif self.rules == 'OPTEXT':
449 result += '0)' # Default value for extensions is not implemented
451 result += '&%s_default)' % (self.struct_name + self.name)
455 def largest_field_value(self):
456 '''Determine if this field needs 16bit or 32bit pb_field_t structure to compile properly.
457 Returns numeric value or a C-expression for assert.'''
458 if self.pbtype == 'MESSAGE':
459 if self.rules == 'REPEATED' and self.allocation == 'STATIC':
460 return 'pb_membersize(%s, %s[0])' % (self.struct_name, self.name)
461 elif self.rules == 'ONEOF':
462 return 'pb_membersize(%s, %s.%s)' % (self.struct_name, self.union_name, self.name)
464 return 'pb_membersize(%s, %s)' % (self.struct_name, self.name)
466 return max(self.tag, self.max_size, self.max_count)
468 def encoded_size(self, allmsgs):
469 '''Return the maximum size that this field can take when encoded,
470 including the field tag. If the size cannot be determined, returns
473 if self.allocation != 'STATIC':
476 if self.pbtype == 'MESSAGE':
478 if msg.name == self.submsgname:
479 encsize = msg.encoded_size(allmsgs)
481 return None # Submessage size is indeterminate
483 # Include submessage length prefix
484 encsize += varint_max_size(encsize.upperlimit())
487 # Submessage cannot be found, this currently occurs when
488 # the submessage type is defined in a different file.
489 # Instead of direct numeric value, reference the size that
490 # has been #defined in the other file.
491 encsize = EncodedSize(self.submsgname + 'size')
493 # We will have to make a conservative assumption on the length
494 # prefix size, though.
497 elif self.enc_size is None:
498 raise RuntimeError("Could not determine encoded size for %s.%s"
499 % (self.struct_name, self.name))
501 encsize = EncodedSize(self.enc_size)
503 encsize += varint_max_size(self.tag << 3) # Tag + wire type
505 if self.rules == 'REPEATED':
506 # Decoders must be always able to handle unpacked arrays.
507 # Therefore we have to reserve space for it, even though
508 # we emit packed arrays ourselves.
509 encsize *= self.max_count
514 class ExtensionRange(Field):
515 def __init__(self, struct_name, range_start, field_options):
516 '''Implements a special pb_extension_t* field in an extensible message
517 structure. The range_start signifies the index at which the extensions
518 start. Not necessarily all tags above this are extensions, it is merely
519 a speed optimization.
521 self.tag = range_start
522 self.struct_name = struct_name
523 self.name = 'extensions'
524 self.pbtype = 'EXTENSION'
525 self.rules = 'OPTIONAL'
526 self.allocation = 'CALLBACK'
527 self.ctype = 'pb_extension_t'
534 return ' pb_extension_t *extensions;'
542 def encoded_size(self, allmsgs):
543 # We exclude extensions from the count, because they cannot be known
544 # until runtime. Other option would be to return None here, but this
545 # way the value remains useful if extensions are not used.
546 return EncodedSize(0)
548 class ExtensionField(Field):
549 def __init__(self, struct_name, desc, field_options):
550 self.fullname = struct_name + desc.name
551 self.extendee_name = names_from_type_name(desc.extendee)
552 Field.__init__(self, self.fullname + 'struct', desc, field_options)
554 if self.rules != 'OPTIONAL':
558 self.rules = 'OPTEXT'
561 '''Return the #define for the tag number of this field.'''
562 identifier = '%s_tag' % self.fullname
563 return '#define %-40s %d\n' % (identifier, self.tag)
565 def extension_decl(self):
566 '''Declaration of the extension type in the .pb.h file'''
568 msg = '/* Extension field %s was skipped because only "optional"\n' % self.fullname
569 msg +=' type of extension fields is currently supported. */\n'
572 return ('extern const pb_extension_type_t %s; /* field type: %s */\n' %
573 (self.fullname, str(self).strip()))
575 def extension_def(self):
576 '''Definition of the extension type in the .pb.c file'''
581 result = 'typedef struct {\n'
583 result += '\n} %s;\n\n' % self.struct_name
584 result += ('static const pb_field_t %s_field = \n %s;\n\n' %
585 (self.fullname, self.pb_field_t(None)))
586 result += 'const pb_extension_type_t %s = {\n' % self.fullname
589 result += ' &%s_field\n' % self.fullname
594 # ---------------------------------------------------------------------------
595 # Generation of oneofs (unions)
596 # ---------------------------------------------------------------------------
599 def __init__(self, struct_name, oneof_desc):
600 self.struct_name = struct_name
601 self.name = oneof_desc.name
604 self.allocation = 'ONEOF'
608 def add_field(self, field):
609 if field.allocation == 'CALLBACK':
610 raise Exception("Callback fields inside of oneof are not supported"
611 + " (field %s)" % field.name)
613 field.union_name = self.name
614 field.rules = 'ONEOF'
615 self.fields.append(field)
616 self.fields.sort(key = lambda f: f.tag)
618 # Sort by the lowest tag number inside union
619 self.tag = min([f.tag for f in self.fields])
621 def __cmp__(self, other):
622 return cmp(self.tag, other.tag)
627 result += ' pb_size_t which_' + self.name + ";\n"
628 result += ' union {\n'
629 for f in self.fields:
630 result += ' ' + str(f).replace('\n', '\n ') + '\n'
631 result += ' } ' + self.name + ';'
635 return ''.join([f.types() for f in self.fields])
637 def get_dependencies(self):
639 for f in self.fields:
640 deps += f.get_dependencies()
643 def get_initializer(self, null_init):
644 return '0, {' + self.fields[0].get_initializer(null_init) + '}'
646 def default_decl(self, declaration_only = False):
650 return '\n'.join([f.tags() for f in self.fields])
652 def pb_field_t(self, prev_field_name):
653 result = ',\n'.join([f.pb_field_t(prev_field_name) for f in self.fields])
656 def largest_field_value(self):
657 return max([f.largest_field_value() for f in self.fields])
659 def encoded_size(self, allmsgs):
660 largest = EncodedSize(0)
661 for f in self.fields:
662 size = f.encoded_size(allmsgs)
666 return None # Cannot resolve maximum of symbols
667 elif size.value > largest.value:
672 # ---------------------------------------------------------------------------
673 # Generation of messages (structures)
674 # ---------------------------------------------------------------------------
678 def __init__(self, names, desc, message_options):
684 if message_options.msgid:
685 self.msgid = message_options.msgid
687 if hasattr(desc, 'oneof_decl'):
688 for i, f in enumerate(desc.oneof_decl):
689 oneof_options = get_nanopb_suboptions(desc, message_options, self.name + f.name)
690 if oneof_options.no_unions:
691 no_unions.append(i) # No union, but add fields normally
692 elif oneof_options.type == nanopb_pb2.FT_IGNORE:
693 pass # No union and skip fields also
695 oneof = OneOf(self.name, f)
696 self.oneofs[i] = oneof
697 self.fields.append(oneof)
700 field_options = get_nanopb_suboptions(f, message_options, self.name + f.name)
701 if field_options.type == nanopb_pb2.FT_IGNORE:
704 field = Field(self.name, f, field_options)
705 if (hasattr(f, 'oneof_index') and
706 f.HasField('oneof_index') and
707 f.oneof_index not in no_unions):
708 if f.oneof_index in self.oneofs:
709 self.oneofs[f.oneof_index].add_field(field)
711 self.fields.append(field)
713 if len(desc.extension_range) > 0:
714 field_options = get_nanopb_suboptions(desc, message_options, self.name + 'extensions')
715 range_start = min([r.start for r in desc.extension_range])
716 if field_options.type != nanopb_pb2.FT_IGNORE:
717 self.fields.append(ExtensionRange(self.name, range_start, field_options))
719 self.packed = message_options.packed_struct
720 self.ordered_fields = self.fields[:]
721 self.ordered_fields.sort()
723 def get_dependencies(self):
724 '''Get list of type names that this structure refers to.'''
726 for f in self.fields:
727 deps += f.get_dependencies()
731 result = 'typedef struct _%s {\n' % self.name
733 if not self.ordered_fields:
734 # Empty structs are not allowed in C standard.
735 # Therefore add a dummy field if an empty message occurs.
736 result += ' uint8_t dummy_field;'
738 result += '\n'.join([str(f) for f in self.ordered_fields])
742 result += ' pb_packed'
744 result += ' %s;' % self.name
747 result = 'PB_PACKED_STRUCT_START\n' + result
748 result += '\nPB_PACKED_STRUCT_END'
753 return ''.join([f.types() for f in self.fields])
755 def get_initializer(self, null_init):
756 if not self.ordered_fields:
760 for field in self.ordered_fields:
761 parts.append(field.get_initializer(null_init))
762 return '{' + ', '.join(parts) + '}'
764 def default_decl(self, declaration_only = False):
766 for field in self.fields:
767 default = field.default_decl(declaration_only)
768 if default is not None:
769 result += default + '\n'
772 def count_required_fields(self):
773 '''Returns number of required fields inside this message'''
775 for f in self.fields:
776 if not isinstance(f, OneOf):
777 if f.rules == 'REQUIRED':
781 def count_all_fields(self):
783 for f in self.fields:
784 if isinstance(f, OneOf):
785 count += len(f.fields)
790 def fields_declaration(self):
791 result = 'extern const pb_field_t %s_fields[%d];' % (self.name, self.count_all_fields() + 1)
794 def fields_definition(self):
795 result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, self.count_all_fields() + 1)
798 for field in self.ordered_fields:
799 result += field.pb_field_t(prev)
801 if isinstance(field, OneOf):
802 prev = field.name + '.' + field.fields[-1].name
806 result += ' PB_LAST_FIELD\n};'
809 def encoded_size(self, allmsgs):
810 '''Return the maximum size that this message can take when encoded.
811 If the size cannot be determined, returns None.
813 size = EncodedSize(0)
814 for field in self.fields:
815 fsize = field.encoded_size(allmsgs)
823 # ---------------------------------------------------------------------------
824 # Processing of entire .proto files
825 # ---------------------------------------------------------------------------
828 def iterate_messages(desc, names = Names()):
829 '''Recursively find all messages. For each, yield name, DescriptorProto.'''
830 if hasattr(desc, 'message_type'):
831 submsgs = desc.message_type
833 submsgs = desc.nested_type
835 for submsg in submsgs:
836 sub_names = names + submsg.name
837 yield sub_names, submsg
839 for x in iterate_messages(submsg, sub_names):
842 def iterate_extensions(desc, names = Names()):
843 '''Recursively find all extensions.
844 For each, yield name, FieldDescriptorProto.
846 for extension in desc.extension:
847 yield names, extension
849 for subname, subdesc in iterate_messages(desc, names):
850 for extension in subdesc.extension:
851 yield subname, extension
853 def parse_file(fdesc, file_options):
854 '''Takes a FileDescriptorProto and returns tuple (enums, messages, extensions).'''
861 base_name = Names(fdesc.package.split('.'))
865 for enum in fdesc.enum_type:
866 enum_options = get_nanopb_suboptions(enum, file_options, base_name + enum.name)
867 enums.append(Enum(base_name, enum, enum_options))
869 for names, message in iterate_messages(fdesc, base_name):
870 message_options = get_nanopb_suboptions(message, file_options, names)
872 if message_options.skip_message:
875 messages.append(Message(names, message, message_options))
876 for enum in message.enum_type:
877 enum_options = get_nanopb_suboptions(enum, message_options, names + enum.name)
878 enums.append(Enum(names, enum, enum_options))
880 for names, extension in iterate_extensions(fdesc, base_name):
881 field_options = get_nanopb_suboptions(extension, file_options, names + extension.name)
882 if field_options.type != nanopb_pb2.FT_IGNORE:
883 extensions.append(ExtensionField(names, extension, field_options))
885 # Fix field default values where enum short names are used.
887 if not enum.options.long_names:
888 for message in messages:
889 for field in message.fields:
890 if field.default in enum.value_longnames:
891 idx = enum.value_longnames.index(field.default)
892 field.default = enum.values[idx][0]
894 return enums, messages, extensions
898 From http://code.activestate.com/recipes/577413-topological-sort/
899 This function is under the MIT license.
901 for k, v in data.items():
902 v.discard(k) # Ignore self dependencies
903 extra_items_in_deps = reduce(set.union, data.values(), set()) - set(data.keys())
904 data.update(dict([(item, set()) for item in extra_items_in_deps]))
906 ordered = set(item for item,dep in data.items() if not dep)
909 for item in sorted(ordered):
911 data = dict([(item, (dep - ordered)) for item,dep in data.items()
912 if item not in ordered])
913 assert not data, "A cyclic dependency exists amongst %r" % data
915 def sort_dependencies(messages):
916 '''Sort a list of Messages based on dependencies.'''
919 for message in messages:
920 dependencies[str(message.name)] = set(message.get_dependencies())
921 message_by_name[str(message.name)] = message
923 for msgname in toposort2(dependencies):
924 if msgname in message_by_name:
925 yield message_by_name[msgname]
927 def make_identifier(headername):
928 '''Make #ifndef identifier that contains uppercase A-Z and digits 0-9'''
930 for c in headername.upper():
937 def generate_header(dependencies, headername, enums, messages, extensions, options):
938 '''Generate content for a header file.
939 Generates strings, which should be concatenated and stored to file.
942 yield '/* Automatically generated nanopb header */\n'
943 if options.notimestamp:
944 yield '/* Generated by %s */\n\n' % (nanopb_version)
946 yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
948 symbol = make_identifier(headername)
949 yield '#ifndef PB_%s_INCLUDED\n' % symbol
950 yield '#define PB_%s_INCLUDED\n' % symbol
952 yield options.libformat % ('pb.h')
954 # no %s specified - use whatever was passed in as options.libformat
955 yield options.libformat
958 for dependency in dependencies:
959 noext = os.path.splitext(dependency)[0]
960 yield options.genformat % (noext + options.extension + '.h')
963 yield '#if PB_PROTO_HEADER_VERSION != 30\n'
964 yield '#error Regenerate this file with the current version of nanopb generator.\n'
968 yield '#ifdef __cplusplus\n'
969 yield 'extern "C" {\n'
972 yield '/* Enum definitions */\n'
974 yield str(enum) + '\n\n'
976 yield '/* Struct definitions */\n'
977 for msg in sort_dependencies(messages):
979 yield str(msg) + '\n\n'
982 yield '/* Extensions */\n'
983 for extension in extensions:
984 yield extension.extension_decl()
987 yield '/* Default values for struct fields */\n'
989 yield msg.default_decl(True)
992 yield '/* Initializer values for message structs */\n'
994 identifier = '%s_init_default' % msg.name
995 yield '#define %-40s %s\n' % (identifier, msg.get_initializer(False))
997 identifier = '%s_init_zero' % msg.name
998 yield '#define %-40s %s\n' % (identifier, msg.get_initializer(True))
1001 yield '/* Field tags (for use in manual encoding/decoding) */\n'
1002 for msg in sort_dependencies(messages):
1003 for field in msg.fields:
1005 for extension in extensions:
1006 yield extension.tags()
1009 yield '/* Struct field encoding specification for nanopb */\n'
1010 for msg in messages:
1011 yield msg.fields_declaration() + '\n'
1014 yield '/* Maximum encoded size of messages (where known) */\n'
1015 for msg in messages:
1016 msize = msg.encoded_size(messages)
1017 if msize is not None:
1018 identifier = '%s_size' % msg.name
1019 yield '#define %-40s %s\n' % (identifier, msize)
1022 yield '/* Message IDs (where set with "msgid" option) */\n'
1024 yield '#ifdef PB_MSGID\n'
1025 for msg in messages:
1026 if hasattr(msg,'msgid'):
1027 yield '#define PB_MSG_%d %s\n' % (msg.msgid, msg.name)
1030 symbol = make_identifier(headername.split('.')[0])
1031 yield '#define %s_MESSAGES \\\n' % symbol
1033 for msg in messages:
1035 msize = msg.encoded_size(messages)
1036 if msize is not None:
1038 if hasattr(msg,'msgid'):
1039 yield '\tPB_MSG(%d,%s,%s) \\\n' % (msg.msgid, m, msg.name)
1042 for msg in messages:
1043 if hasattr(msg,'msgid'):
1044 yield '#define %s_msgid %d\n' % (msg.name, msg.msgid)
1050 yield '#ifdef __cplusplus\n'
1051 yield '} /* extern "C" */\n'
1057 def generate_source(headername, enums, messages, extensions, options):
1058 '''Generate content for a source file.'''
1060 yield '/* Automatically generated nanopb constant definitions */\n'
1061 if options.notimestamp:
1062 yield '/* Generated by %s */\n\n' % (nanopb_version)
1064 yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
1065 yield options.genformat % (headername)
1068 yield '#if PB_PROTO_HEADER_VERSION != 30\n'
1069 yield '#error Regenerate this file with the current version of nanopb generator.\n'
1073 for msg in messages:
1074 yield msg.default_decl(False)
1078 for msg in messages:
1079 yield msg.fields_definition() + '\n\n'
1081 for ext in extensions:
1082 yield ext.extension_def() + '\n'
1084 # Add checks for numeric limits
1086 largest_msg = max(messages, key = lambda m: m.count_required_fields())
1087 largest_count = largest_msg.count_required_fields()
1088 if largest_count > 64:
1089 yield '\n/* Check that missing required fields will be properly detected */\n'
1090 yield '#if PB_MAX_REQUIRED_FIELDS < %d\n' % largest_count
1091 yield '#error Properly detecting missing required fields in %s requires \\\n' % largest_msg.name
1092 yield ' setting PB_MAX_REQUIRED_FIELDS to %d or more.\n' % largest_count
1098 checks_msgnames = []
1099 for msg in messages:
1100 checks_msgnames.append(msg.name)
1101 for field in msg.fields:
1102 status = field.largest_field_value()
1103 if isinstance(status, (str, unicode)):
1104 checks.append(status)
1105 elif status > worst:
1107 worst_field = str(field.struct_name) + '.' + str(field.name)
1109 if worst > 255 or checks:
1110 yield '\n/* Check that field information fits in pb_field_t */\n'
1112 if worst > 65535 or checks:
1113 yield '#if !defined(PB_FIELD_32BIT)\n'
1115 yield '#error Field descriptor for %s is too large. Define PB_FIELD_32BIT to fix this.\n' % worst_field
1117 assertion = ' && '.join(str(c) + ' < 65536' for c in checks)
1118 msgs = '_'.join(str(n) for n in checks_msgnames)
1119 yield '/* If you get an error here, it means that you need to define PB_FIELD_32BIT\n'
1120 yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1122 yield ' * The reason you need to do this is that some of your messages contain tag\n'
1123 yield ' * numbers or field sizes that are larger than what can fit in 8 or 16 bit\n'
1124 yield ' * field descriptors.\n'
1126 yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_32BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1130 yield '#if !defined(PB_FIELD_16BIT) && !defined(PB_FIELD_32BIT)\n'
1132 yield '#error Field descriptor for %s is too large. Define PB_FIELD_16BIT to fix this.\n' % worst_field
1134 assertion = ' && '.join(str(c) + ' < 256' for c in checks)
1135 msgs = '_'.join(str(n) for n in checks_msgnames)
1136 yield '/* If you get an error here, it means that you need to define PB_FIELD_16BIT\n'
1137 yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1139 yield ' * The reason you need to do this is that some of your messages contain tag\n'
1140 yield ' * numbers or field sizes that are larger than what can fit in the default\n'
1141 yield ' * 8 bit descriptors.\n'
1143 yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_16BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1146 # Add check for sizeof(double)
1148 for msg in messages:
1149 for field in msg.fields:
1150 if field.ctype == 'double':
1155 yield '/* On some platforms (such as AVR), double is really float.\n'
1156 yield ' * These are not directly supported by nanopb, but see example_avr_double.\n'
1157 yield ' * To get rid of this error, remove any double fields from your .proto.\n'
1159 yield 'PB_STATIC_ASSERT(sizeof(double) == 8, DOUBLE_MUST_BE_8_BYTES)\n'
1163 # ---------------------------------------------------------------------------
1164 # Options parsing for the .proto files
1165 # ---------------------------------------------------------------------------
1167 from fnmatch import fnmatch
1169 def read_options_file(infile):
1170 '''Parse a separate options file to list:
1171 [(namemask, options), ...]
1174 data = infile.read()
1175 data = re.sub('/\*.*?\*/', '', data, flags = re.MULTILINE)
1176 data = re.sub('//.*?$', '', data, flags = re.MULTILINE)
1177 data = re.sub('#.*?$', '', data, flags = re.MULTILINE)
1178 for i, line in enumerate(data.split('\n')):
1183 parts = line.split(None, 1)
1186 sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1187 "Option lines should have space between field name and options. " +
1188 "Skipping line: '%s'\n" % line)
1191 opts = nanopb_pb2.NanoPBOptions()
1194 text_format.Merge(parts[1], opts)
1195 except Exception, e:
1196 sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1197 "Unparseable option line: '%s'. " % line +
1198 "Error: %s\n" % str(e))
1200 results.append((parts[0], opts))
1205 '''Ugly global variables, should find a good way to pass these.'''
1206 verbose_options = False
1207 separate_options = []
1208 matched_namemasks = set()
1210 def get_nanopb_suboptions(subdesc, options, name):
1211 '''Get copy of options, and merge information from subdesc.'''
1212 new_options = nanopb_pb2.NanoPBOptions()
1213 new_options.CopyFrom(options)
1215 # Handle options defined in a separate file
1216 dotname = '.'.join(name.parts)
1217 for namemask, options in Globals.separate_options:
1218 if fnmatch(dotname, namemask):
1219 Globals.matched_namemasks.add(namemask)
1220 new_options.MergeFrom(options)
1222 # Handle options defined in .proto
1223 if isinstance(subdesc.options, descriptor.FieldOptions):
1224 ext_type = nanopb_pb2.nanopb
1225 elif isinstance(subdesc.options, descriptor.FileOptions):
1226 ext_type = nanopb_pb2.nanopb_fileopt
1227 elif isinstance(subdesc.options, descriptor.MessageOptions):
1228 ext_type = nanopb_pb2.nanopb_msgopt
1229 elif isinstance(subdesc.options, descriptor.EnumOptions):
1230 ext_type = nanopb_pb2.nanopb_enumopt
1232 raise Exception("Unknown options type")
1234 if subdesc.options.HasExtension(ext_type):
1235 ext = subdesc.options.Extensions[ext_type]
1236 new_options.MergeFrom(ext)
1238 if Globals.verbose_options:
1239 sys.stderr.write("Options for " + dotname + ": ")
1240 sys.stderr.write(text_format.MessageToString(new_options) + "\n")
1245 # ---------------------------------------------------------------------------
1246 # Command line interface
1247 # ---------------------------------------------------------------------------
1251 from optparse import OptionParser
1253 optparser = OptionParser(
1254 usage = "Usage: nanopb_generator.py [options] file.pb ...",
1255 epilog = "Compile file.pb from file.proto by: 'protoc -ofile.pb file.proto'. " +
1256 "Output will be written to file.pb.h and file.pb.c.")
1257 optparser.add_option("-x", dest="exclude", metavar="FILE", action="append", default=[],
1258 help="Exclude file from generated #include list.")
1259 optparser.add_option("-e", "--extension", dest="extension", metavar="EXTENSION", default=".pb",
1260 help="Set extension to use instead of '.pb' for generated files. [default: %default]")
1261 optparser.add_option("-f", "--options-file", dest="options_file", metavar="FILE", default="%s.options",
1262 help="Set name of a separate generator options file.")
1263 optparser.add_option("-Q", "--generated-include-format", dest="genformat",
1264 metavar="FORMAT", default='#include "%s"\n',
1265 help="Set format string to use for including other .pb.h files. [default: %default]")
1266 optparser.add_option("-L", "--library-include-format", dest="libformat",
1267 metavar="FORMAT", default='#include <%s>\n',
1268 help="Set format string to use for including the nanopb pb.h header. [default: %default]")
1269 optparser.add_option("-T", "--no-timestamp", dest="notimestamp", action="store_true", default=False,
1270 help="Don't add timestamp to .pb.h and .pb.c preambles")
1271 optparser.add_option("-q", "--quiet", dest="quiet", action="store_true", default=False,
1272 help="Don't print anything except errors.")
1273 optparser.add_option("-v", "--verbose", dest="verbose", action="store_true", default=False,
1274 help="Print more information.")
1275 optparser.add_option("-s", dest="settings", metavar="OPTION:VALUE", action="append", default=[],
1276 help="Set generator option (max_size, max_count etc.).")
1278 def process_file(filename, fdesc, options):
1279 '''Process a single file.
1280 filename: The full path to the .proto or .pb source file, as string.
1281 fdesc: The loaded FileDescriptorSet, or None to read from the input file.
1282 options: Command line options as they come from OptionsParser.
1285 {'headername': Name of header file,
1286 'headerdata': Data for the .h header file,
1287 'sourcename': Name of the source code file,
1288 'sourcedata': Data for the .c source code file
1291 toplevel_options = nanopb_pb2.NanoPBOptions()
1292 for s in options.settings:
1293 text_format.Merge(s, toplevel_options)
1296 data = open(filename, 'rb').read()
1297 fdesc = descriptor.FileDescriptorSet.FromString(data).file[0]
1299 # Check if there is a separate .options file
1302 optfilename = options.options_file % os.path.splitext(filename)[0]
1304 # No %s specified, use the filename as-is
1305 optfilename = options.options_file
1308 if os.path.isfile(optfilename):
1310 sys.stderr.write('Reading options from ' + optfilename + '\n')
1312 Globals.separate_options = read_options_file(open(optfilename, "rU"))
1314 # If we are given a full filename and it does not exist, give an error.
1315 # However, don't give error when we automatically look for .options file
1316 # with the same name as .proto.
1317 if options.verbose or had_abspath:
1318 sys.stderr.write('Options file not found: ' + optfilename)
1320 Globals.separate_options = []
1322 Globals.matched_namemasks = set()
1325 file_options = get_nanopb_suboptions(fdesc, toplevel_options, Names([filename]))
1326 enums, messages, extensions = parse_file(fdesc, file_options)
1328 # Decide the file names
1329 noext = os.path.splitext(filename)[0]
1330 headername = noext + options.extension + '.h'
1331 sourcename = noext + options.extension + '.c'
1332 headerbasename = os.path.basename(headername)
1334 # List of .proto files that should not be included in the C header file
1335 # even if they are mentioned in the source .proto.
1336 excludes = ['nanopb.proto', 'google/protobuf/descriptor.proto'] + options.exclude
1337 dependencies = [d for d in fdesc.dependency if d not in excludes]
1339 headerdata = ''.join(generate_header(dependencies, headerbasename, enums,
1340 messages, extensions, options))
1342 sourcedata = ''.join(generate_source(headerbasename, enums,
1343 messages, extensions, options))
1345 # Check if there were any lines in .options that did not match a member
1346 unmatched = [n for n,o in Globals.separate_options if n not in Globals.matched_namemasks]
1347 if unmatched and not options.quiet:
1348 sys.stderr.write("Following patterns in " + optfilename + " did not match any fields: "
1349 + ', '.join(unmatched) + "\n")
1350 if not Globals.verbose_options:
1351 sys.stderr.write("Use protoc --nanopb-out=-v:. to see a list of the field names.\n")
1353 return {'headername': headername, 'headerdata': headerdata,
1354 'sourcename': sourcename, 'sourcedata': sourcedata}
1357 '''Main function when invoked directly from the command line.'''
1359 options, filenames = optparser.parse_args()
1362 optparser.print_help()
1366 options.verbose = False
1368 Globals.verbose_options = options.verbose
1370 for filename in filenames:
1371 results = process_file(filename, None, options)
1373 if not options.quiet:
1374 sys.stderr.write("Writing to " + results['headername'] + " and "
1375 + results['sourcename'] + "\n")
1377 open(results['headername'], 'w').write(results['headerdata'])
1378 open(results['sourcename'], 'w').write(results['sourcedata'])
1381 '''Main function when invoked as a protoc plugin.'''
1384 if sys.platform == "win32":
1386 # Set stdin and stdout to binary mode
1387 msvcrt.setmode(sys.stdin.fileno(), os.O_BINARY)
1388 msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)
1390 data = sys.stdin.read()
1391 request = plugin_pb2.CodeGeneratorRequest.FromString(data)
1394 # Versions of Python prior to 2.7.3 do not support unicode
1395 # input to shlex.split(). Try to convert to str if possible.
1396 params = str(request.parameter)
1397 except UnicodeEncodeError:
1398 params = request.parameter
1401 args = shlex.split(params)
1402 options, dummy = optparser.parse_args(args)
1404 Globals.verbose_options = options.verbose
1406 response = plugin_pb2.CodeGeneratorResponse()
1408 for filename in request.file_to_generate:
1409 for fdesc in request.proto_file:
1410 if fdesc.name == filename:
1411 results = process_file(filename, fdesc, options)
1413 f = response.file.add()
1414 f.name = results['headername']
1415 f.content = results['headerdata']
1417 f = response.file.add()
1418 f.name = results['sourcename']
1419 f.content = results['sourcedata']
1421 sys.stdout.write(response.SerializeToString())
1423 if __name__ == '__main__':
1424 # Check if we are running as a plugin under protoc
1425 if 'protoc-gen-' in sys.argv[0] or '--protoc-plugin' in sys.argv: