1#!/usr/bin/env python
2
3from __future__ import unicode_literals
4
5'''Generate header file for nanopb from a ProtoBuf FileDescriptorSet.'''
6nanopb_version = "nanopb-0.3.9.1"
7
8import sys
9import re
10import codecs
11from functools import reduce
12
13try:
14    # Add some dummy imports to keep packaging tools happy.
15    import google, distutils.util # bbfreeze seems to need these
16    import pkg_resources # pyinstaller / protobuf 2.5 seem to need these
17except:
18    # Don't care, we will error out later if it is actually important.
19    pass
20
21try:
22    import google.protobuf.text_format as text_format
23    import google.protobuf.descriptor_pb2 as descriptor
24except:
25    sys.stderr.write('''
26         *************************************************************
27         *** Could not import the Google protobuf Python libraries ***
28         *** Try installing package 'python-protobuf' or similar.  ***
29         *************************************************************
30    ''' + '\n')
31    raise
32
33try:
34    import proto.nanopb_pb2 as nanopb_pb2
35    import proto.plugin_pb2 as plugin_pb2
36except TypeError:
37    sys.stderr.write('''
38         ****************************************************************************
39         *** Got TypeError when importing the protocol definitions for generator. ***
40         *** This usually means that the protoc in your path doesn't match the    ***
41         *** Python protobuf library version.                                     ***
42         ***                                                                      ***
43         *** Please check the output of the following commands:                   ***
44         *** which protoc                                                         ***
45         *** protoc --version                                                     ***
46         *** python -c 'import google.protobuf; print(google.protobuf.__file__)'  ***
47         ****************************************************************************
48    ''' + '\n')
49    raise
50except:
51    sys.stderr.write('''
52         ********************************************************************
53         *** Failed to import the protocol definitions for generator.     ***
54         *** You have to run 'make' in the nanopb/generator/proto folder. ***
55         ********************************************************************
56    ''' + '\n')
57    raise
58
59# ---------------------------------------------------------------------------
60#                     Generation of single fields
61# ---------------------------------------------------------------------------
62
63import time
64import os.path
65
66# Values are tuple (c type, pb type, encoded size, int_size_allowed)
67FieldD = descriptor.FieldDescriptorProto
68datatypes = {
69    FieldD.TYPE_BOOL:       ('bool',     'BOOL',        1,  False),
70    FieldD.TYPE_DOUBLE:     ('double',   'DOUBLE',      8,  False),
71    FieldD.TYPE_FIXED32:    ('uint32_t', 'FIXED32',     4,  False),
72    FieldD.TYPE_FIXED64:    ('uint64_t', 'FIXED64',     8,  False),
73    FieldD.TYPE_FLOAT:      ('float',    'FLOAT',       4,  False),
74    FieldD.TYPE_INT32:      ('int32_t',  'INT32',      10,  True),
75    FieldD.TYPE_INT64:      ('int64_t',  'INT64',      10,  True),
76    FieldD.TYPE_SFIXED32:   ('int32_t',  'SFIXED32',    4,  False),
77    FieldD.TYPE_SFIXED64:   ('int64_t',  'SFIXED64',    8,  False),
78    FieldD.TYPE_SINT32:     ('int32_t',  'SINT32',      5,  True),
79    FieldD.TYPE_SINT64:     ('int64_t',  'SINT64',     10,  True),
80    FieldD.TYPE_UINT32:     ('uint32_t', 'UINT32',      5,  True),
81    FieldD.TYPE_UINT64:     ('uint64_t', 'UINT64',     10,  True)
82}
83
84# Integer size overrides (from .proto settings)
85intsizes = {
86    nanopb_pb2.IS_8:     'int8_t',
87    nanopb_pb2.IS_16:    'int16_t',
88    nanopb_pb2.IS_32:    'int32_t',
89    nanopb_pb2.IS_64:    'int64_t',
90}
91
92# String types (for python 2 / python 3 compatibility)
93try:
94    strtypes = (unicode, str)
95except NameError:
96    strtypes = (str, )
97
98class Names:
99    '''Keeps a set of nested names and formats them to C identifier.'''
100    def __init__(self, parts = ()):
101        if isinstance(parts, Names):
102            parts = parts.parts
103        self.parts = tuple(parts)
104
105    def __str__(self):
106        return '_'.join(self.parts)
107
108    def __add__(self, other):
109        if isinstance(other, strtypes):
110            return Names(self.parts + (other,))
111        elif isinstance(other, tuple):
112            return Names(self.parts + other)
113        else:
114            raise ValueError("Name parts should be of type str")
115
116    def __eq__(self, other):
117        return isinstance(other, Names) and self.parts == other.parts
118
119def names_from_type_name(type_name):
120    '''Parse Names() from FieldDescriptorProto type_name'''
121    if type_name[0] != '.':
122        raise NotImplementedError("Lookup of non-absolute type names is not supported")
123    return Names(type_name[1:].split('.'))
124
125def varint_max_size(max_value):
126    '''Returns the maximum number of bytes a varint can take when encoded.'''
127    if max_value < 0:
128        max_value = 2**64 - max_value
129    for i in range(1, 11):
130        if (max_value >> (i * 7)) == 0:
131            return i
132    raise ValueError("Value too large for varint: " + str(max_value))
133
134assert varint_max_size(-1) == 10
135assert varint_max_size(0) == 1
136assert varint_max_size(127) == 1
137assert varint_max_size(128) == 2
138
139class EncodedSize:
140    '''Class used to represent the encoded size of a field or a message.
141    Consists of a combination of symbolic sizes and integer sizes.'''
142    def __init__(self, value = 0, symbols = []):
143        if isinstance(value, EncodedSize):
144            self.value = value.value
145            self.symbols = value.symbols
146        elif isinstance(value, strtypes + (Names,)):
147            self.symbols = [str(value)]
148            self.value = 0
149        else:
150            self.value = value
151            self.symbols = symbols
152
153    def __add__(self, other):
154        if isinstance(other, int):
155            return EncodedSize(self.value + other, self.symbols)
156        elif isinstance(other, strtypes + (Names,)):
157            return EncodedSize(self.value, self.symbols + [str(other)])
158        elif isinstance(other, EncodedSize):
159            return EncodedSize(self.value + other.value, self.symbols + other.symbols)
160        else:
161            raise ValueError("Cannot add size: " + repr(other))
162
163    def __mul__(self, other):
164        if isinstance(other, int):
165            return EncodedSize(self.value * other, [str(other) + '*' + s for s in self.symbols])
166        else:
167            raise ValueError("Cannot multiply size: " + repr(other))
168
169    def __str__(self):
170        if not self.symbols:
171            return str(self.value)
172        else:
173            return '(' + str(self.value) + ' + ' + ' + '.join(self.symbols) + ')'
174
175    def upperlimit(self):
176        if not self.symbols:
177            return self.value
178        else:
179            return 2**32 - 1
180
181class Enum:
182    def __init__(self, names, desc, enum_options):
183        '''desc is EnumDescriptorProto'''
184
185        self.options = enum_options
186        self.names = names + desc.name
187
188        if enum_options.long_names:
189            self.values = [(self.names + x.name, x.number) for x in desc.value]
190        else:
191            self.values = [(names + x.name, x.number) for x in desc.value]
192
193        self.value_longnames = [self.names + x.name for x in desc.value]
194        self.packed = enum_options.packed_enum
195
196    def has_negative(self):
197        for n, v in self.values:
198            if v < 0:
199                return True
200        return False
201
202    def encoded_size(self):
203        return max([varint_max_size(v) for n,v in self.values])
204
205    def __str__(self):
206        result = 'typedef enum _%s {\n' % self.names
207        result += ',\n'.join(["    %s = %d" % x for x in self.values])
208        result += '\n}'
209
210        if self.packed:
211            result += ' pb_packed'
212
213        result += ' %s;' % self.names
214
215        result += '\n#define _%s_MIN %s' % (self.names, self.values[0][0])
216        result += '\n#define _%s_MAX %s' % (self.names, self.values[-1][0])
217        result += '\n#define _%s_ARRAYSIZE ((%s)(%s+1))' % (self.names, self.names, self.values[-1][0])
218
219        if not self.options.long_names:
220            # Define the long names always so that enum value references
221            # from other files work properly.
222            for i, x in enumerate(self.values):
223                result += '\n#define %s %s' % (self.value_longnames[i], x[0])
224
225        if self.options.enum_to_string:
226            result += '\nconst char *%s_name(%s v);\n' % (self.names, self.names)
227
228        return result
229
230    def enum_to_string_definition(self):
231        if not self.options.enum_to_string:
232            return ""
233
234        result = 'const char *%s_name(%s v) {\n' % (self.names, self.names)
235        result += '    switch (v) {\n'
236
237        for ((enumname, _), strname) in zip(self.values, self.value_longnames):
238            # Strip off the leading type name from the string value.
239            strval = str(strname)[len(str(self.names)) + 1:]
240            result += '        case %s: return "%s";\n' % (enumname, strval)
241
242        result += '    }\n'
243        result += '    return "unknown";\n'
244        result += '}\n'
245
246        return result
247
248class FieldMaxSize:
249    def __init__(self, worst = 0, checks = [], field_name = 'undefined'):
250        if isinstance(worst, list):
251            self.worst = max(i for i in worst if i is not None)
252        else:
253            self.worst = worst
254
255        self.worst_field = field_name
256        self.checks = list(checks)
257
258    def extend(self, extend, field_name = None):
259        self.worst = max(self.worst, extend.worst)
260
261        if self.worst == extend.worst:
262            self.worst_field = extend.worst_field
263
264        self.checks.extend(extend.checks)
265
266class Field:
267    def __init__(self, struct_name, desc, field_options):
268        '''desc is FieldDescriptorProto'''
269        self.tag = desc.number
270        self.struct_name = struct_name
271        self.union_name = None
272        self.name = desc.name
273        self.default = None
274        self.max_size = None
275        self.max_count = None
276        self.array_decl = ""
277        self.enc_size = None
278        self.ctype = None
279        self.fixed_count = False
280
281        if field_options.type == nanopb_pb2.FT_INLINE:
282            # Before nanopb-0.3.8, fixed length bytes arrays were specified
283            # by setting type to FT_INLINE. But to handle pointer typed fields,
284            # it makes sense to have it as a separate option.
285            field_options.type = nanopb_pb2.FT_STATIC
286            field_options.fixed_length = True
287
288        # Parse field options
289        if field_options.HasField("max_size"):
290            self.max_size = field_options.max_size
291
292        if desc.type == FieldD.TYPE_STRING and field_options.HasField("max_length"):
293            # max_length overrides max_size for strings
294            self.max_size = field_options.max_length + 1
295
296        if field_options.HasField("max_count"):
297            self.max_count = field_options.max_count
298
299        if desc.HasField('default_value'):
300            self.default = desc.default_value
301
302        # Check field rules, i.e. required/optional/repeated.
303        can_be_static = True
304        if desc.label == FieldD.LABEL_REPEATED:
305            self.rules = 'REPEATED'
306            if self.max_count is None:
307                can_be_static = False
308            else:
309                self.array_decl = '[%d]' % self.max_count
310                self.fixed_count = field_options.fixed_count
311
312        elif field_options.proto3:
313            self.rules = 'SINGULAR'
314        elif desc.label == FieldD.LABEL_REQUIRED:
315            self.rules = 'REQUIRED'
316        elif desc.label == FieldD.LABEL_OPTIONAL:
317            self.rules = 'OPTIONAL'
318        else:
319            raise NotImplementedError(desc.label)
320
321        # Check if the field can be implemented with static allocation
322        # i.e. whether the data size is known.
323        if desc.type == FieldD.TYPE_STRING and self.max_size is None:
324            can_be_static = False
325
326        if desc.type == FieldD.TYPE_BYTES and self.max_size is None:
327            can_be_static = False
328
329        # Decide how the field data will be allocated
330        if field_options.type == nanopb_pb2.FT_DEFAULT:
331            if can_be_static:
332                field_options.type = nanopb_pb2.FT_STATIC
333            else:
334                field_options.type = nanopb_pb2.FT_CALLBACK
335
336        if field_options.type == nanopb_pb2.FT_STATIC and not can_be_static:
337            raise Exception("Field '%s' is defined as static, but max_size or "
338                            "max_count is not given." % self.name)
339
340        if field_options.fixed_count and self.max_count is None:
341            raise Exception("Field '%s' is defined as fixed count, "
342                            "but max_count is not given." % self.name)
343
344        if field_options.type == nanopb_pb2.FT_STATIC:
345            self.allocation = 'STATIC'
346        elif field_options.type == nanopb_pb2.FT_POINTER:
347            self.allocation = 'POINTER'
348        elif field_options.type == nanopb_pb2.FT_CALLBACK:
349            self.allocation = 'CALLBACK'
350        else:
351            raise NotImplementedError(field_options.type)
352
353        # Decide the C data type to use in the struct.
354        if desc.type in datatypes:
355            self.ctype, self.pbtype, self.enc_size, isa = datatypes[desc.type]
356
357            # Override the field size if user wants to use smaller integers
358            if isa and field_options.int_size != nanopb_pb2.IS_DEFAULT:
359                self.ctype = intsizes[field_options.int_size]
360                if desc.type == FieldD.TYPE_UINT32 or desc.type == FieldD.TYPE_UINT64:
361                    self.ctype = 'u' + self.ctype;
362        elif desc.type == FieldD.TYPE_ENUM:
363            self.pbtype = 'ENUM'
364            self.ctype = names_from_type_name(desc.type_name)
365            if self.default is not None:
366                self.default = self.ctype + self.default
367            self.enc_size = None # Needs to be filled in when enum values are known
368        elif desc.type == FieldD.TYPE_STRING:
369            self.pbtype = 'STRING'
370            self.ctype = 'char'
371            if self.allocation == 'STATIC':
372                self.ctype = 'char'
373                self.array_decl += '[%d]' % self.max_size
374                self.enc_size = varint_max_size(self.max_size) + self.max_size
375        elif desc.type == FieldD.TYPE_BYTES:
376            if field_options.fixed_length:
377                self.pbtype = 'FIXED_LENGTH_BYTES'
378
379                if self.max_size is None:
380                    raise Exception("Field '%s' is defined as fixed length, "
381                                    "but max_size is not given." % self.name)
382
383                self.enc_size = varint_max_size(self.max_size) + self.max_size
384                self.ctype = 'pb_byte_t'
385                self.array_decl += '[%d]' % self.max_size
386            else:
387                self.pbtype = 'BYTES'
388                self.ctype = 'pb_bytes_array_t'
389                if self.allocation == 'STATIC':
390                    self.ctype = self.struct_name + self.name + 't'
391                    self.enc_size = varint_max_size(self.max_size) + self.max_size
392        elif desc.type == FieldD.TYPE_MESSAGE:
393            self.pbtype = 'MESSAGE'
394            self.ctype = self.submsgname = names_from_type_name(desc.type_name)
395            self.enc_size = None # Needs to be filled in after the message type is available
396        else:
397            raise NotImplementedError(desc.type)
398
399    def __lt__(self, other):
400        return self.tag < other.tag
401
402    def __str__(self):
403        result = ''
404        if self.allocation == 'POINTER':
405            if self.rules == 'REPEATED':
406                result += '    pb_size_t ' + self.name + '_count;\n'
407
408            if self.pbtype == 'MESSAGE':
409                # Use struct definition, so recursive submessages are possible
410                result += '    struct _%s *%s;' % (self.ctype, self.name)
411            elif self.pbtype == 'FIXED_LENGTH_BYTES':
412                # Pointer to fixed size array
413                result += '    %s (*%s)%s;' % (self.ctype, self.name, self.array_decl)
414            elif self.rules == 'REPEATED' and self.pbtype in ['STRING', 'BYTES']:
415                # String/bytes arrays need to be defined as pointers to pointers
416                result += '    %s **%s;' % (self.ctype, self.name)
417            else:
418                result += '    %s *%s;' % (self.ctype, self.name)
419        elif self.allocation == 'CALLBACK':
420            result += '    pb_callback_t %s;' % self.name
421        else:
422            if self.rules == 'OPTIONAL' and self.allocation == 'STATIC':
423                result += '    bool has_' + self.name + ';\n'
424            elif (self.rules == 'REPEATED' and
425                  self.allocation == 'STATIC' and
426                  not self.fixed_count):
427                result += '    pb_size_t ' + self.name + '_count;\n'
428            result += '    %s %s%s;' % (self.ctype, self.name, self.array_decl)
429        return result
430
431    def types(self):
432        '''Return definitions for any special types this field might need.'''
433        if self.pbtype == 'BYTES' and self.allocation == 'STATIC':
434            result = 'typedef PB_BYTES_ARRAY_T(%d) %s;\n' % (self.max_size, self.ctype)
435        else:
436            result = ''
437        return result
438
439    def get_dependencies(self):
440        '''Get list of type names used by this field.'''
441        if self.allocation == 'STATIC':
442            return [str(self.ctype)]
443        else:
444            return []
445
446    def get_initializer(self, null_init, inner_init_only = False):
447        '''Return literal expression for this field's default value.
448        null_init: If True, initialize to a 0 value instead of default from .proto
449        inner_init_only: If True, exclude initialization for any count/has fields
450        '''
451
452        inner_init = None
453        if self.pbtype == 'MESSAGE':
454            if null_init:
455                inner_init = '%s_init_zero' % self.ctype
456            else:
457                inner_init = '%s_init_default' % self.ctype
458        elif self.default is None or null_init:
459            if self.pbtype == 'STRING':
460                inner_init = '""'
461            elif self.pbtype == 'BYTES':
462                inner_init = '{0, {0}}'
463            elif self.pbtype == 'FIXED_LENGTH_BYTES':
464                inner_init = '{0}'
465            elif self.pbtype in ('ENUM', 'UENUM'):
466                inner_init = '_%s_MIN' % self.ctype
467            else:
468                inner_init = '0'
469        else:
470            if self.pbtype == 'STRING':
471                data = codecs.escape_encode(self.default.encode('utf-8'))[0]
472                inner_init = '"' + data.decode('ascii') + '"'
473            elif self.pbtype == 'BYTES':
474                data = codecs.escape_decode(self.default)[0]
475                data = ["0x%02x" % c for c in bytearray(data)]
476                if len(data) == 0:
477                    inner_init = '{0, {0}}'
478                else:
479                    inner_init = '{%d, {%s}}' % (len(data), ','.join(data))
480            elif self.pbtype == 'FIXED_LENGTH_BYTES':
481                data = codecs.escape_decode(self.default)[0]
482                data = ["0x%02x" % c for c in bytearray(data)]
483                if len(data) == 0:
484                    inner_init = '{0}'
485                else:
486                    inner_init = '{%s}' % ','.join(data)
487            elif self.pbtype in ['FIXED32', 'UINT32']:
488                inner_init = str(self.default) + 'u'
489            elif self.pbtype in ['FIXED64', 'UINT64']:
490                inner_init = str(self.default) + 'ull'
491            elif self.pbtype in ['SFIXED64', 'INT64']:
492                inner_init = str(self.default) + 'll'
493            else:
494                inner_init = str(self.default)
495
496        if inner_init_only:
497            return inner_init
498
499        outer_init = None
500        if self.allocation == 'STATIC':
501            if self.rules == 'REPEATED':
502                outer_init = ''
503                if not self.fixed_count:
504                    outer_init += '0, '
505                outer_init += '{'
506                outer_init += ', '.join([inner_init] * self.max_count)
507                outer_init += '}'
508            elif self.rules == 'OPTIONAL':
509                outer_init = 'false, ' + inner_init
510            else:
511                outer_init = inner_init
512        elif self.allocation == 'POINTER':
513            if self.rules == 'REPEATED':
514                outer_init = '0, NULL'
515            else:
516                outer_init = 'NULL'
517        elif self.allocation == 'CALLBACK':
518            if self.pbtype == 'EXTENSION':
519                outer_init = 'NULL'
520            else:
521                outer_init = '{{NULL}, NULL}'
522
523        return outer_init
524
525    def default_decl(self, declaration_only = False):
526        '''Return definition for this field's default value.'''
527        if self.default is None:
528            return None
529
530        ctype = self.ctype
531        default = self.get_initializer(False, True)
532        array_decl = ''
533
534        if self.pbtype == 'STRING':
535            if self.allocation != 'STATIC':
536                return None # Not implemented
537            array_decl = '[%d]' % self.max_size
538        elif self.pbtype == 'BYTES':
539            if self.allocation != 'STATIC':
540                return None # Not implemented
541        elif self.pbtype == 'FIXED_LENGTH_BYTES':
542            if self.allocation != 'STATIC':
543                return None # Not implemented
544            array_decl = '[%d]' % self.max_size
545
546        if declaration_only:
547            return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
548        else:
549            return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
550
551    def tags(self):
552        '''Return the #define for the tag number of this field.'''
553        identifier = '%s_%s_tag' % (self.struct_name, self.name)
554        return '#define %-40s %d\n' % (identifier, self.tag)
555
556    def pb_field_t(self, prev_field_name, union_index = None):
557        '''Return the pb_field_t initializer to use in the constant array.
558        prev_field_name is the name of the previous field or None. For OneOf
559        unions, union_index is the index of this field inside the OneOf.
560        '''
561
562        if self.rules == 'ONEOF':
563            if self.anonymous:
564                result = '    PB_ANONYMOUS_ONEOF_FIELD(%s, ' % self.union_name
565            else:
566                result = '    PB_ONEOF_FIELD(%s, ' % self.union_name
567        elif self.fixed_count:
568            result = '    PB_REPEATED_FIXED_COUNT('
569        else:
570            result = '    PB_FIELD('
571
572        result += '%3d, ' % self.tag
573        result += '%-8s, ' % self.pbtype
574        if not self.fixed_count:
575            result += '%s, ' % self.rules
576            result += '%-8s, ' % self.allocation
577
578        if union_index is not None and union_index > 0:
579            result += 'UNION, '
580        elif prev_field_name is None:
581            result += 'FIRST, '
582        else:
583            result += 'OTHER, '
584
585        result += '%s, ' % self.struct_name
586        result += '%s, ' % self.name
587        result += '%s, ' % (prev_field_name or self.name)
588
589        if self.pbtype == 'MESSAGE':
590            result += '&%s_fields)' % self.submsgname
591        elif self.default is None:
592            result += '0)'
593        elif self.pbtype in ['BYTES', 'STRING', 'FIXED_LENGTH_BYTES'] and self.allocation != 'STATIC':
594            result += '0)' # Arbitrary size default values not implemented
595        elif self.rules == 'OPTEXT':
596            result += '0)' # Default value for extensions is not implemented
597        else:
598            result += '&%s_default)' % (self.struct_name + self.name)
599
600        return result
601
602    def get_last_field_name(self):
603        return self.name
604
605    def largest_field_value(self):
606        '''Determine if this field needs 16bit or 32bit pb_field_t structure to compile properly.
607        Returns numeric value or a C-expression for assert.'''
608        check = []
609        if self.pbtype == 'MESSAGE' and self.allocation == 'STATIC':
610            if self.rules == 'REPEATED':
611                check.append('pb_membersize(%s, %s[0])' % (self.struct_name, self.name))
612            elif self.rules == 'ONEOF':
613                if self.anonymous:
614                    check.append('pb_membersize(%s, %s)' % (self.struct_name, self.name))
615                else:
616                    check.append('pb_membersize(%s, %s.%s)' % (self.struct_name, self.union_name, self.name))
617            else:
618                check.append('pb_membersize(%s, %s)' % (self.struct_name, self.name))
619        elif self.pbtype == 'BYTES' and self.allocation == 'STATIC':
620            if self.max_size > 251:
621                check.append('pb_membersize(%s, %s)' % (self.struct_name, self.name))
622
623        return FieldMaxSize([self.tag, self.max_size, self.max_count],
624                            check,
625                            ('%s.%s' % (self.struct_name, self.name)))
626
627    def encoded_size(self, dependencies):
628        '''Return the maximum size that this field can take when encoded,
629        including the field tag. If the size cannot be determined, returns
630        None.'''
631
632        if self.allocation != 'STATIC':
633            return None
634
635        if self.pbtype == 'MESSAGE':
636            encsize = None
637            if str(self.submsgname) in dependencies:
638                submsg = dependencies[str(self.submsgname)]
639                encsize = submsg.encoded_size(dependencies)
640                if encsize is not None:
641                    # Include submessage length prefix
642                    encsize += varint_max_size(encsize.upperlimit())
643
644            if encsize is None:
645                # Submessage or its size cannot be found.
646                # This can occur if submessage is defined in different
647                # file, and it or its .options could not be found.
648                # Instead of direct numeric value, reference the size that
649                # has been #defined in the other file.
650                encsize = EncodedSize(self.submsgname + 'size')
651
652                # We will have to make a conservative assumption on the length
653                # prefix size, though.
654                encsize += 5
655
656        elif self.pbtype in ['ENUM', 'UENUM']:
657            if str(self.ctype) in dependencies:
658                enumtype = dependencies[str(self.ctype)]
659                encsize = enumtype.encoded_size()
660            else:
661                # Conservative assumption
662                encsize = 10
663
664        elif self.enc_size is None:
665            raise RuntimeError("Could not determine encoded size for %s.%s"
666                               % (self.struct_name, self.name))
667        else:
668            encsize = EncodedSize(self.enc_size)
669
670        encsize += varint_max_size(self.tag << 3) # Tag + wire type
671
672        if self.rules == 'REPEATED':
673            # Decoders must be always able to handle unpacked arrays.
674            # Therefore we have to reserve space for it, even though
675            # we emit packed arrays ourselves. For length of 1, packed
676            # arrays are larger however so we need to add allowance
677            # for the length byte.
678            encsize *= self.max_count
679
680            if self.max_count == 1:
681                encsize += 1
682
683        return encsize
684
685
686class ExtensionRange(Field):
687    def __init__(self, struct_name, range_start, field_options):
688        '''Implements a special pb_extension_t* field in an extensible message
689        structure. The range_start signifies the index at which the extensions
690        start. Not necessarily all tags above this are extensions, it is merely
691        a speed optimization.
692        '''
693        self.tag = range_start
694        self.struct_name = struct_name
695        self.name = 'extensions'
696        self.pbtype = 'EXTENSION'
697        self.rules = 'OPTIONAL'
698        self.allocation = 'CALLBACK'
699        self.ctype = 'pb_extension_t'
700        self.array_decl = ''
701        self.default = None
702        self.max_size = 0
703        self.max_count = 0
704        self.fixed_count = False
705
706    def __str__(self):
707        return '    pb_extension_t *extensions;'
708
709    def types(self):
710        return ''
711
712    def tags(self):
713        return ''
714
715    def encoded_size(self, dependencies):
716        # We exclude extensions from the count, because they cannot be known
717        # until runtime. Other option would be to return None here, but this
718        # way the value remains useful if extensions are not used.
719        return EncodedSize(0)
720
721class ExtensionField(Field):
722    def __init__(self, struct_name, desc, field_options):
723        self.fullname = struct_name + desc.name
724        self.extendee_name = names_from_type_name(desc.extendee)
725        Field.__init__(self, self.fullname + 'struct', desc, field_options)
726
727        if self.rules != 'OPTIONAL':
728            self.skip = True
729        else:
730            self.skip = False
731            self.rules = 'OPTEXT'
732
733    def tags(self):
734        '''Return the #define for the tag number of this field.'''
735        identifier = '%s_tag' % self.fullname
736        return '#define %-40s %d\n' % (identifier, self.tag)
737
738    def extension_decl(self):
739        '''Declaration of the extension type in the .pb.h file'''
740        if self.skip:
741            msg = '/* Extension field %s was skipped because only "optional"\n' % self.fullname
742            msg +='   type of extension fields is currently supported. */\n'
743            return msg
744
745        return ('extern const pb_extension_type_t %s; /* field type: %s */\n' %
746            (self.fullname, str(self).strip()))
747
748    def extension_def(self):
749        '''Definition of the extension type in the .pb.c file'''
750
751        if self.skip:
752            return ''
753
754        result  = 'typedef struct {\n'
755        result += str(self)
756        result += '\n} %s;\n\n' % self.struct_name
757        result += ('static const pb_field_t %s_field = \n  %s;\n\n' %
758                    (self.fullname, self.pb_field_t(None)))
759        result += 'const pb_extension_type_t %s = {\n' % self.fullname
760        result += '    NULL,\n'
761        result += '    NULL,\n'
762        result += '    &%s_field\n' % self.fullname
763        result += '};\n'
764        return result
765
766
767# ---------------------------------------------------------------------------
768#                   Generation of oneofs (unions)
769# ---------------------------------------------------------------------------
770
771class OneOf(Field):
772    def __init__(self, struct_name, oneof_desc):
773        self.struct_name = struct_name
774        self.name = oneof_desc.name
775        self.ctype = 'union'
776        self.pbtype = 'oneof'
777        self.fields = []
778        self.allocation = 'ONEOF'
779        self.default = None
780        self.rules = 'ONEOF'
781        self.anonymous = False
782
783    def add_field(self, field):
784        if field.allocation == 'CALLBACK':
785            raise Exception("Callback fields inside of oneof are not supported"
786                            + " (field %s)" % field.name)
787
788        field.union_name = self.name
789        field.rules = 'ONEOF'
790        field.anonymous = self.anonymous
791        self.fields.append(field)
792        self.fields.sort(key = lambda f: f.tag)
793
794        # Sort by the lowest tag number inside union
795        self.tag = min([f.tag for f in self.fields])
796
797    def __str__(self):
798        result = ''
799        if self.fields:
800            result += '    pb_size_t which_' + self.name + ";\n"
801            result += '    union {\n'
802            for f in self.fields:
803                result += '    ' + str(f).replace('\n', '\n    ') + '\n'
804            if self.anonymous:
805                result += '    };'
806            else:
807                result += '    } ' + self.name + ';'
808        return result
809
810    def types(self):
811        return ''.join([f.types() for f in self.fields])
812
813    def get_dependencies(self):
814        deps = []
815        for f in self.fields:
816            deps += f.get_dependencies()
817        return deps
818
819    def get_initializer(self, null_init):
820        return '0, {' + self.fields[0].get_initializer(null_init) + '}'
821
822    def default_decl(self, declaration_only = False):
823        return None
824
825    def tags(self):
826        return ''.join([f.tags() for f in self.fields])
827
828    def pb_field_t(self, prev_field_name):
829        parts = []
830        for union_index, field in enumerate(self.fields):
831            parts.append(field.pb_field_t(prev_field_name, union_index))
832        return ',\n'.join(parts)
833
834    def get_last_field_name(self):
835        if self.anonymous:
836            return self.fields[-1].name
837        else:
838            return self.name + '.' + self.fields[-1].name
839
840    def largest_field_value(self):
841        largest = FieldMaxSize()
842        for f in self.fields:
843            largest.extend(f.largest_field_value())
844        return largest
845
846    def encoded_size(self, dependencies):
847        '''Returns the size of the largest oneof field.'''
848        largest = EncodedSize(0)
849        symbols = set()
850        for f in self.fields:
851            size = EncodedSize(f.encoded_size(dependencies))
852            if size.value is None:
853                return None
854            elif size.symbols:
855                symbols.add(EncodedSize(f.submsgname + 'size').symbols[0])
856            elif size.value > largest.value:
857                largest = size
858
859        if not symbols:
860            return largest
861
862        symbols = list(symbols)
863        symbols.append(str(largest))
864        max_size = lambda x, y: '({0} > {1} ? {0} : {1})'.format(x, y)
865        return reduce(max_size, symbols)
866
867# ---------------------------------------------------------------------------
868#                   Generation of messages (structures)
869# ---------------------------------------------------------------------------
870
871
872class Message:
873    def __init__(self, names, desc, message_options):
874        self.name = names
875        self.fields = []
876        self.oneofs = {}
877        no_unions = []
878
879        if message_options.msgid:
880            self.msgid = message_options.msgid
881
882        if hasattr(desc, 'oneof_decl'):
883            for i, f in enumerate(desc.oneof_decl):
884                oneof_options = get_nanopb_suboptions(desc, message_options, self.name + f.name)
885                if oneof_options.no_unions:
886                    no_unions.append(i) # No union, but add fields normally
887                elif oneof_options.type == nanopb_pb2.FT_IGNORE:
888                    pass # No union and skip fields also
889                else:
890                    oneof = OneOf(self.name, f)
891                    if oneof_options.anonymous_oneof:
892                        oneof.anonymous = True
893                    self.oneofs[i] = oneof
894                    self.fields.append(oneof)
895        else:
896            sys.stderr.write('Note: This Python protobuf library has no OneOf support\n')
897
898        for f in desc.field:
899            field_options = get_nanopb_suboptions(f, message_options, self.name + f.name)
900            if field_options.type == nanopb_pb2.FT_IGNORE:
901                continue
902
903            field = Field(self.name, f, field_options)
904            if (hasattr(f, 'oneof_index') and
905                f.HasField('oneof_index') and
906                f.oneof_index not in no_unions):
907                if f.oneof_index in self.oneofs:
908                    self.oneofs[f.oneof_index].add_field(field)
909            else:
910                self.fields.append(field)
911
912        if len(desc.extension_range) > 0:
913            field_options = get_nanopb_suboptions(desc, message_options, self.name + 'extensions')
914            range_start = min([r.start for r in desc.extension_range])
915            if field_options.type != nanopb_pb2.FT_IGNORE:
916                self.fields.append(ExtensionRange(self.name, range_start, field_options))
917
918        self.packed = message_options.packed_struct
919        self.ordered_fields = self.fields[:]
920        self.ordered_fields.sort()
921
922    def get_dependencies(self):
923        '''Get list of type names that this structure refers to.'''
924        deps = []
925        for f in self.fields:
926            deps += f.get_dependencies()
927        return deps
928
929    def __str__(self):
930        result = 'typedef struct _%s {\n' % self.name
931
932        if not self.ordered_fields:
933            # Empty structs are not allowed in C standard.
934            # Therefore add a dummy field if an empty message occurs.
935            result += '    char dummy_field;'
936
937        result += '\n'.join([str(f) for f in self.ordered_fields])
938        result += '\n/* @@protoc_insertion_point(struct:%s) */' % self.name
939        result += '\n}'
940
941        if self.packed:
942            result += ' pb_packed'
943
944        result += ' %s;' % self.name
945
946        if self.packed:
947            result = 'PB_PACKED_STRUCT_START\n' + result
948            result += '\nPB_PACKED_STRUCT_END'
949
950        return result
951
952    def types(self):
953        return ''.join([f.types() for f in self.fields])
954
955    def get_initializer(self, null_init):
956        if not self.ordered_fields:
957            return '{0}'
958
959        parts = []
960        for field in self.ordered_fields:
961            parts.append(field.get_initializer(null_init))
962        return '{' + ', '.join(parts) + '}'
963
964    def default_decl(self, declaration_only = False):
965        result = ""
966        for field in self.fields:
967            default = field.default_decl(declaration_only)
968            if default is not None:
969                result += default + '\n'
970        return result
971
972    def count_required_fields(self):
973        '''Returns number of required fields inside this message'''
974        count = 0
975        for f in self.fields:
976            if not isinstance(f, OneOf):
977                if f.rules == 'REQUIRED':
978                    count += 1
979        return count
980
981    def count_all_fields(self):
982        count = 0
983        for f in self.fields:
984            if isinstance(f, OneOf):
985                count += len(f.fields)
986            else:
987                count += 1
988        return count
989
990    def fields_declaration(self):
991        result = 'extern const pb_field_t %s_fields[%d];' % (self.name, self.count_all_fields() + 1)
992        return result
993
994    def fields_definition(self):
995        result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, self.count_all_fields() + 1)
996
997        prev = None
998        for field in self.ordered_fields:
999            result += field.pb_field_t(prev)
1000            result += ',\n'
1001            prev = field.get_last_field_name()
1002
1003        result += '    PB_LAST_FIELD\n};'
1004        return result
1005
1006    def encoded_size(self, dependencies):
1007        '''Return the maximum size that this message can take when encoded.
1008        If the size cannot be determined, returns None.
1009        '''
1010        size = EncodedSize(0)
1011        for field in self.fields:
1012            fsize = field.encoded_size(dependencies)
1013            if fsize is None:
1014                return None
1015            size += fsize
1016
1017        return size
1018
1019
1020# ---------------------------------------------------------------------------
1021#                    Processing of entire .proto files
1022# ---------------------------------------------------------------------------
1023
1024def iterate_messages(desc, names = Names()):
1025    '''Recursively find all messages. For each, yield name, DescriptorProto.'''
1026    if hasattr(desc, 'message_type'):
1027        submsgs = desc.message_type
1028    else:
1029        submsgs = desc.nested_type
1030
1031    for submsg in submsgs:
1032        sub_names = names + submsg.name
1033        yield sub_names, submsg
1034
1035        for x in iterate_messages(submsg, sub_names):
1036            yield x
1037
1038def iterate_extensions(desc, names = Names()):
1039    '''Recursively find all extensions.
1040    For each, yield name, FieldDescriptorProto.
1041    '''
1042    for extension in desc.extension:
1043        yield names, extension
1044
1045    for subname, subdesc in iterate_messages(desc, names):
1046        for extension in subdesc.extension:
1047            yield subname, extension
1048
1049def toposort2(data):
1050    '''Topological sort.
1051    From http://code.activestate.com/recipes/577413-topological-sort/
1052    This function is under the MIT license.
1053    '''
1054    for k, v in list(data.items()):
1055        v.discard(k) # Ignore self dependencies
1056    extra_items_in_deps = reduce(set.union, list(data.values()), set()) - set(data.keys())
1057    data.update(dict([(item, set()) for item in extra_items_in_deps]))
1058    while True:
1059        ordered = set(item for item,dep in list(data.items()) if not dep)
1060        if not ordered:
1061            break
1062        for item in sorted(ordered):
1063            yield item
1064        data = dict([(item, (dep - ordered)) for item,dep in list(data.items())
1065                if item not in ordered])
1066    assert not data, "A cyclic dependency exists amongst %r" % data
1067
1068def sort_dependencies(messages):
1069    '''Sort a list of Messages based on dependencies.'''
1070    dependencies = {}
1071    message_by_name = {}
1072    for message in messages:
1073        dependencies[str(message.name)] = set(message.get_dependencies())
1074        message_by_name[str(message.name)] = message
1075
1076    for msgname in toposort2(dependencies):
1077        if msgname in message_by_name:
1078            yield message_by_name[msgname]
1079
1080def make_identifier(headername):
1081    '''Make #ifndef identifier that contains uppercase A-Z and digits 0-9'''
1082    result = ""
1083    for c in headername.upper():
1084        if c.isalnum():
1085            result += c
1086        else:
1087            result += '_'
1088    return result
1089
1090class ProtoFile:
1091    def __init__(self, fdesc, file_options):
1092        '''Takes a FileDescriptorProto and parses it.'''
1093        self.fdesc = fdesc
1094        self.file_options = file_options
1095        self.dependencies = {}
1096        self.parse()
1097
1098        # Some of types used in this file probably come from the file itself.
1099        # Thus it has implicit dependency on itself.
1100        self.add_dependency(self)
1101
1102    def parse(self):
1103        self.enums = []
1104        self.messages = []
1105        self.extensions = []
1106
1107        if self.fdesc.package:
1108            base_name = Names(self.fdesc.package.split('.'))
1109        else:
1110            base_name = Names()
1111
1112        for enum in self.fdesc.enum_type:
1113            enum_options = get_nanopb_suboptions(enum, self.file_options, base_name + enum.name)
1114            self.enums.append(Enum(base_name, enum, enum_options))
1115
1116        for names, message in iterate_messages(self.fdesc, base_name):
1117            message_options = get_nanopb_suboptions(message, self.file_options, names)
1118
1119            if message_options.skip_message:
1120                continue
1121
1122            self.messages.append(Message(names, message, message_options))
1123            for enum in message.enum_type:
1124                enum_options = get_nanopb_suboptions(enum, message_options, names + enum.name)
1125                self.enums.append(Enum(names, enum, enum_options))
1126
1127        for names, extension in iterate_extensions(self.fdesc, base_name):
1128            field_options = get_nanopb_suboptions(extension, self.file_options, names + extension.name)
1129            if field_options.type != nanopb_pb2.FT_IGNORE:
1130                self.extensions.append(ExtensionField(names, extension, field_options))
1131
1132    def add_dependency(self, other):
1133        for enum in other.enums:
1134            self.dependencies[str(enum.names)] = enum
1135
1136        for msg in other.messages:
1137            self.dependencies[str(msg.name)] = msg
1138
1139        # Fix field default values where enum short names are used.
1140        for enum in other.enums:
1141            if not enum.options.long_names:
1142                for message in self.messages:
1143                    for field in message.fields:
1144                        if field.default in enum.value_longnames:
1145                            idx = enum.value_longnames.index(field.default)
1146                            field.default = enum.values[idx][0]
1147
1148        # Fix field data types where enums have negative values.
1149        for enum in other.enums:
1150            if not enum.has_negative():
1151                for message in self.messages:
1152                    for field in message.fields:
1153                        if field.pbtype == 'ENUM' and field.ctype == enum.names:
1154                            field.pbtype = 'UENUM'
1155
1156    def generate_header(self, includes, headername, options):
1157        '''Generate content for a header file.
1158        Generates strings, which should be concatenated and stored to file.
1159        '''
1160
1161        yield '/* Automatically generated nanopb header */\n'
1162        if options.notimestamp:
1163            yield '/* Generated by %s */\n\n' % (nanopb_version)
1164        else:
1165            yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
1166
1167        if self.fdesc.package:
1168            symbol = make_identifier(self.fdesc.package + '_' + headername)
1169        else:
1170            symbol = make_identifier(headername)
1171        yield '#ifndef PB_%s_INCLUDED\n' % symbol
1172        yield '#define PB_%s_INCLUDED\n' % symbol
1173        try:
1174            yield options.libformat % ('pb.h')
1175        except TypeError:
1176            # no %s specified - use whatever was passed in as options.libformat
1177            yield options.libformat
1178        yield '\n'
1179
1180        for incfile in includes:
1181            noext = os.path.splitext(incfile)[0]
1182            yield options.genformat % (noext + options.extension + options.header_extension)
1183            yield '\n'
1184
1185        yield '/* @@protoc_insertion_point(includes) */\n'
1186
1187        yield '#if PB_PROTO_HEADER_VERSION != 30\n'
1188        yield '#error Regenerate this file with the current version of nanopb generator.\n'
1189        yield '#endif\n'
1190        yield '\n'
1191
1192        yield '#ifdef __cplusplus\n'
1193        yield 'extern "C" {\n'
1194        yield '#endif\n\n'
1195
1196        if self.enums:
1197            yield '/* Enum definitions */\n'
1198            for enum in self.enums:
1199                yield str(enum) + '\n\n'
1200
1201        if self.messages:
1202            yield '/* Struct definitions */\n'
1203            for msg in sort_dependencies(self.messages):
1204                yield msg.types()
1205                yield str(msg) + '\n\n'
1206
1207        if self.extensions:
1208            yield '/* Extensions */\n'
1209            for extension in self.extensions:
1210                yield extension.extension_decl()
1211            yield '\n'
1212
1213        if self.messages:
1214            yield '/* Default values for struct fields */\n'
1215            for msg in self.messages:
1216                yield msg.default_decl(True)
1217            yield '\n'
1218
1219            yield '/* Initializer values for message structs */\n'
1220            for msg in self.messages:
1221                identifier = '%s_init_default' % msg.name
1222                yield '#define %-40s %s\n' % (identifier, msg.get_initializer(False))
1223            for msg in self.messages:
1224                identifier = '%s_init_zero' % msg.name
1225                yield '#define %-40s %s\n' % (identifier, msg.get_initializer(True))
1226            yield '\n'
1227
1228            yield '/* Field tags (for use in manual encoding/decoding) */\n'
1229            for msg in sort_dependencies(self.messages):
1230                for field in msg.fields:
1231                    yield field.tags()
1232            for extension in self.extensions:
1233                yield extension.tags()
1234            yield '\n'
1235
1236            yield '/* Struct field encoding specification for nanopb */\n'
1237            for msg in self.messages:
1238                yield msg.fields_declaration() + '\n'
1239            yield '\n'
1240
1241            yield '/* Maximum encoded size of messages (where known) */\n'
1242            for msg in self.messages:
1243                msize = msg.encoded_size(self.dependencies)
1244                identifier = '%s_size' % msg.name
1245                if msize is not None:
1246                    yield '#define %-40s %s\n' % (identifier, msize)
1247                else:
1248                    yield '/* %s depends on runtime parameters */\n' % identifier
1249            yield '\n'
1250
1251            yield '/* Message IDs (where set with "msgid" option) */\n'
1252
1253            yield '#ifdef PB_MSGID\n'
1254            for msg in self.messages:
1255                if hasattr(msg,'msgid'):
1256                    yield '#define PB_MSG_%d %s\n' % (msg.msgid, msg.name)
1257            yield '\n'
1258
1259            symbol = make_identifier(headername.split('.')[0])
1260            yield '#define %s_MESSAGES \\\n' % symbol
1261
1262            for msg in self.messages:
1263                m = "-1"
1264                msize = msg.encoded_size(self.dependencies)
1265                if msize is not None:
1266                    m = msize
1267                if hasattr(msg,'msgid'):
1268                    yield '\tPB_MSG(%d,%s,%s) \\\n' % (msg.msgid, m, msg.name)
1269            yield '\n'
1270
1271            for msg in self.messages:
1272                if hasattr(msg,'msgid'):
1273                    yield '#define %s_msgid %d\n' % (msg.name, msg.msgid)
1274            yield '\n'
1275
1276            yield '#endif\n\n'
1277
1278        yield '#ifdef __cplusplus\n'
1279        yield '} /* extern "C" */\n'
1280        yield '#endif\n'
1281
1282        # End of header
1283        yield '/* @@protoc_insertion_point(eof) */\n'
1284        yield '\n#endif\n'
1285
1286    def generate_source(self, headername, options):
1287        '''Generate content for a source file.'''
1288
1289        yield '/* Automatically generated nanopb constant definitions */\n'
1290        if options.notimestamp:
1291            yield '/* Generated by %s */\n\n' % (nanopb_version)
1292        else:
1293            yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
1294        yield options.genformat % (headername)
1295        yield '\n'
1296        yield '/* @@protoc_insertion_point(includes) */\n'
1297
1298        yield '#if PB_PROTO_HEADER_VERSION != 30\n'
1299        yield '#error Regenerate this file with the current version of nanopb generator.\n'
1300        yield '#endif\n'
1301        yield '\n'
1302
1303        for msg in self.messages:
1304            yield msg.default_decl(False)
1305
1306        yield '\n\n'
1307
1308        for msg in self.messages:
1309            yield msg.fields_definition() + '\n\n'
1310
1311        for ext in self.extensions:
1312            yield ext.extension_def() + '\n'
1313
1314        for enum in self.enums:
1315            yield enum.enum_to_string_definition() + '\n'
1316
1317        # Add checks for numeric limits
1318        if self.messages:
1319            largest_msg = max(self.messages, key = lambda m: m.count_required_fields())
1320            largest_count = largest_msg.count_required_fields()
1321            if largest_count > 64:
1322                yield '\n/* Check that missing required fields will be properly detected */\n'
1323                yield '#if PB_MAX_REQUIRED_FIELDS < %d\n' % largest_count
1324                yield '#error Properly detecting missing required fields in %s requires \\\n' % largest_msg.name
1325                yield '       setting PB_MAX_REQUIRED_FIELDS to %d or more.\n' % largest_count
1326                yield '#endif\n'
1327
1328        max_field = FieldMaxSize()
1329        checks_msgnames = []
1330        for msg in self.messages:
1331            checks_msgnames.append(msg.name)
1332            for field in msg.fields:
1333                max_field.extend(field.largest_field_value())
1334        for field in self.extensions:
1335            max_field.extend(field.largest_field_value())
1336
1337        worst = max_field.worst
1338        worst_field = max_field.worst_field
1339        checks = max_field.checks
1340
1341        if worst > 255 or checks:
1342            yield '\n/* Check that field information fits in pb_field_t */\n'
1343
1344            if worst > 65535 or checks:
1345                yield '#if !defined(PB_FIELD_32BIT)\n'
1346                if worst > 65535:
1347                    yield '#error Field descriptor for %s is too large. Define PB_FIELD_32BIT to fix this.\n' % worst_field
1348                else:
1349                    assertion = ' && '.join(str(c) + ' < 65536' for c in checks)
1350                    msgs = '_'.join(str(n) for n in checks_msgnames)
1351                    yield '/* If you get an error here, it means that you need to define PB_FIELD_32BIT\n'
1352                    yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1353                    yield ' * \n'
1354                    yield ' * The reason you need to do this is that some of your messages contain tag\n'
1355                    yield ' * numbers or field sizes that are larger than what can fit in 8 or 16 bit\n'
1356                    yield ' * field descriptors.\n'
1357                    yield ' */\n'
1358                    yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_32BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1359                yield '#endif\n\n'
1360
1361            if worst < 65536:
1362                yield '#if !defined(PB_FIELD_16BIT) && !defined(PB_FIELD_32BIT)\n'
1363                if worst > 255:
1364                    yield '#error Field descriptor for %s is too large. Define PB_FIELD_16BIT to fix this.\n' % worst_field
1365                else:
1366                    assertion = ' && '.join(str(c) + ' < 256' for c in checks)
1367                    msgs = '_'.join(str(n) for n in checks_msgnames)
1368                    yield '/* If you get an error here, it means that you need to define PB_FIELD_16BIT\n'
1369                    yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1370                    yield ' * \n'
1371                    yield ' * The reason you need to do this is that some of your messages contain tag\n'
1372                    yield ' * numbers or field sizes that are larger than what can fit in the default\n'
1373                    yield ' * 8 bit descriptors.\n'
1374                    yield ' */\n'
1375                    yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_16BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1376                yield '#endif\n\n'
1377
1378        # Add check for sizeof(double)
1379        has_double = False
1380        for msg in self.messages:
1381            for field in msg.fields:
1382                if field.ctype == 'double':
1383                    has_double = True
1384
1385        if has_double:
1386            yield '\n'
1387            yield '/* On some platforms (such as AVR), double is really float.\n'
1388            yield ' * These are not directly supported by nanopb, but see example_avr_double.\n'
1389            yield ' * To get rid of this error, remove any double fields from your .proto.\n'
1390            yield ' */\n'
1391            yield 'PB_STATIC_ASSERT(sizeof(double) == 8, DOUBLE_MUST_BE_8_BYTES)\n'
1392
1393        yield '\n'
1394        yield '/* @@protoc_insertion_point(eof) */\n'
1395
1396# ---------------------------------------------------------------------------
1397#                    Options parsing for the .proto files
1398# ---------------------------------------------------------------------------
1399
1400from fnmatch import fnmatch
1401
1402def read_options_file(infile):
1403    '''Parse a separate options file to list:
1404        [(namemask, options), ...]
1405    '''
1406    results = []
1407    data = infile.read()
1408    data = re.sub('/\*.*?\*/', '', data, flags = re.MULTILINE)
1409    data = re.sub('//.*?$', '', data, flags = re.MULTILINE)
1410    data = re.sub('#.*?$', '', data, flags = re.MULTILINE)
1411    for i, line in enumerate(data.split('\n')):
1412        line = line.strip()
1413        if not line:
1414            continue
1415
1416        parts = line.split(None, 1)
1417
1418        if len(parts) < 2:
1419            sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1420                             "Option lines should have space between field name and options. " +
1421                             "Skipping line: '%s'\n" % line)
1422            continue
1423
1424        opts = nanopb_pb2.NanoPBOptions()
1425
1426        try:
1427            text_format.Merge(parts[1], opts)
1428        except Exception as e:
1429            sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1430                             "Unparseable option line: '%s'. " % line +
1431                             "Error: %s\n" % str(e))
1432            continue
1433        results.append((parts[0], opts))
1434
1435    return results
1436
1437class Globals:
1438    '''Ugly global variables, should find a good way to pass these.'''
1439    verbose_options = False
1440    separate_options = []
1441    matched_namemasks = set()
1442
1443def get_nanopb_suboptions(subdesc, options, name):
1444    '''Get copy of options, and merge information from subdesc.'''
1445    new_options = nanopb_pb2.NanoPBOptions()
1446    new_options.CopyFrom(options)
1447
1448    if hasattr(subdesc, 'syntax') and subdesc.syntax == "proto3":
1449        new_options.proto3 = True
1450
1451    # Handle options defined in a separate file
1452    dotname = '.'.join(name.parts)
1453    for namemask, options in Globals.separate_options:
1454        if fnmatch(dotname, namemask):
1455            Globals.matched_namemasks.add(namemask)
1456            new_options.MergeFrom(options)
1457
1458    # Handle options defined in .proto
1459    if isinstance(subdesc.options, descriptor.FieldOptions):
1460        ext_type = nanopb_pb2.nanopb
1461    elif isinstance(subdesc.options, descriptor.FileOptions):
1462        ext_type = nanopb_pb2.nanopb_fileopt
1463    elif isinstance(subdesc.options, descriptor.MessageOptions):
1464        ext_type = nanopb_pb2.nanopb_msgopt
1465    elif isinstance(subdesc.options, descriptor.EnumOptions):
1466        ext_type = nanopb_pb2.nanopb_enumopt
1467    else:
1468        raise Exception("Unknown options type")
1469
1470    if subdesc.options.HasExtension(ext_type):
1471        ext = subdesc.options.Extensions[ext_type]
1472        new_options.MergeFrom(ext)
1473
1474    if Globals.verbose_options:
1475        sys.stderr.write("Options for " + dotname + ": ")
1476        sys.stderr.write(text_format.MessageToString(new_options) + "\n")
1477
1478    return new_options
1479
1480
1481# ---------------------------------------------------------------------------
1482#                         Command line interface
1483# ---------------------------------------------------------------------------
1484
1485import sys
1486import os.path
1487from optparse import OptionParser
1488
1489optparser = OptionParser(
1490    usage = "Usage: nanopb_generator.py [options] file.pb ...",
1491    epilog = "Compile file.pb from file.proto by: 'protoc -ofile.pb file.proto'. " +
1492             "Output will be written to file.pb.h and file.pb.c.")
1493optparser.add_option("-x", dest="exclude", metavar="FILE", action="append", default=[],
1494    help="Exclude file from generated #include list.")
1495optparser.add_option("-e", "--extension", dest="extension", metavar="EXTENSION", default=".pb",
1496    help="Set extension to use instead of '.pb' for generated files. [default: %default]")
1497optparser.add_option("-H", "--header-extension", dest="header_extension", metavar="EXTENSION", default=".h",
1498    help="Set extension to use for generated header files. [default: %default]")
1499optparser.add_option("-S", "--source-extension", dest="source_extension", metavar="EXTENSION", default=".c",
1500    help="Set extension to use for generated source files. [default: %default]")
1501optparser.add_option("-f", "--options-file", dest="options_file", metavar="FILE", default="%s.options",
1502    help="Set name of a separate generator options file.")
1503optparser.add_option("-I", "--options-path", dest="options_path", metavar="DIR",
1504    action="append", default = [],
1505    help="Search for .options files additionally in this path")
1506optparser.add_option("-D", "--output-dir", dest="output_dir",
1507                     metavar="OUTPUTDIR", default=None,
1508                     help="Output directory of .pb.h and .pb.c files")
1509optparser.add_option("-Q", "--generated-include-format", dest="genformat",
1510    metavar="FORMAT", default='#include "%s"\n',
1511    help="Set format string to use for including other .pb.h files. [default: %default]")
1512optparser.add_option("-L", "--library-include-format", dest="libformat",
1513    metavar="FORMAT", default='#include <%s>\n',
1514    help="Set format string to use for including the nanopb pb.h header. [default: %default]")
1515optparser.add_option("-T", "--no-timestamp", dest="notimestamp", action="store_true", default=False,
1516    help="Don't add timestamp to .pb.h and .pb.c preambles")
1517optparser.add_option("-q", "--quiet", dest="quiet", action="store_true", default=False,
1518    help="Don't print anything except errors.")
1519optparser.add_option("-v", "--verbose", dest="verbose", action="store_true", default=False,
1520    help="Print more information.")
1521optparser.add_option("-s", dest="settings", metavar="OPTION:VALUE", action="append", default=[],
1522    help="Set generator option (max_size, max_count etc.).")
1523
1524def parse_file(filename, fdesc, options):
1525    '''Parse a single file. Returns a ProtoFile instance.'''
1526    toplevel_options = nanopb_pb2.NanoPBOptions()
1527    for s in options.settings:
1528        text_format.Merge(s, toplevel_options)
1529
1530    if not fdesc:
1531        data = open(filename, 'rb').read()
1532        fdesc = descriptor.FileDescriptorSet.FromString(data).file[0]
1533
1534    # Check if there is a separate .options file
1535    had_abspath = False
1536    try:
1537        optfilename = options.options_file % os.path.splitext(filename)[0]
1538    except TypeError:
1539        # No %s specified, use the filename as-is
1540        optfilename = options.options_file
1541        had_abspath = True
1542
1543    paths = ['.'] + options.options_path
1544    for p in paths:
1545        if os.path.isfile(os.path.join(p, optfilename)):
1546            optfilename = os.path.join(p, optfilename)
1547            if options.verbose:
1548                sys.stderr.write('Reading options from ' + optfilename + '\n')
1549            Globals.separate_options = read_options_file(open(optfilename, "rU"))
1550            break
1551    else:
1552        # If we are given a full filename and it does not exist, give an error.
1553        # However, don't give error when we automatically look for .options file
1554        # with the same name as .proto.
1555        if options.verbose or had_abspath:
1556            sys.stderr.write('Options file not found: ' + optfilename + '\n')
1557        Globals.separate_options = []
1558
1559    Globals.matched_namemasks = set()
1560
1561    # Parse the file
1562    file_options = get_nanopb_suboptions(fdesc, toplevel_options, Names([filename]))
1563    f = ProtoFile(fdesc, file_options)
1564    f.optfilename = optfilename
1565
1566    return f
1567
1568def process_file(filename, fdesc, options, other_files = {}):
1569    '''Process a single file.
1570    filename: The full path to the .proto or .pb source file, as string.
1571    fdesc: The loaded FileDescriptorSet, or None to read from the input file.
1572    options: Command line options as they come from OptionsParser.
1573
1574    Returns a dict:
1575        {'headername': Name of header file,
1576         'headerdata': Data for the .h header file,
1577         'sourcename': Name of the source code file,
1578         'sourcedata': Data for the .c source code file
1579        }
1580    '''
1581    f = parse_file(filename, fdesc, options)
1582
1583    # Provide dependencies if available
1584    for dep in f.fdesc.dependency:
1585        if dep in other_files:
1586            f.add_dependency(other_files[dep])
1587
1588    # Decide the file names
1589    noext = os.path.splitext(filename)[0]
1590    headername = noext + options.extension + options.header_extension
1591    sourcename = noext + options.extension + options.source_extension
1592    headerbasename = os.path.basename(headername)
1593
1594    # List of .proto files that should not be included in the C header file
1595    # even if they are mentioned in the source .proto.
1596    excludes = ['nanopb.proto', 'google/protobuf/descriptor.proto'] + options.exclude
1597    includes = [d for d in f.fdesc.dependency if d not in excludes]
1598
1599    headerdata = ''.join(f.generate_header(includes, headerbasename, options))
1600    sourcedata = ''.join(f.generate_source(headerbasename, options))
1601
1602    # Check if there were any lines in .options that did not match a member
1603    unmatched = [n for n,o in Globals.separate_options if n not in Globals.matched_namemasks]
1604    if unmatched and not options.quiet:
1605        sys.stderr.write("Following patterns in " + f.optfilename + " did not match any fields: "
1606                         + ', '.join(unmatched) + "\n")
1607        if not Globals.verbose_options:
1608            sys.stderr.write("Use  protoc --nanopb-out=-v:.   to see a list of the field names.\n")
1609
1610    return {'headername': headername, 'headerdata': headerdata,
1611            'sourcename': sourcename, 'sourcedata': sourcedata}
1612
1613def main_cli():
1614    '''Main function when invoked directly from the command line.'''
1615
1616    options, filenames = optparser.parse_args()
1617
1618    if not filenames:
1619        optparser.print_help()
1620        sys.exit(1)
1621
1622    if options.quiet:
1623        options.verbose = False
1624
1625    if options.output_dir and not os.path.exists(options.output_dir):
1626        optparser.print_help()
1627        sys.stderr.write("\noutput_dir does not exist: %s\n" % options.output_dir)
1628        sys.exit(1)
1629
1630    if options.verbose:
1631        sys.stderr.write('Google Python protobuf library imported from %s, version %s\n'
1632                         % (google.protobuf.__file__, google.protobuf.__version__))
1633
1634    Globals.verbose_options = options.verbose
1635    for filename in filenames:
1636        results = process_file(filename, None, options)
1637
1638        base_dir = options.output_dir or ''
1639        to_write = [
1640            (os.path.join(base_dir, results['headername']), results['headerdata']),
1641            (os.path.join(base_dir, results['sourcename']), results['sourcedata']),
1642        ]
1643
1644        if not options.quiet:
1645            paths = " and ".join([x[0] for x in to_write])
1646            sys.stderr.write("Writing to %s\n" % paths)
1647
1648        for path, data in to_write:
1649            with open(path, 'w') as f:
1650                f.write(data)
1651
1652def main_plugin():
1653    '''Main function when invoked as a protoc plugin.'''
1654
1655    import io, sys
1656    if sys.platform == "win32":
1657        import os, msvcrt
1658        # Set stdin and stdout to binary mode
1659        msvcrt.setmode(sys.stdin.fileno(), os.O_BINARY)
1660        msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)
1661
1662    data = io.open(sys.stdin.fileno(), "rb").read()
1663
1664    request = plugin_pb2.CodeGeneratorRequest.FromString(data)
1665
1666    try:
1667        # Versions of Python prior to 2.7.3 do not support unicode
1668        # input to shlex.split(). Try to convert to str if possible.
1669        params = str(request.parameter)
1670    except UnicodeEncodeError:
1671        params = request.parameter
1672
1673    import shlex
1674    args = shlex.split(params)
1675    options, dummy = optparser.parse_args(args)
1676
1677    Globals.verbose_options = options.verbose
1678
1679    if options.verbose:
1680        sys.stderr.write('Google Python protobuf library imported from %s, version %s\n'
1681                         % (google.protobuf.__file__, google.protobuf.__version__))
1682
1683    response = plugin_pb2.CodeGeneratorResponse()
1684
1685    # Google's protoc does not currently indicate the full path of proto files.
1686    # Instead always add the main file path to the search dirs, that works for
1687    # the common case.
1688    import os.path
1689    options.options_path.append(os.path.dirname(request.file_to_generate[0]))
1690
1691    # Process any include files first, in order to have them
1692    # available as dependencies
1693    other_files = {}
1694    for fdesc in request.proto_file:
1695        other_files[fdesc.name] = parse_file(fdesc.name, fdesc, options)
1696
1697    for filename in request.file_to_generate:
1698        for fdesc in request.proto_file:
1699            if fdesc.name == filename:
1700                results = process_file(filename, fdesc, options, other_files)
1701
1702                f = response.file.add()
1703                f.name = results['headername']
1704                f.content = results['headerdata']
1705
1706                f = response.file.add()
1707                f.name = results['sourcename']
1708                f.content = results['sourcedata']
1709
1710    io.open(sys.stdout.fileno(), "wb").write(response.SerializeToString())
1711
1712if __name__ == '__main__':
1713    # Check if we are running as a plugin under protoc
1714    if 'protoc-gen-' in sys.argv[0] or '--protoc-plugin' in sys.argv:
1715        main_plugin()
1716    else:
1717        main_cli()
1718