xref: /aosp_15_r20/external/nanopb-c/generator/nanopb_generator.py (revision c8d645cafcee3f91213d30caa0fe303887010b9b)
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-0.3.9.8"
7
8import sys
9import re
10import codecs
11import copy
12from functools import reduce
13
14try:
15    # Add some dummy imports to keep packaging tools happy.
16    import google, distutils.util # bbfreeze seems to need these
17    import pkg_resources # pyinstaller / protobuf 2.5 seem to need these
18    import proto.nanopb_pb2 as nanopb_pb2 # pyinstaller seems to need this
19    import proto.plugin_pb2 as plugin_pb2
20except:
21    # Don't care, we will error out later if it is actually important.
22    pass
23
24try:
25    import google.protobuf.text_format as text_format
26    import google.protobuf.descriptor_pb2 as descriptor
27except:
28    sys.stderr.write('''
29         *************************************************************
30         *** Could not import the Google protobuf Python libraries ***
31         *** Try installing package 'python-protobuf' or similar.  ***
32         *************************************************************
33    ''' + '\n')
34    raise
35
36try:
37    from .proto import nanopb_pb2, plugin_pb2
38except TypeError:
39    sys.stderr.write('''
40         ****************************************************************************
41         *** Got TypeError when importing the protocol definitions for generator. ***
42         *** This usually means that the protoc in your path doesn't match the    ***
43         *** Python protobuf library version.                                     ***
44         ***                                                                      ***
45         *** Please check the output of the following commands:                   ***
46         *** which protoc                                                         ***
47         *** protoc --version                                                     ***
48         *** python -c 'import google.protobuf; print(google.protobuf.__file__)'  ***
49         *** If you are not able to find the python protobuf version using the    ***
50         *** above command, use this command.                                     ***
51         *** pip freeze | grep -i protobuf                                        ***
52         ****************************************************************************
53    ''' + '\n')
54    raise
55except (ValueError, SystemError, ImportError):
56    # Probably invoked directly instead of via installed scripts.
57    import proto.nanopb_pb2 as nanopb_pb2
58    import proto.plugin_pb2 as plugin_pb2
59except:
60    sys.stderr.write('''
61         ********************************************************************
62         *** Failed to import the protocol definitions for generator.     ***
63         *** You have to run 'make' in the nanopb/generator/proto folder. ***
64         ********************************************************************
65    ''' + '\n')
66    raise
67
68# ---------------------------------------------------------------------------
69#                     Generation of single fields
70# ---------------------------------------------------------------------------
71
72import time
73import os.path
74
75# Values are tuple (c type, pb type, encoded size, int_size_allowed)
76FieldD = descriptor.FieldDescriptorProto
77datatypes = {
78    FieldD.TYPE_BOOL:       ('bool',     'BOOL',        1,  False),
79    FieldD.TYPE_DOUBLE:     ('double',   'DOUBLE',      8,  False),
80    FieldD.TYPE_FIXED32:    ('uint32_t', 'FIXED32',     4,  False),
81    FieldD.TYPE_FIXED64:    ('uint64_t', 'FIXED64',     8,  False),
82    FieldD.TYPE_FLOAT:      ('float',    'FLOAT',       4,  False),
83    FieldD.TYPE_INT32:      ('int32_t',  'INT32',      10,  True),
84    FieldD.TYPE_INT64:      ('int64_t',  'INT64',      10,  True),
85    FieldD.TYPE_SFIXED32:   ('int32_t',  'SFIXED32',    4,  False),
86    FieldD.TYPE_SFIXED64:   ('int64_t',  'SFIXED64',    8,  False),
87    FieldD.TYPE_SINT32:     ('int32_t',  'SINT32',      5,  True),
88    FieldD.TYPE_SINT64:     ('int64_t',  'SINT64',     10,  True),
89    FieldD.TYPE_UINT32:     ('uint32_t', 'UINT32',      5,  True),
90    FieldD.TYPE_UINT64:     ('uint64_t', 'UINT64',     10,  True)
91}
92
93# Integer size overrides (from .proto settings)
94intsizes = {
95    nanopb_pb2.IS_8:     'int8_t',
96    nanopb_pb2.IS_16:    'int16_t',
97    nanopb_pb2.IS_32:    'int32_t',
98    nanopb_pb2.IS_64:    'int64_t',
99}
100
101# String types (for python 2 / python 3 compatibility)
102try:
103    strtypes = (unicode, str)
104except NameError:
105    strtypes = (str, )
106
107
108class Names:
109    '''Keeps a set of nested names and formats them to C identifier.'''
110    def __init__(self, parts = ()):
111        if isinstance(parts, Names):
112            parts = parts.parts
113        elif isinstance(parts, strtypes):
114            parts = (parts,)
115        self.parts = tuple(parts)
116
117    def __str__(self):
118        return '_'.join(self.parts)
119
120    def __add__(self, other):
121        if isinstance(other, strtypes):
122            return Names(self.parts + (other,))
123        elif isinstance(other, Names):
124            return Names(self.parts + other.parts)
125        elif isinstance(other, tuple):
126            return Names(self.parts + other)
127        else:
128            raise ValueError("Name parts should be of type str")
129
130    def __eq__(self, other):
131        return isinstance(other, Names) and self.parts == other.parts
132
133def names_from_type_name(type_name):
134    '''Parse Names() from FieldDescriptorProto type_name'''
135    if type_name[0] != '.':
136        raise NotImplementedError("Lookup of non-absolute type names is not supported")
137    return Names(type_name[1:].split('.'))
138
139def varint_max_size(max_value):
140    '''Returns the maximum number of bytes a varint can take when encoded.'''
141    if max_value < 0:
142        max_value = 2**64 - max_value
143    for i in range(1, 11):
144        if (max_value >> (i * 7)) == 0:
145            return i
146    raise ValueError("Value too large for varint: " + str(max_value))
147
148assert varint_max_size(-1) == 10
149assert varint_max_size(0) == 1
150assert varint_max_size(127) == 1
151assert varint_max_size(128) == 2
152
153class EncodedSize:
154    '''Class used to represent the encoded size of a field or a message.
155    Consists of a combination of symbolic sizes and integer sizes.'''
156    def __init__(self, value = 0, symbols = []):
157        if isinstance(value, EncodedSize):
158            self.value = value.value
159            self.symbols = value.symbols
160        elif isinstance(value, strtypes + (Names,)):
161            self.symbols = [str(value)]
162            self.value = 0
163        else:
164            self.value = value
165            self.symbols = symbols
166
167    def __add__(self, other):
168        if isinstance(other, int):
169            return EncodedSize(self.value + other, self.symbols)
170        elif isinstance(other, strtypes + (Names,)):
171            return EncodedSize(self.value, self.symbols + [str(other)])
172        elif isinstance(other, EncodedSize):
173            return EncodedSize(self.value + other.value, self.symbols + other.symbols)
174        else:
175            raise ValueError("Cannot add size: " + repr(other))
176
177    def __mul__(self, other):
178        if isinstance(other, int):
179            return EncodedSize(self.value * other, [str(other) + '*' + s for s in self.symbols])
180        else:
181            raise ValueError("Cannot multiply size: " + repr(other))
182
183    def __str__(self):
184        if not self.symbols:
185            return str(self.value)
186        else:
187            return '(' + str(self.value) + ' + ' + ' + '.join(self.symbols) + ')'
188
189    def upperlimit(self):
190        if not self.symbols:
191            return self.value
192        else:
193            return 2**32 - 1
194
195class Enum:
196    def __init__(self, names, desc, enum_options):
197        '''desc is EnumDescriptorProto'''
198
199        self.options = enum_options
200        self.names = names
201
202        # by definition, `names` include this enum's name
203        base_name = Names(names.parts[:-1])
204
205        if enum_options.long_names:
206            self.values = [(names + x.name, x.number) for x in desc.value]
207        else:
208            self.values = [(base_name + x.name, x.number) for x in desc.value]
209
210        self.value_longnames = [self.names + x.name for x in desc.value]
211        self.packed = enum_options.packed_enum
212
213    def has_negative(self):
214        for n, v in self.values:
215            if v < 0:
216                return True
217        return False
218
219    def encoded_size(self):
220        return max([varint_max_size(v) for n,v in self.values])
221
222    def __str__(self):
223        result = 'typedef enum _%s {\n' % self.names
224        result += ',\n'.join(["    %s = %d" % x for x in self.values])
225        result += '\n}'
226
227        if self.packed:
228            result += ' pb_packed'
229
230        result += ' %s;' % self.names
231
232        # sort the enum by value
233        sorted_values = sorted(self.values, key = lambda x: (x[1], x[0]))
234
235        result += '\n#define _%s_MIN %s' % (self.names, sorted_values[0][0])
236        result += '\n#define _%s_MAX %s' % (self.names, sorted_values[-1][0])
237        result += '\n#define _%s_ARRAYSIZE ((%s)(%s+1))' % (self.names, self.names, sorted_values[-1][0])
238
239        if not self.options.long_names:
240            # Define the long names always so that enum value references
241            # from other files work properly.
242            for i, x in enumerate(self.values):
243                result += '\n#define %s %s' % (self.value_longnames[i], x[0])
244
245        if self.options.enum_to_string:
246            result += '\nconst char *%s_name(%s v);\n' % (self.names, self.names)
247
248        return result
249
250    def enum_to_string_definition(self):
251        if not self.options.enum_to_string:
252            return ""
253
254        result = 'const char *%s_name(%s v) {\n' % (self.names, self.names)
255        result += '    switch (v) {\n'
256
257        for ((enumname, _), strname) in zip(self.values, self.value_longnames):
258            # Strip off the leading type name from the string value.
259            strval = str(strname)[len(str(self.names)) + 1:]
260            result += '        case %s: return "%s";\n' % (enumname, strval)
261
262        result += '    }\n'
263        result += '    return "unknown";\n'
264        result += '}\n'
265
266        return result
267
268class FieldMaxSize:
269    def __init__(self, worst = 0, checks = [], field_name = 'undefined'):
270        if isinstance(worst, list):
271            self.worst = max(i for i in worst if i is not None)
272        else:
273            self.worst = worst
274
275        self.worst_field = field_name
276        self.checks = list(checks)
277
278    def extend(self, extend, field_name = None):
279        self.worst = max(self.worst, extend.worst)
280
281        if self.worst == extend.worst:
282            self.worst_field = extend.worst_field
283
284        self.checks.extend(extend.checks)
285
286class Field:
287    def __init__(self, struct_name, desc, field_options):
288        '''desc is FieldDescriptorProto'''
289        self.tag = desc.number
290        self.struct_name = struct_name
291        self.union_name = None
292        self.name = desc.name
293        self.default = None
294        self.max_size = None
295        self.max_count = None
296        self.array_decl = ""
297        self.enc_size = None
298        self.ctype = None
299        self.fixed_count = False
300
301        if field_options.type == nanopb_pb2.FT_INLINE:
302            # Before nanopb-0.3.8, fixed length bytes arrays were specified
303            # by setting type to FT_INLINE. But to handle pointer typed fields,
304            # it makes sense to have it as a separate option.
305            field_options.type = nanopb_pb2.FT_STATIC
306            field_options.fixed_length = True
307
308        # Parse field options
309        if field_options.HasField("max_size"):
310            self.max_size = field_options.max_size
311
312        if desc.type == FieldD.TYPE_STRING and field_options.HasField("max_length"):
313            # max_length overrides max_size for strings
314            self.max_size = field_options.max_length + 1
315
316        if field_options.HasField("max_count"):
317            self.max_count = field_options.max_count
318
319        if desc.HasField('default_value'):
320            self.default = desc.default_value
321
322        # Check field rules, i.e. required/optional/repeated.
323        can_be_static = True
324        if desc.label == FieldD.LABEL_REPEATED:
325            self.rules = 'REPEATED'
326            if self.max_count is None:
327                can_be_static = False
328            else:
329                self.array_decl = '[%d]' % self.max_count
330                self.fixed_count = field_options.fixed_count
331
332        elif field_options.proto3:
333            self.rules = 'SINGULAR'
334        elif desc.label == FieldD.LABEL_REQUIRED:
335            self.rules = 'REQUIRED'
336        elif desc.label == FieldD.LABEL_OPTIONAL:
337            self.rules = 'OPTIONAL'
338        else:
339            raise NotImplementedError(desc.label)
340
341        # Check if the field can be implemented with static allocation
342        # i.e. whether the data size is known.
343        if desc.type == FieldD.TYPE_STRING and self.max_size is None:
344            can_be_static = False
345
346        if desc.type == FieldD.TYPE_BYTES and self.max_size is None:
347            can_be_static = False
348
349        # Decide how the field data will be allocated
350        if field_options.type == nanopb_pb2.FT_DEFAULT:
351            if can_be_static:
352                field_options.type = nanopb_pb2.FT_STATIC
353            else:
354                field_options.type = nanopb_pb2.FT_CALLBACK
355
356        if field_options.type == nanopb_pb2.FT_STATIC and not can_be_static:
357            raise Exception("Field '%s' is defined as static, but max_size or "
358                            "max_count is not given." % self.name)
359
360        if field_options.fixed_count and self.max_count is None:
361            raise Exception("Field '%s' is defined as fixed count, "
362                            "but max_count is not given." % self.name)
363
364        if field_options.type == nanopb_pb2.FT_STATIC:
365            self.allocation = 'STATIC'
366        elif field_options.type == nanopb_pb2.FT_POINTER:
367            self.allocation = 'POINTER'
368        elif field_options.type == nanopb_pb2.FT_CALLBACK:
369            self.allocation = 'CALLBACK'
370        else:
371            raise NotImplementedError(field_options.type)
372
373        # Decide the C data type to use in the struct.
374        if desc.type in datatypes:
375            self.ctype, self.pbtype, self.enc_size, isa = datatypes[desc.type]
376
377            # Override the field size if user wants to use smaller integers
378            if isa and field_options.int_size != nanopb_pb2.IS_DEFAULT:
379                self.ctype = intsizes[field_options.int_size]
380                if desc.type == FieldD.TYPE_UINT32 or desc.type == FieldD.TYPE_UINT64:
381                    self.ctype = 'u' + self.ctype;
382        elif desc.type == FieldD.TYPE_ENUM:
383            self.pbtype = 'ENUM'
384            self.ctype = names_from_type_name(desc.type_name)
385            if self.default is not None:
386                self.default = self.ctype + self.default
387            self.enc_size = None # Needs to be filled in when enum values are known
388        elif desc.type == FieldD.TYPE_STRING:
389            self.pbtype = 'STRING'
390            self.ctype = 'char'
391            if self.allocation == 'STATIC':
392                self.ctype = 'char'
393                self.array_decl += '[%d]' % self.max_size
394                self.enc_size = varint_max_size(self.max_size) + self.max_size
395        elif desc.type == FieldD.TYPE_BYTES:
396            if field_options.fixed_length:
397                self.pbtype = 'FIXED_LENGTH_BYTES'
398
399                if self.max_size is None:
400                    raise Exception("Field '%s' is defined as fixed length, "
401                                    "but max_size is not given." % self.name)
402
403                self.enc_size = varint_max_size(self.max_size) + self.max_size
404                self.ctype = 'pb_byte_t'
405                self.array_decl += '[%d]' % self.max_size
406            else:
407                self.pbtype = 'BYTES'
408                self.ctype = 'pb_bytes_array_t'
409                if self.allocation == 'STATIC':
410                    self.ctype = self.struct_name + self.name + 't'
411                    self.enc_size = varint_max_size(self.max_size) + self.max_size
412        elif desc.type == FieldD.TYPE_MESSAGE:
413            self.pbtype = 'MESSAGE'
414            self.ctype = self.submsgname = names_from_type_name(desc.type_name)
415            self.enc_size = None # Needs to be filled in after the message type is available
416        else:
417            raise NotImplementedError(desc.type)
418
419    def __lt__(self, other):
420        return self.tag < other.tag
421
422    def __str__(self):
423        result = ''
424        if self.allocation == 'POINTER':
425            if self.rules == 'REPEATED':
426                result += '    pb_size_t ' + self.name + '_count;\n'
427
428            if self.pbtype == 'MESSAGE':
429                # Use struct definition, so recursive submessages are possible
430                result += '    struct _%s *%s;' % (self.ctype, self.name)
431            elif self.pbtype == 'FIXED_LENGTH_BYTES':
432                # Pointer to fixed size array
433                result += '    %s (*%s)%s;' % (self.ctype, self.name, self.array_decl)
434            elif self.rules == 'REPEATED' and self.pbtype in ['STRING', 'BYTES']:
435                # String/bytes arrays need to be defined as pointers to pointers
436                result += '    %s **%s;' % (self.ctype, self.name)
437            else:
438                result += '    %s *%s;' % (self.ctype, self.name)
439        elif self.allocation == 'CALLBACK':
440            result += '    pb_callback_t %s;' % self.name
441        else:
442            if self.rules == 'OPTIONAL' and self.allocation == 'STATIC':
443                result += '    bool has_' + self.name + ';\n'
444            elif (self.rules == 'REPEATED' and
445                  self.allocation == 'STATIC' and
446                  not self.fixed_count):
447                result += '    pb_size_t ' + self.name + '_count;\n'
448            result += '    %s %s%s;' % (self.ctype, self.name, self.array_decl)
449        return result
450
451    def types(self):
452        '''Return definitions for any special types this field might need.'''
453        if self.pbtype == 'BYTES' and self.allocation == 'STATIC':
454            result = 'typedef PB_BYTES_ARRAY_T(%d) %s;\n' % (self.max_size, self.ctype)
455        else:
456            result = ''
457        return result
458
459    def get_dependencies(self):
460        '''Get list of type names used by this field.'''
461        if self.allocation == 'STATIC':
462            return [str(self.ctype)]
463        else:
464            return []
465
466    def get_initializer(self, null_init, inner_init_only = False):
467        '''Return literal expression for this field's default value.
468        null_init: If True, initialize to a 0 value instead of default from .proto
469        inner_init_only: If True, exclude initialization for any count/has fields
470        '''
471
472        inner_init = None
473        if self.pbtype == 'MESSAGE':
474            if null_init:
475                inner_init = '%s_init_zero' % self.ctype
476            else:
477                inner_init = '%s_init_default' % self.ctype
478        elif self.default is None or null_init:
479            if self.pbtype == 'STRING':
480                inner_init = '""'
481            elif self.pbtype == 'BYTES':
482                inner_init = '{0, {0}}'
483            elif self.pbtype == 'FIXED_LENGTH_BYTES':
484                inner_init = '{0}'
485            elif self.pbtype in ('ENUM', 'UENUM'):
486                inner_init = '_%s_MIN' % self.ctype
487            else:
488                inner_init = '0'
489        else:
490            if self.pbtype == 'STRING':
491                data = codecs.escape_encode(self.default.encode('utf-8'))[0]
492                inner_init = '"' + data.decode('ascii') + '"'
493            elif self.pbtype == 'BYTES':
494                data = codecs.escape_decode(self.default)[0]
495                data = ["0x%02x" % c for c in bytearray(data)]
496                if len(data) == 0:
497                    inner_init = '{0, {0}}'
498                else:
499                    inner_init = '{%d, {%s}}' % (len(data), ','.join(data))
500            elif self.pbtype == 'FIXED_LENGTH_BYTES':
501                data = codecs.escape_decode(self.default)[0]
502                data = ["0x%02x" % c for c in bytearray(data)]
503                if len(data) == 0:
504                    inner_init = '{0}'
505                else:
506                    inner_init = '{%s}' % ','.join(data)
507            elif self.pbtype in ['FIXED32', 'UINT32']:
508                inner_init = str(self.default) + 'u'
509            elif self.pbtype in ['FIXED64', 'UINT64']:
510                inner_init = str(self.default) + 'ull'
511            elif self.pbtype in ['SFIXED64', 'INT64']:
512                inner_init = str(self.default) + 'll'
513            else:
514                inner_init = str(self.default)
515
516        if inner_init_only:
517            return inner_init
518
519        outer_init = None
520        if self.allocation == 'STATIC':
521            if self.rules == 'REPEATED':
522                outer_init = ''
523                if not self.fixed_count:
524                    outer_init += '0, '
525                outer_init += '{'
526                outer_init += ', '.join([inner_init] * self.max_count)
527                outer_init += '}'
528            elif self.rules == 'OPTIONAL':
529                outer_init = 'false, ' + inner_init
530            else:
531                outer_init = inner_init
532        elif self.allocation == 'POINTER':
533            if self.rules == 'REPEATED':
534                outer_init = '0, NULL'
535            else:
536                outer_init = 'NULL'
537        elif self.allocation == 'CALLBACK':
538            if self.pbtype == 'EXTENSION':
539                outer_init = 'NULL'
540            else:
541                outer_init = '{{NULL}, NULL}'
542
543        return outer_init
544
545    def default_decl(self, declaration_only = False):
546        '''Return definition for this field's default value.'''
547        if self.default is None:
548            return None
549
550        ctype = self.ctype
551        default = self.get_initializer(False, True)
552        array_decl = ''
553
554        if self.pbtype == 'STRING':
555            if self.allocation != 'STATIC':
556                return None # Not implemented
557            array_decl = '[%d]' % self.max_size
558        elif self.pbtype == 'BYTES':
559            if self.allocation != 'STATIC':
560                return None # Not implemented
561        elif self.pbtype == 'FIXED_LENGTH_BYTES':
562            if self.allocation != 'STATIC':
563                return None # Not implemented
564            array_decl = '[%d]' % self.max_size
565
566        if declaration_only:
567            return 'extern const %s %s_default%s;' % (ctype, self.struct_name + self.name, array_decl)
568        else:
569            return 'const %s %s_default%s = %s;' % (ctype, self.struct_name + self.name, array_decl, default)
570
571    def tags(self):
572        '''Return the #define for the tag number of this field.'''
573        identifier = '%s_%s_tag' % (self.struct_name, self.name)
574        return '#define %-40s %d\n' % (identifier, self.tag)
575
576    def pb_field_t(self, prev_field_name, union_index = None):
577        '''Return the pb_field_t initializer to use in the constant array.
578        prev_field_name is the name of the previous field or None. For OneOf
579        unions, union_index is the index of this field inside the OneOf.
580        '''
581
582        if self.rules == 'ONEOF':
583            if self.anonymous:
584                result = '    PB_ANONYMOUS_ONEOF_FIELD(%s, ' % self.union_name
585            else:
586                result = '    PB_ONEOF_FIELD(%s, ' % self.union_name
587        elif self.fixed_count:
588            result = '    PB_REPEATED_FIXED_COUNT('
589        else:
590            result = '    PB_FIELD('
591
592        result += '%3d, ' % self.tag
593        result += '%-8s, ' % self.pbtype
594        if not self.fixed_count:
595            result += '%s, ' % self.rules
596            result += '%-8s, ' % self.allocation
597
598        if union_index is not None and union_index > 0:
599            result += 'UNION, '
600        elif prev_field_name is None:
601            result += 'FIRST, '
602        else:
603            result += 'OTHER, '
604
605        result += '%s, ' % self.struct_name
606        result += '%s, ' % self.name
607        result += '%s, ' % (prev_field_name or self.name)
608
609        if self.pbtype == 'MESSAGE':
610            result += '&%s_fields)' % self.submsgname
611        elif self.default is None:
612            result += '0)'
613        elif self.pbtype in ['BYTES', 'STRING', 'FIXED_LENGTH_BYTES'] and self.allocation != 'STATIC':
614            result += '0)' # Arbitrary size default values not implemented
615        elif self.rules == 'OPTEXT':
616            result += '0)' # Default value for extensions is not implemented
617        else:
618            result += '&%s_default)' % (self.struct_name + self.name)
619
620        return result
621
622    def get_last_field_name(self):
623        return self.name
624
625    def largest_field_value(self):
626        '''Determine if this field needs 16bit or 32bit pb_field_t structure to compile properly.
627        Returns numeric value or a C-expression for assert.'''
628        check = []
629
630        need_check = False
631
632        if self.pbtype == 'BYTES' and self.allocation == 'STATIC' and self.max_size > 251:
633            need_check = True
634        elif self.pbtype == 'MESSAGE' and self.allocation == 'STATIC':
635            need_check = True
636
637        if need_check:
638            if self.rules == 'REPEATED':
639                check.append('pb_membersize(%s, %s[0])' % (self.struct_name, self.name))
640            elif self.rules == 'ONEOF':
641                if self.anonymous:
642                    check.append('pb_membersize(%s, %s)' % (self.struct_name, self.name))
643                else:
644                    check.append('pb_membersize(%s, %s.%s)' % (self.struct_name, self.union_name, self.name))
645            else:
646                check.append('pb_membersize(%s, %s)' % (self.struct_name, self.name))
647
648        return FieldMaxSize([self.tag, self.max_size, self.max_count],
649                            check,
650                            ('%s.%s' % (self.struct_name, self.name)))
651
652    def encoded_size(self, dependencies):
653        '''Return the maximum size that this field can take when encoded,
654        including the field tag. If the size cannot be determined, returns
655        None.'''
656
657        if self.allocation != 'STATIC':
658            return None
659
660        if self.pbtype == 'MESSAGE':
661            encsize = None
662            if str(self.submsgname) in dependencies:
663                submsg = dependencies[str(self.submsgname)]
664                encsize = submsg.encoded_size(dependencies)
665                if encsize is not None:
666                    # Include submessage length prefix
667                    encsize += varint_max_size(encsize.upperlimit())
668                else:
669                    my_msg = dependencies.get(str(self.struct_name))
670                    if my_msg and submsg.protofile == my_msg.protofile:
671                        # The dependency is from the same file and size cannot be
672                        # determined for it, thus we know it will not be possible
673                        # in runtime either.
674                        return None
675
676            if encsize is None:
677                # Submessage or its size cannot be found.
678                # This can occur if submessage is defined in different
679                # file, and it or its .options could not be found.
680                # Instead of direct numeric value, reference the size that
681                # has been #defined in the other file.
682                encsize = EncodedSize(self.submsgname + 'size')
683
684                # We will have to make a conservative assumption on the length
685                # prefix size, though.
686                encsize += 5
687
688        elif self.pbtype in ['ENUM', 'UENUM']:
689            if str(self.ctype) in dependencies:
690                enumtype = dependencies[str(self.ctype)]
691                encsize = enumtype.encoded_size()
692            else:
693                # Conservative assumption
694                encsize = 10
695
696        elif self.enc_size is None:
697            raise RuntimeError("Could not determine encoded size for %s.%s"
698                               % (self.struct_name, self.name))
699        else:
700            encsize = EncodedSize(self.enc_size)
701
702        encsize += varint_max_size(self.tag << 3) # Tag + wire type
703
704        if self.rules == 'REPEATED':
705            # Decoders must be always able to handle unpacked arrays.
706            # Therefore we have to reserve space for it, even though
707            # we emit packed arrays ourselves. For length of 1, packed
708            # arrays are larger however so we need to add allowance
709            # for the length byte.
710            encsize *= self.max_count
711
712            if self.max_count == 1:
713                encsize += 1
714
715        return encsize
716
717
718class ExtensionRange(Field):
719    def __init__(self, struct_name, range_start, field_options):
720        '''Implements a special pb_extension_t* field in an extensible message
721        structure. The range_start signifies the index at which the extensions
722        start. Not necessarily all tags above this are extensions, it is merely
723        a speed optimization.
724        '''
725        self.tag = range_start
726        self.struct_name = struct_name
727        self.name = 'extensions'
728        self.pbtype = 'EXTENSION'
729        self.rules = 'OPTIONAL'
730        self.allocation = 'CALLBACK'
731        self.ctype = 'pb_extension_t'
732        self.array_decl = ''
733        self.default = None
734        self.max_size = 0
735        self.max_count = 0
736        self.fixed_count = False
737
738    def __str__(self):
739        return '    pb_extension_t *extensions;'
740
741    def types(self):
742        return ''
743
744    def tags(self):
745        return ''
746
747    def encoded_size(self, dependencies):
748        # We exclude extensions from the count, because they cannot be known
749        # until runtime. Other option would be to return None here, but this
750        # way the value remains useful if extensions are not used.
751        return EncodedSize(0)
752
753class ExtensionField(Field):
754    def __init__(self, fullname, desc, field_options):
755        self.fullname = fullname
756        self.extendee_name = names_from_type_name(desc.extendee)
757        Field.__init__(self, self.fullname + 'struct', desc, field_options)
758
759        if self.rules != 'OPTIONAL':
760            self.skip = True
761        else:
762            self.skip = False
763            self.rules = 'OPTEXT'
764
765    def tags(self):
766        '''Return the #define for the tag number of this field.'''
767        identifier = '%s_tag' % self.fullname
768        return '#define %-40s %d\n' % (identifier, self.tag)
769
770    def extension_decl(self):
771        '''Declaration of the extension type in the .pb.h file'''
772        if self.skip:
773            msg = '/* Extension field %s was skipped because only "optional"\n' % self.fullname
774            msg +='   type of extension fields is currently supported. */\n'
775            return msg
776
777        return ('extern const pb_extension_type_t %s; /* field type: %s */\n' %
778            (self.fullname, str(self).strip()))
779
780    def extension_def(self):
781        '''Definition of the extension type in the .pb.c file'''
782
783        if self.skip:
784            return ''
785
786        result  = 'typedef struct {\n'
787        result += str(self)
788        result += '\n} %s;\n\n' % self.struct_name
789        result += ('static const pb_field_t %s_field = \n  %s;\n\n' %
790                    (self.fullname, self.pb_field_t(None)))
791        result += 'const pb_extension_type_t %s = {\n' % self.fullname
792        result += '    NULL,\n'
793        result += '    NULL,\n'
794        result += '    &%s_field\n' % self.fullname
795        result += '};\n'
796        return result
797
798
799# ---------------------------------------------------------------------------
800#                   Generation of oneofs (unions)
801# ---------------------------------------------------------------------------
802
803class OneOf(Field):
804    def __init__(self, struct_name, oneof_desc):
805        self.struct_name = struct_name
806        self.name = oneof_desc.name
807        self.ctype = 'union'
808        self.pbtype = 'oneof'
809        self.fields = []
810        self.allocation = 'ONEOF'
811        self.default = None
812        self.rules = 'ONEOF'
813        self.anonymous = False
814
815    def add_field(self, field):
816        if field.allocation == 'CALLBACK':
817            raise Exception("Callback fields inside of oneof are not supported"
818                            + " (field %s)" % field.name)
819
820        field.union_name = self.name
821        field.rules = 'ONEOF'
822        field.anonymous = self.anonymous
823        self.fields.append(field)
824        self.fields.sort(key = lambda f: f.tag)
825
826        # Sort by the lowest tag number inside union
827        self.tag = min([f.tag for f in self.fields])
828
829    def __str__(self):
830        result = ''
831        if self.fields:
832            result += '    pb_size_t which_' + self.name + ";\n"
833            result += '    union {\n'
834            for f in self.fields:
835                result += '    ' + str(f).replace('\n', '\n    ') + '\n'
836            if self.anonymous:
837                result += '    };'
838            else:
839                result += '    } ' + self.name + ';'
840        return result
841
842    def types(self):
843        return ''.join([f.types() for f in self.fields])
844
845    def get_dependencies(self):
846        deps = []
847        for f in self.fields:
848            deps += f.get_dependencies()
849        return deps
850
851    def get_initializer(self, null_init):
852        return '0, {' + self.fields[0].get_initializer(null_init) + '}'
853
854    def default_decl(self, declaration_only = False):
855        return None
856
857    def tags(self):
858        return ''.join([f.tags() for f in self.fields])
859
860    def pb_field_t(self, prev_field_name):
861        parts = []
862        for union_index, field in enumerate(self.fields):
863            parts.append(field.pb_field_t(prev_field_name, union_index))
864        return ',\n'.join(parts)
865
866    def get_last_field_name(self):
867        if self.anonymous:
868            return self.fields[-1].name
869        else:
870            return self.name + '.' + self.fields[-1].name
871
872    def largest_field_value(self):
873        largest = FieldMaxSize()
874        for f in self.fields:
875            largest.extend(f.largest_field_value())
876        return largest
877
878    def encoded_size(self, dependencies):
879        '''Returns the size of the largest oneof field.'''
880        largest = 0
881        symbols = []
882        for f in self.fields:
883            size = EncodedSize(f.encoded_size(dependencies))
884            if size is None or size.value is None:
885                return None
886            elif size.symbols:
887                symbols.append((f.tag, size.symbols[0]))
888            elif size.value > largest:
889                largest = size.value
890
891        if not symbols:
892            # Simple case, all sizes were known at generator time
893            return largest
894
895        if largest > 0:
896            # Some sizes were known, some were not
897            symbols.insert(0, (0, largest))
898
899        if len(symbols) == 1:
900            # Only one symbol was needed
901            return EncodedSize(5, [symbols[0][1]])
902        else:
903            # Use sizeof(union{}) construct to find the maximum size of
904            # submessages.
905            union_def = ' '.join('char f%d[%s];' % s for s in symbols)
906            return EncodedSize(5, ['sizeof(union{%s})' % union_def])
907
908# ---------------------------------------------------------------------------
909#                   Generation of messages (structures)
910# ---------------------------------------------------------------------------
911
912
913class Message:
914    def __init__(self, names, desc, message_options):
915        self.name = names
916        self.fields = []
917        self.oneofs = {}
918        no_unions = []
919
920        if message_options.msgid:
921            self.msgid = message_options.msgid
922
923        if hasattr(desc, 'oneof_decl'):
924            for i, f in enumerate(desc.oneof_decl):
925                oneof_options = get_nanopb_suboptions(desc, message_options, self.name + f.name)
926                if oneof_options.no_unions:
927                    no_unions.append(i) # No union, but add fields normally
928                elif oneof_options.type == nanopb_pb2.FT_IGNORE:
929                    pass # No union and skip fields also
930                else:
931                    oneof = OneOf(self.name, f)
932                    if oneof_options.anonymous_oneof:
933                        oneof.anonymous = True
934                    self.oneofs[i] = oneof
935                    self.fields.append(oneof)
936        else:
937            sys.stderr.write('Note: This Python protobuf library has no OneOf support\n')
938
939        for f in desc.field:
940            field_options = get_nanopb_suboptions(f, message_options, self.name + f.name)
941            if field_options.type == nanopb_pb2.FT_IGNORE:
942                continue
943
944            field = Field(self.name, f, field_options)
945            if (hasattr(f, 'oneof_index') and
946                f.HasField('oneof_index') and
947                f.oneof_index not in no_unions):
948                if f.oneof_index in self.oneofs:
949                    self.oneofs[f.oneof_index].add_field(field)
950            else:
951                self.fields.append(field)
952
953        if len(desc.extension_range) > 0:
954            field_options = get_nanopb_suboptions(desc, message_options, self.name + 'extensions')
955            range_start = min([r.start for r in desc.extension_range])
956            if field_options.type != nanopb_pb2.FT_IGNORE:
957                self.fields.append(ExtensionRange(self.name, range_start, field_options))
958
959        self.packed = message_options.packed_struct
960        self.ordered_fields = self.fields[:]
961        self.ordered_fields.sort()
962
963    def get_dependencies(self):
964        '''Get list of type names that this structure refers to.'''
965        deps = []
966        for f in self.fields:
967            deps += f.get_dependencies()
968        return deps
969
970    def __str__(self):
971        result = 'typedef struct _%s {\n' % self.name
972
973        if not self.ordered_fields:
974            # Empty structs are not allowed in C standard.
975            # Therefore add a dummy field if an empty message occurs.
976            result += '    char dummy_field;'
977
978        result += '\n'.join([str(f) for f in self.ordered_fields])
979        result += '\n/* @@protoc_insertion_point(struct:%s) */' % self.name
980        result += '\n}'
981
982        if self.packed:
983            result += ' pb_packed'
984
985        result += ' %s;' % self.name
986
987        if self.packed:
988            result = 'PB_PACKED_STRUCT_START\n' + result
989            result += '\nPB_PACKED_STRUCT_END'
990
991        return result
992
993    def types(self):
994        return ''.join([f.types() for f in self.fields])
995
996    def get_initializer(self, null_init):
997        if not self.ordered_fields:
998            return '{0}'
999
1000        parts = []
1001        for field in self.ordered_fields:
1002            parts.append(field.get_initializer(null_init))
1003        return '{' + ', '.join(parts) + '}'
1004
1005    def default_decl(self, declaration_only = False):
1006        result = ""
1007        for field in self.fields:
1008            default = field.default_decl(declaration_only)
1009            if default is not None:
1010                result += default + '\n'
1011        return result
1012
1013    def all_fields(self):
1014        '''Iterate over all fields in this message, including nested OneOfs.'''
1015        for f in self.fields:
1016            if isinstance(f, OneOf):
1017                for f2 in f.fields:
1018                    yield f2
1019            else:
1020                yield f
1021
1022    def count_required_fields(self):
1023        '''Returns number of required fields inside this message'''
1024        count = 0
1025        for f in self.fields:
1026            if not isinstance(f, OneOf):
1027                if f.rules == 'REQUIRED':
1028                    count += 1
1029        return count
1030
1031    def count_all_fields(self):
1032        count = 0
1033        for f in self.fields:
1034            if isinstance(f, OneOf):
1035                count += len(f.fields)
1036            else:
1037                count += 1
1038        return count
1039
1040    def fields_declaration(self):
1041        result = 'extern const pb_field_t %s_fields[%d];' % (self.name, self.count_all_fields() + 1)
1042        return result
1043
1044    def fields_definition(self):
1045        result = 'const pb_field_t %s_fields[%d] = {\n' % (self.name, self.count_all_fields() + 1)
1046
1047        prev = None
1048        for field in self.ordered_fields:
1049            result += field.pb_field_t(prev)
1050            result += ',\n'
1051            prev = field.get_last_field_name()
1052
1053        result += '    PB_LAST_FIELD\n};'
1054        return result
1055
1056    def encoded_size(self, dependencies):
1057        '''Return the maximum size that this message can take when encoded.
1058        If the size cannot be determined, returns None.
1059        '''
1060        size = EncodedSize(0)
1061        for field in self.fields:
1062            fsize = field.encoded_size(dependencies)
1063            if fsize is None:
1064                return None
1065            size += fsize
1066
1067        return size
1068
1069
1070# ---------------------------------------------------------------------------
1071#                    Processing of entire .proto files
1072# ---------------------------------------------------------------------------
1073
1074def iterate_messages(desc, flatten = False, names = Names()):
1075    '''Recursively find all messages. For each, yield name, DescriptorProto.'''
1076    if hasattr(desc, 'message_type'):
1077        submsgs = desc.message_type
1078    else:
1079        submsgs = desc.nested_type
1080
1081    for submsg in submsgs:
1082        sub_names = names + submsg.name
1083        if flatten:
1084            yield Names(submsg.name), submsg
1085        else:
1086            yield sub_names, submsg
1087
1088        for x in iterate_messages(submsg, flatten, sub_names):
1089            yield x
1090
1091def iterate_extensions(desc, flatten = False, names = Names()):
1092    '''Recursively find all extensions.
1093    For each, yield name, FieldDescriptorProto.
1094    '''
1095    for extension in desc.extension:
1096        yield names, extension
1097
1098    for subname, subdesc in iterate_messages(desc, flatten, names):
1099        for extension in subdesc.extension:
1100            yield subname, extension
1101
1102def toposort2(data):
1103    '''Topological sort.
1104    From http://code.activestate.com/recipes/577413-topological-sort/
1105    This function is under the MIT license.
1106    '''
1107    for k, v in list(data.items()):
1108        v.discard(k) # Ignore self dependencies
1109    extra_items_in_deps = reduce(set.union, list(data.values()), set()) - set(data.keys())
1110    data.update(dict([(item, set()) for item in extra_items_in_deps]))
1111    while True:
1112        ordered = set(item for item,dep in list(data.items()) if not dep)
1113        if not ordered:
1114            break
1115        for item in sorted(ordered):
1116            yield item
1117        data = dict([(item, (dep - ordered)) for item,dep in list(data.items())
1118                if item not in ordered])
1119    assert not data, "A cyclic dependency exists amongst %r" % data
1120
1121def sort_dependencies(messages):
1122    '''Sort a list of Messages based on dependencies.'''
1123    dependencies = {}
1124    message_by_name = {}
1125    for message in messages:
1126        dependencies[str(message.name)] = set(message.get_dependencies())
1127        message_by_name[str(message.name)] = message
1128
1129    for msgname in toposort2(dependencies):
1130        if msgname in message_by_name:
1131            yield message_by_name[msgname]
1132
1133def make_identifier(headername):
1134    '''Make #ifndef identifier that contains uppercase A-Z and digits 0-9'''
1135    result = ""
1136    for c in headername.upper():
1137        if c.isalnum():
1138            result += c
1139        else:
1140            result += '_'
1141    return result
1142
1143class ProtoFile:
1144    def __init__(self, fdesc, file_options):
1145        '''Takes a FileDescriptorProto and parses it.'''
1146        self.fdesc = fdesc
1147        self.file_options = file_options
1148        self.dependencies = {}
1149        self.parse()
1150
1151        # Some of types used in this file probably come from the file itself.
1152        # Thus it has implicit dependency on itself.
1153        self.add_dependency(self)
1154
1155    def parse(self):
1156        self.enums = []
1157        self.messages = []
1158        self.extensions = []
1159
1160        mangle_names = self.file_options.mangle_names
1161        flatten = mangle_names == nanopb_pb2.M_FLATTEN
1162        strip_prefix = None
1163        if mangle_names == nanopb_pb2.M_STRIP_PACKAGE:
1164            strip_prefix = "." + self.fdesc.package
1165
1166        def create_name(names):
1167            if mangle_names == nanopb_pb2.M_NONE:
1168                return base_name + names
1169            elif mangle_names == nanopb_pb2.M_STRIP_PACKAGE:
1170                return Names(names)
1171            else:
1172                single_name = names
1173                if isinstance(names, Names):
1174                    single_name = names.parts[-1]
1175                return Names(single_name)
1176
1177        def mangle_field_typename(typename):
1178            if mangle_names == nanopb_pb2.M_FLATTEN:
1179                return "." + typename.split(".")[-1]
1180            elif strip_prefix is not None and typename.startswith(strip_prefix):
1181                return typename[len(strip_prefix):]
1182            else:
1183                return typename
1184
1185        if self.fdesc.package:
1186            base_name = Names(self.fdesc.package.split('.'))
1187        else:
1188            base_name = Names()
1189
1190        for enum in self.fdesc.enum_type:
1191            name = create_name(enum.name)
1192            enum_options = get_nanopb_suboptions(enum, self.file_options, name)
1193            self.enums.append(Enum(name, enum, enum_options))
1194
1195        for names, message in iterate_messages(self.fdesc, flatten):
1196            name = create_name(names)
1197            message_options = get_nanopb_suboptions(message, self.file_options, name)
1198
1199            if message_options.skip_message:
1200                continue
1201
1202            message = copy.deepcopy(message)
1203            for field in message.field:
1204                if field.type in (FieldD.TYPE_MESSAGE, FieldD.TYPE_ENUM):
1205                    field.type_name = mangle_field_typename(field.type_name)
1206
1207            self.messages.append(Message(name, message, message_options))
1208            for enum in message.enum_type:
1209                name = create_name(names + enum.name)
1210                enum_options = get_nanopb_suboptions(enum, message_options, name)
1211                self.enums.append(Enum(name, enum, enum_options))
1212
1213        for names, extension in iterate_extensions(self.fdesc, flatten):
1214            name = create_name(names + extension.name)
1215            field_options = get_nanopb_suboptions(extension, self.file_options, name)
1216            if field_options.type != nanopb_pb2.FT_IGNORE:
1217                self.extensions.append(ExtensionField(name, extension, field_options))
1218
1219    def add_dependency(self, other):
1220        for enum in other.enums:
1221            self.dependencies[str(enum.names)] = enum
1222            enum.protofile = other
1223
1224        for msg in other.messages:
1225            self.dependencies[str(msg.name)] = msg
1226            msg.protofile = other
1227
1228        # Fix field default values where enum short names are used.
1229        for enum in other.enums:
1230            if not enum.options.long_names:
1231                for message in self.messages:
1232                    for field in message.all_fields():
1233                        if field.default in enum.value_longnames:
1234                            idx = enum.value_longnames.index(field.default)
1235                            field.default = enum.values[idx][0]
1236
1237        # Fix field data types where enums have negative values.
1238        for enum in other.enums:
1239            if not enum.has_negative():
1240                for message in self.messages:
1241                    for field in message.all_fields():
1242                        if field.pbtype == 'ENUM' and field.ctype == enum.names:
1243                            field.pbtype = 'UENUM'
1244
1245    def generate_header(self, includes, headername, options):
1246        '''Generate content for a header file.
1247        Generates strings, which should be concatenated and stored to file.
1248        '''
1249
1250        yield '/* Automatically generated nanopb header */\n'
1251        if options.notimestamp:
1252            yield '/* Generated by %s */\n\n' % (nanopb_version)
1253        else:
1254            yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
1255
1256        if self.fdesc.package:
1257            symbol = make_identifier(self.fdesc.package + '_' + headername)
1258        else:
1259            symbol = make_identifier(headername)
1260        yield '#ifndef PB_%s_INCLUDED\n' % symbol
1261        yield '#define PB_%s_INCLUDED\n' % symbol
1262        try:
1263            yield options.libformat % ('pb.h')
1264        except TypeError:
1265            # no %s specified - use whatever was passed in as options.libformat
1266            yield options.libformat
1267        yield '\n'
1268
1269        for incfile in includes:
1270            noext = os.path.splitext(incfile)[0]
1271            yield options.genformat % (noext + options.extension + options.header_extension)
1272            yield '\n'
1273
1274        yield '/* @@protoc_insertion_point(includes) */\n'
1275
1276        yield '#if PB_PROTO_HEADER_VERSION != 30\n'
1277        yield '#error Regenerate this file with the current version of nanopb generator.\n'
1278        yield '#endif\n'
1279        yield '\n'
1280
1281        yield '#ifdef __cplusplus\n'
1282        yield 'extern "C" {\n'
1283        yield '#endif\n\n'
1284
1285        if self.enums:
1286            yield '/* Enum definitions */\n'
1287            for enum in self.enums:
1288                yield str(enum) + '\n\n'
1289
1290        if self.messages:
1291            yield '/* Struct definitions */\n'
1292            for msg in sort_dependencies(self.messages):
1293                yield msg.types()
1294                yield str(msg) + '\n\n'
1295
1296        if self.extensions:
1297            yield '/* Extensions */\n'
1298            for extension in self.extensions:
1299                yield extension.extension_decl()
1300            yield '\n'
1301
1302        if self.messages:
1303            yield '/* Default values for struct fields */\n'
1304            for msg in self.messages:
1305                yield msg.default_decl(True)
1306            yield '\n'
1307
1308            yield '/* Initializer values for message structs */\n'
1309            for msg in self.messages:
1310                identifier = '%s_init_default' % msg.name
1311                yield '#define %-40s %s\n' % (identifier, msg.get_initializer(False))
1312            for msg in self.messages:
1313                identifier = '%s_init_zero' % msg.name
1314                yield '#define %-40s %s\n' % (identifier, msg.get_initializer(True))
1315            yield '\n'
1316
1317            yield '/* Field tags (for use in manual encoding/decoding) */\n'
1318            for msg in sort_dependencies(self.messages):
1319                for field in msg.fields:
1320                    yield field.tags()
1321            for extension in self.extensions:
1322                yield extension.tags()
1323            yield '\n'
1324
1325            yield '/* Struct field encoding specification for nanopb */\n'
1326            for msg in self.messages:
1327                yield msg.fields_declaration() + '\n'
1328            yield '\n'
1329
1330            yield '/* Maximum encoded size of messages (where known) */\n'
1331            for msg in self.messages:
1332                msize = msg.encoded_size(self.dependencies)
1333                identifier = '%s_size' % msg.name
1334                if msize is not None:
1335                    yield '#define %-40s %s\n' % (identifier, msize)
1336                else:
1337                    yield '/* %s depends on runtime parameters */\n' % identifier
1338            yield '\n'
1339
1340            yield '/* Message IDs (where set with "msgid" option) */\n'
1341
1342            yield '#ifdef PB_MSGID\n'
1343            for msg in self.messages:
1344                if hasattr(msg,'msgid'):
1345                    yield '#define PB_MSG_%d %s\n' % (msg.msgid, msg.name)
1346            yield '\n'
1347
1348            symbol = make_identifier(headername.split('.')[0])
1349            yield '#define %s_MESSAGES \\\n' % symbol
1350
1351            for msg in self.messages:
1352                m = "-1"
1353                msize = msg.encoded_size(self.dependencies)
1354                if msize is not None:
1355                    m = msize
1356                if hasattr(msg,'msgid'):
1357                    yield '\tPB_MSG(%d,%s,%s) \\\n' % (msg.msgid, m, msg.name)
1358            yield '\n'
1359
1360            for msg in self.messages:
1361                if hasattr(msg,'msgid'):
1362                    yield '#define %s_msgid %d\n' % (msg.name, msg.msgid)
1363            yield '\n'
1364
1365            yield '#endif\n\n'
1366
1367        yield '#ifdef __cplusplus\n'
1368        yield '} /* extern "C" */\n'
1369        yield '#endif\n'
1370
1371        # End of header
1372        yield '/* @@protoc_insertion_point(eof) */\n'
1373        yield '\n#endif\n'
1374
1375    def generate_source(self, headername, options):
1376        '''Generate content for a source file.'''
1377
1378        yield '/* Automatically generated nanopb constant definitions */\n'
1379        if options.notimestamp:
1380            yield '/* Generated by %s */\n\n' % (nanopb_version)
1381        else:
1382            yield '/* Generated by %s at %s. */\n\n' % (nanopb_version, time.asctime())
1383        yield options.genformat % (headername)
1384        yield '\n'
1385        yield '/* @@protoc_insertion_point(includes) */\n'
1386
1387        yield '#if PB_PROTO_HEADER_VERSION != 30\n'
1388        yield '#error Regenerate this file with the current version of nanopb generator.\n'
1389        yield '#endif\n'
1390        yield '\n'
1391
1392        for msg in self.messages:
1393            yield msg.default_decl(False)
1394
1395        yield '\n\n'
1396
1397        for msg in self.messages:
1398            yield msg.fields_definition() + '\n\n'
1399
1400        for ext in self.extensions:
1401            yield ext.extension_def() + '\n'
1402
1403        for enum in self.enums:
1404            yield enum.enum_to_string_definition() + '\n'
1405
1406        # Add checks for numeric limits
1407        if self.messages:
1408            largest_msg = max(self.messages, key = lambda m: m.count_required_fields())
1409            largest_count = largest_msg.count_required_fields()
1410            if largest_count > 64:
1411                yield '\n/* Check that missing required fields will be properly detected */\n'
1412                yield '#if PB_MAX_REQUIRED_FIELDS < %d\n' % largest_count
1413                yield '#error Properly detecting missing required fields in %s requires \\\n' % largest_msg.name
1414                yield '       setting PB_MAX_REQUIRED_FIELDS to %d or more.\n' % largest_count
1415                yield '#endif\n'
1416
1417        max_field = FieldMaxSize()
1418        checks_msgnames = []
1419        for msg in self.messages:
1420            checks_msgnames.append(msg.name)
1421            for field in msg.fields:
1422                max_field.extend(field.largest_field_value())
1423        for field in self.extensions:
1424            max_field.extend(field.largest_field_value())
1425
1426        worst = max_field.worst
1427        worst_field = max_field.worst_field
1428        checks = max_field.checks
1429
1430        if worst > 255 or checks:
1431            yield '\n/* Check that field information fits in pb_field_t */\n'
1432
1433            if worst > 65535 or checks:
1434                yield '#if !defined(PB_FIELD_32BIT)\n'
1435                if worst > 65535:
1436                    yield '#error Field descriptor for %s is too large. Define PB_FIELD_32BIT to fix this.\n' % worst_field
1437                else:
1438                    assertion = ' && '.join(str(c) + ' < 65536' for c in checks)
1439                    msgs = '_'.join(str(n) for n in checks_msgnames)
1440                    yield '/* If you get an error here, it means that you need to define PB_FIELD_32BIT\n'
1441                    yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1442                    yield ' *\n'
1443                    yield ' * The reason you need to do this is that some of your messages contain tag\n'
1444                    yield ' * numbers or field sizes that are larger than what can fit in 8 or 16 bit\n'
1445                    yield ' * field descriptors.\n'
1446                    yield ' */\n'
1447                    yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_32BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1448                yield '#endif\n\n'
1449
1450            if worst < 65536:
1451                yield '#if !defined(PB_FIELD_16BIT) && !defined(PB_FIELD_32BIT)\n'
1452                if worst > 255:
1453                    yield '#error Field descriptor for %s is too large. Define PB_FIELD_16BIT to fix this.\n' % worst_field
1454                else:
1455                    assertion = ' && '.join(str(c) + ' < 256' for c in checks)
1456                    msgs = '_'.join(str(n) for n in checks_msgnames)
1457                    yield '/* If you get an error here, it means that you need to define PB_FIELD_16BIT\n'
1458                    yield ' * compile-time option. You can do that in pb.h or on compiler command line.\n'
1459                    yield ' *\n'
1460                    yield ' * The reason you need to do this is that some of your messages contain tag\n'
1461                    yield ' * numbers or field sizes that are larger than what can fit in the default\n'
1462                    yield ' * 8 bit descriptors.\n'
1463                    yield ' */\n'
1464                    yield 'PB_STATIC_ASSERT((%s), YOU_MUST_DEFINE_PB_FIELD_16BIT_FOR_MESSAGES_%s)\n'%(assertion,msgs)
1465                yield '#endif\n\n'
1466
1467        # Add check for sizeof(double)
1468        has_double = False
1469        for msg in self.messages:
1470            for field in msg.all_fields():
1471                if field.ctype == 'double':
1472                    has_double = True
1473
1474        if has_double:
1475            yield '\n'
1476            yield '/* On some platforms (such as AVR), double is really float.\n'
1477            yield ' * These are not directly supported by nanopb, but see example_avr_double.\n'
1478            yield ' * To get rid of this error, remove any double fields from your .proto.\n'
1479            yield ' */\n'
1480            yield 'PB_STATIC_ASSERT(sizeof(double) == 8, DOUBLE_MUST_BE_8_BYTES)\n'
1481
1482        yield '\n'
1483        yield '/* @@protoc_insertion_point(eof) */\n'
1484
1485# ---------------------------------------------------------------------------
1486#                    Options parsing for the .proto files
1487# ---------------------------------------------------------------------------
1488
1489from fnmatch import fnmatch
1490
1491def read_options_file(infile):
1492    '''Parse a separate options file to list:
1493        [(namemask, options), ...]
1494    '''
1495    results = []
1496    data = infile.read()
1497    data = re.sub('/\*.*?\*/', '', data, flags = re.MULTILINE)
1498    data = re.sub('//.*?$', '', data, flags = re.MULTILINE)
1499    data = re.sub('#.*?$', '', data, flags = re.MULTILINE)
1500    for i, line in enumerate(data.split('\n')):
1501        line = line.strip()
1502        if not line:
1503            continue
1504
1505        parts = line.split(None, 1)
1506
1507        if len(parts) < 2:
1508            sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1509                             "Option lines should have space between field name and options. " +
1510                             "Skipping line: '%s'\n" % line)
1511            continue
1512
1513        opts = nanopb_pb2.NanoPBOptions()
1514
1515        try:
1516            text_format.Merge(parts[1], opts)
1517        except Exception as e:
1518            sys.stderr.write("%s:%d: " % (infile.name, i + 1) +
1519                             "Unparseable option line: '%s'. " % line +
1520                             "Error: %s\n" % str(e))
1521            continue
1522        results.append((parts[0], opts))
1523
1524    return results
1525
1526class Globals:
1527    '''Ugly global variables, should find a good way to pass these.'''
1528    verbose_options = False
1529    separate_options = []
1530    matched_namemasks = set()
1531
1532def get_nanopb_suboptions(subdesc, options, name):
1533    '''Get copy of options, and merge information from subdesc.'''
1534    new_options = nanopb_pb2.NanoPBOptions()
1535    new_options.CopyFrom(options)
1536
1537    if hasattr(subdesc, 'syntax') and subdesc.syntax == "proto3":
1538        new_options.proto3 = True
1539
1540    # Handle options defined in a separate file
1541    dotname = '.'.join(name.parts)
1542    for namemask, options in Globals.separate_options:
1543        if fnmatch(dotname, namemask):
1544            Globals.matched_namemasks.add(namemask)
1545            new_options.MergeFrom(options)
1546
1547    # Handle options defined in .proto
1548    if isinstance(subdesc.options, descriptor.FieldOptions):
1549        ext_type = nanopb_pb2.nanopb
1550    elif isinstance(subdesc.options, descriptor.FileOptions):
1551        ext_type = nanopb_pb2.nanopb_fileopt
1552    elif isinstance(subdesc.options, descriptor.MessageOptions):
1553        ext_type = nanopb_pb2.nanopb_msgopt
1554    elif isinstance(subdesc.options, descriptor.EnumOptions):
1555        ext_type = nanopb_pb2.nanopb_enumopt
1556    else:
1557        raise Exception("Unknown options type")
1558
1559    if subdesc.options.HasExtension(ext_type):
1560        ext = subdesc.options.Extensions[ext_type]
1561        new_options.MergeFrom(ext)
1562
1563    if Globals.verbose_options:
1564        sys.stderr.write("Options for " + dotname + ": ")
1565        sys.stderr.write(text_format.MessageToString(new_options) + "\n")
1566
1567    return new_options
1568
1569
1570# ---------------------------------------------------------------------------
1571#                         Command line interface
1572# ---------------------------------------------------------------------------
1573
1574import sys
1575import os.path
1576from optparse import OptionParser
1577
1578optparser = OptionParser(
1579    usage = "Usage: nanopb_generator.py [options] file.pb ...",
1580    epilog = "Compile file.pb from file.proto by: 'protoc -ofile.pb file.proto'. " +
1581             "Output will be written to file.pb.h and file.pb.c.")
1582optparser.add_option("--version", dest="version", action="store_true",
1583    help="Show version info and exit")
1584optparser.add_option("-x", dest="exclude", metavar="FILE", action="append", default=[],
1585    help="Exclude file from generated #include list.")
1586optparser.add_option("-e", "--extension", dest="extension", metavar="EXTENSION", default=".pb",
1587    help="Set extension to use instead of '.pb' for generated files. [default: %default]")
1588optparser.add_option("-H", "--header-extension", dest="header_extension", metavar="EXTENSION", default=".h",
1589    help="Set extension to use for generated header files. [default: %default]")
1590optparser.add_option("-S", "--source-extension", dest="source_extension", metavar="EXTENSION", default=".c",
1591    help="Set extension to use for generated source files. [default: %default]")
1592optparser.add_option("-f", "--options-file", dest="options_file", metavar="FILE", default="%s.options",
1593    help="Set name of a separate generator options file.")
1594optparser.add_option("-I", "--options-path", dest="options_path", metavar="DIR",
1595    action="append", default = [],
1596    help="Search for .options files additionally in this path")
1597optparser.add_option("-D", "--output-dir", dest="output_dir",
1598                     metavar="OUTPUTDIR", default=None,
1599                     help="Output directory of .pb.h and .pb.c files")
1600optparser.add_option("-Q", "--generated-include-format", dest="genformat",
1601    metavar="FORMAT", default='#include "%s"\n',
1602    help="Set format string to use for including other .pb.h files. [default: %default]")
1603optparser.add_option("-L", "--library-include-format", dest="libformat",
1604    metavar="FORMAT", default='#include <%s>\n',
1605    help="Set format string to use for including the nanopb pb.h header. [default: %default]")
1606optparser.add_option("--strip-path", dest="strip_path", action="store_true", default=True,
1607    help="Strip directory path from #included .pb.h file name [default: %default]")
1608optparser.add_option("--no-strip-path", dest="strip_path", action="store_false",
1609    help="Opposite of --strip-path")
1610optparser.add_option("-T", "--no-timestamp", dest="notimestamp", action="store_true", default=False,
1611    help="Don't add timestamp to .pb.h and .pb.c preambles")
1612optparser.add_option("-q", "--quiet", dest="quiet", action="store_true", default=False,
1613    help="Don't print anything except errors.")
1614optparser.add_option("-v", "--verbose", dest="verbose", action="store_true", default=False,
1615    help="Print more information.")
1616optparser.add_option("-s", dest="settings", metavar="OPTION:VALUE", action="append", default=[],
1617    help="Set generator option (max_size, max_count etc.).")
1618
1619def parse_file(filename, fdesc, options):
1620    '''Parse a single file. Returns a ProtoFile instance.'''
1621    toplevel_options = nanopb_pb2.NanoPBOptions()
1622    for s in options.settings:
1623        text_format.Merge(s, toplevel_options)
1624
1625    if not fdesc:
1626        data = open(filename, 'rb').read()
1627        fdesc = descriptor.FileDescriptorSet.FromString(data).file[0]
1628
1629    # Check if there is a separate .options file
1630    had_abspath = False
1631    try:
1632        optfilename = options.options_file % os.path.splitext(filename)[0]
1633    except TypeError:
1634        # No %s specified, use the filename as-is
1635        optfilename = options.options_file
1636        had_abspath = True
1637
1638    paths = ['.'] + options.options_path
1639    for p in paths:
1640        if os.path.isfile(os.path.join(p, optfilename)):
1641            optfilename = os.path.join(p, optfilename)
1642            if options.verbose:
1643                sys.stderr.write('Reading options from ' + optfilename + '\n')
1644            Globals.separate_options = read_options_file(open(optfilename, 'r', encoding = 'utf-8'))
1645            break
1646    else:
1647        # If we are given a full filename and it does not exist, give an error.
1648        # However, don't give error when we automatically look for .options file
1649        # with the same name as .proto.
1650        if options.verbose or had_abspath:
1651            sys.stderr.write('Options file not found: ' + optfilename + '\n')
1652        Globals.separate_options = []
1653
1654    Globals.matched_namemasks = set()
1655
1656    # Parse the file
1657    file_options = get_nanopb_suboptions(fdesc, toplevel_options, Names([filename]))
1658    f = ProtoFile(fdesc, file_options)
1659    f.optfilename = optfilename
1660
1661    return f
1662
1663def process_file(filename, fdesc, options, other_files = {}):
1664    '''Process a single file.
1665    filename: The full path to the .proto or .pb source file, as string.
1666    fdesc: The loaded FileDescriptorSet, or None to read from the input file.
1667    options: Command line options as they come from OptionsParser.
1668
1669    Returns a dict:
1670        {'headername': Name of header file,
1671         'headerdata': Data for the .h header file,
1672         'sourcename': Name of the source code file,
1673         'sourcedata': Data for the .c source code file
1674        }
1675    '''
1676    f = parse_file(filename, fdesc, options)
1677
1678    # Provide dependencies if available
1679    for dep in f.fdesc.dependency:
1680        if dep in other_files:
1681            f.add_dependency(other_files[dep])
1682
1683    # Decide the file names
1684    noext = os.path.splitext(filename)[0]
1685    headername = noext + options.extension + options.header_extension
1686    sourcename = noext + options.extension + options.source_extension
1687
1688    if options.strip_path:
1689        headerbasename = os.path.basename(headername)
1690    else:
1691        headerbasename = headername
1692
1693    # List of .proto files that should not be included in the C header file
1694    # even if they are mentioned in the source .proto.
1695    excludes = ['nanopb.proto', 'google/protobuf/descriptor.proto'] + options.exclude
1696    includes = [d for d in f.fdesc.dependency if d not in excludes]
1697
1698    headerdata = ''.join(f.generate_header(includes, headerbasename, options))
1699    sourcedata = ''.join(f.generate_source(headerbasename, options))
1700
1701    # Check if there were any lines in .options that did not match a member
1702    unmatched = [n for n,o in Globals.separate_options if n not in Globals.matched_namemasks]
1703    if unmatched and not options.quiet:
1704        sys.stderr.write("Following patterns in " + f.optfilename + " did not match any fields: "
1705                         + ', '.join(unmatched) + "\n")
1706        if not Globals.verbose_options:
1707            sys.stderr.write("Use  protoc --nanopb-out=-v:.   to see a list of the field names.\n")
1708
1709    return {'headername': headername, 'headerdata': headerdata,
1710            'sourcename': sourcename, 'sourcedata': sourcedata}
1711
1712def main_cli():
1713    '''Main function when invoked directly from the command line.'''
1714
1715    options, filenames = optparser.parse_args()
1716
1717    if options.version:
1718        print(nanopb_version)
1719        sys.exit(0)
1720
1721    if not filenames:
1722        optparser.print_help()
1723        sys.exit(1)
1724
1725    if options.quiet:
1726        options.verbose = False
1727
1728    if options.output_dir and not os.path.exists(options.output_dir):
1729        optparser.print_help()
1730        sys.stderr.write("\noutput_dir does not exist: %s\n" % options.output_dir)
1731        sys.exit(1)
1732
1733    if options.verbose:
1734        sys.stderr.write("Nanopb version %s\n" % nanopb_version)
1735        sys.stderr.write('Google Python protobuf library imported from %s, version %s\n'
1736                         % (google.protobuf.__file__, google.protobuf.__version__))
1737
1738    Globals.verbose_options = options.verbose
1739    for filename in filenames:
1740        results = process_file(filename, None, options)
1741
1742        base_dir = options.output_dir or ''
1743        to_write = [
1744            (os.path.join(base_dir, results['headername']), results['headerdata']),
1745            (os.path.join(base_dir, results['sourcename']), results['sourcedata']),
1746        ]
1747
1748        if not options.quiet:
1749            paths = " and ".join([x[0] for x in to_write])
1750            sys.stderr.write("Writing to %s\n" % paths)
1751
1752        for path, data in to_write:
1753            with open(path, 'w') as f:
1754                f.write(data)
1755
1756def main_plugin():
1757    '''Main function when invoked as a protoc plugin.'''
1758
1759    import io, sys
1760    if sys.platform == "win32":
1761        import os, msvcrt
1762        # Set stdin and stdout to binary mode
1763        msvcrt.setmode(sys.stdin.fileno(), os.O_BINARY)
1764        msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)
1765
1766    data = io.open(sys.stdin.fileno(), "rb").read()
1767
1768    request = plugin_pb2.CodeGeneratorRequest.FromString(data)
1769
1770    try:
1771        # Versions of Python prior to 2.7.3 do not support unicode
1772        # input to shlex.split(). Try to convert to str if possible.
1773        params = str(request.parameter)
1774    except UnicodeEncodeError:
1775        params = request.parameter
1776
1777    import shlex
1778    args = shlex.split(params)
1779
1780    if len(args) == 1 and ',' in args[0]:
1781        # For compatibility with other protoc plugins, support options
1782        # separated by comma.
1783        lex = shlex.shlex(params)
1784        lex.whitespace_split = True
1785        lex.whitespace = ','
1786        args = list(lex)
1787
1788    optparser.usage = "Usage: protoc --nanopb_out=[options][,more_options]:outdir file.proto"
1789    optparser.epilog = "Output will be written to file.pb.h and file.pb.c."
1790
1791    if '-h' in args or '--help' in args:
1792        # By default optparser prints help to stdout, which doesn't work for
1793        # protoc plugins.
1794        optparser.print_help(sys.stderr)
1795        sys.exit(1)
1796
1797    options, dummy = optparser.parse_args(args)
1798
1799    if options.version:
1800        sys.stderr.write('%s\n' % (nanopb_version))
1801        sys.exit(0)
1802
1803    Globals.verbose_options = options.verbose
1804
1805    if options.verbose:
1806        sys.stderr.write("Nanopb version %s\n" % nanopb_version)
1807        sys.stderr.write('Google Python protobuf library imported from %s, version %s\n'
1808                         % (google.protobuf.__file__, google.protobuf.__version__))
1809
1810    response = plugin_pb2.CodeGeneratorResponse()
1811
1812    # Google's protoc does not currently indicate the full path of proto files.
1813    # Instead always add the main file path to the search dirs, that works for
1814    # the common case.
1815    import os.path
1816    options.options_path.append(os.path.dirname(request.file_to_generate[0]))
1817
1818    # Process any include files first, in order to have them
1819    # available as dependencies
1820    other_files = {}
1821    for fdesc in request.proto_file:
1822        other_files[fdesc.name] = parse_file(fdesc.name, fdesc, options)
1823
1824    for filename in request.file_to_generate:
1825        for fdesc in request.proto_file:
1826            if fdesc.name == filename:
1827                results = process_file(filename, fdesc, options, other_files)
1828
1829                f = response.file.add()
1830                f.name = results['headername']
1831                f.content = results['headerdata']
1832
1833                f = response.file.add()
1834                f.name = results['sourcename']
1835                f.content = results['sourcedata']
1836
1837    io.open(sys.stdout.fileno(), "wb").write(response.SerializeToString())
1838
1839if __name__ == '__main__':
1840    # Check if we are running as a plugin under protoc
1841    if 'protoc-gen-' in sys.argv[0] or '--protoc-plugin' in sys.argv:
1842        main_plugin()
1843    else:
1844        main_cli()
1845