components/policy/tools/syntax_check_policy_template_json.py

   1 #!/usr/bin/env python
   2 # Copyright (c) 2012 The Chromium Authors. All rights reserved.
   3 # Use of this source code is governed by a BSD-style license that can be
   4 # found in the LICENSE file.
   5
   6 '''
   7 Checks a policy_templates.json file for conformity to its syntax specification.
   8 '''
   9
  10 import json
  11 import optparse
  12 import os
  13 import re
  14 import sys
  15
  16
  17 LEADING_WHITESPACE = re.compile('^([ \t]*)')
  18 TRAILING_WHITESPACE = re.compile('.*?([ \t]+)$')
  19 # Matches all non-empty strings that contain no whitespaces.
  20 NO_WHITESPACE = re.compile('[^\s]+$')
  21
  22 # Convert a 'type' to the schema types it may be converted to.
  23 # The 'dict' type represents structured JSON data, and can be converted
  24 # to an 'object' or an 'array'.
  25 TYPE_TO_SCHEMA = {
  26   'int': [ 'integer' ],
  27   'list': [ 'array' ],
  28   'dict': [ 'object', 'array' ],
  29   'main': [ 'boolean' ],
  30   'string': [ 'string' ],
  31   'int-enum': [ 'integer' ],
  32   'string-enum': [ 'string' ],
  33   'external': [ 'object' ],
  34 }
  35
  36 # List of boolean policies that have been introduced with negative polarity in
  37 # the past and should not trigger the negative polarity check.
  38 LEGACY_INVERTED_POLARITY_WHITELIST = [
  39     'DeveloperToolsDisabled',
  40     'DeviceAutoUpdateDisabled',
  41     'Disable3DAPIs',
  42     'DisableAuthNegotiateCnameLookup',
  43     'DisablePluginFinder',
  44     'DisablePrintPreview',
  45     'DisableSafeBrowsingProceedAnyway',
  46     'DisableScreenshots',
  47     'DisableSpdy',
  48     'DisableSSLRecordSplitting',
  49     'DriveDisabled',
  50     'DriveDisabledOverCellular',
  51     'ExternalStorageDisabled',
  52     'SavingBrowserHistoryDisabled',
  53     'SyncDisabled',
  54 ]
  55
  56 class PolicyTemplateChecker(object):
  57
  58   def __init__(self):
  59     self.error_count = 0
  60     self.warning_count = 0
  61     self.num_policies = 0
  62     self.num_groups = 0
  63     self.num_policies_in_groups = 0
  64     self.options = None
  65     self.features = []
  66
  67   def _Error(self, message, parent_element=None, identifier=None,
  68              offending_snippet=None):
  69     self.error_count += 1
  70     error = ''
  71     if identifier is not None and parent_element is not None:
  72       error += 'In %s %s: ' % (parent_element, identifier)
  73     print error + 'Error: ' + message
  74     if offending_snippet is not None:
  75       print '  Offending:', json.dumps(offending_snippet, indent=2)
  76
  77   def _CheckContains(self, container, key, value_type,
  78                      optional=False,
  79                      parent_element='policy',
  80                      container_name=None,
  81                      identifier=None,
  82                      offending='__CONTAINER__',
  83                      regexp_check=None):
  84     '''
  85     Checks |container| for presence of |key| with value of type |value_type|.
  86     If |value_type| is string and |regexp_check| is specified, then an error is
  87     reported when the value does not match the regular expression object.
  88
  89     |value_type| can also be a list, if more than one type is supported.
  90
  91     The other parameters are needed to generate, if applicable, an appropriate
  92     human-readable error message of the following form:
  93
  94     In |parent_element| |identifier|:
  95       (if the key is not present):
  96       Error: |container_name| must have a |value_type| named |key|.
  97       Offending snippet: |offending| (if specified; defaults to |container|)
  98       (if the value does not have the required type):
  99       Error: Value of |key| must be a |value_type|.
 100       Offending snippet: |container[key]|
 101
 102     Returns: |container[key]| if the key is present, None otherwise.
 103     '''
 104     if identifier is None:
 105       try:
 106         identifier = container.get('name')
 107       except:
 108         self._Error('Cannot access container name of "%s".' % container_name)
 109         return None
 110     if container_name is None:
 111       container_name = parent_element
 112     if offending == '__CONTAINER__':
 113       offending = container
 114     if key not in container:
 115       if optional:
 116         return
 117       else:
 118         self._Error('%s must have a %s "%s".' %
 119                     (container_name.title(), value_type.__name__, key),
 120                     container_name, identifier, offending)
 121       return None
 122     value = container[key]
 123     value_types = value_type if isinstance(value_type, list) else [ value_type ]
 124     if not any(isinstance(value, type) for type in value_types):
 125       self._Error('Value of "%s" must one of [ %s ].' %
 126                   (key, ', '.join([type.__name__ for type in value_types])),
 127                   container_name, identifier, value)
 128     if str in value_types and regexp_check and not regexp_check.match(value):
 129       self._Error('Value of "%s" must match "%s".' %
 130                   (key, regexp_check.pattern),
 131                   container_name, identifier, value)
 132     return value
 133
 134   def _AddPolicyID(self, id, policy_ids, policy):
 135     '''
 136     Adds |id| to |policy_ids|. Generates an error message if the
 137     |id| exists already; |policy| is needed for this message.
 138     '''
 139     if id in policy_ids:
 140       self._Error('Duplicate id', 'policy', policy.get('name'),
 141                   id)
 142     else:
 143       policy_ids.add(id)
 144
 145   def _CheckPolicyIDs(self, policy_ids):
 146     '''
 147     Checks a set of policy_ids to make sure it contains a continuous range
 148     of entries (i.e. no holes).
 149     Holes would not be a technical problem, but we want to ensure that nobody
 150     accidentally omits IDs.
 151     '''
 152     for i in range(len(policy_ids)):
 153       if (i + 1) not in policy_ids:
 154         self._Error('No policy with id: %s' % (i + 1))
 155
 156   def _CheckPolicySchema(self, policy, policy_type):
 157     '''Checks that the 'schema' field matches the 'type' field.'''
 158     self._CheckContains(policy, 'schema', dict)
 159     if isinstance(policy.get('schema'), dict):
 160       self._CheckContains(policy['schema'], 'type', str)
 161       schema_type = policy['schema'].get('type')
 162       if schema_type not in TYPE_TO_SCHEMA[policy_type]:
 163         self._Error('Schema type must match the existing type for policy %s' %
 164                     policy.get('name'))
 165
 166       # Checks that boolean policies are not negated (which makes them harder to
 167       # reason about).
 168       if (schema_type == 'boolean' and
 169           'disable' in policy.get('name').lower() and
 170           policy.get('name') not in LEGACY_INVERTED_POLARITY_WHITELIST):
 171         self._Error(('Boolean policy %s uses negative polarity, please make ' +
 172                      'new boolean policies follow the XYZEnabled pattern. ' +
 173                      'See also http://crbug.com/85687') % policy.get('name'))
 174
 175
 176   def _CheckPolicy(self, policy, is_in_group, policy_ids):
 177     if not isinstance(policy, dict):
 178       self._Error('Each policy must be a dictionary.', 'policy', None, policy)
 179       return
 180
 181     # There should not be any unknown keys in |policy|.
 182     for key in policy:
 183       if key not in ('name', 'type', 'caption', 'desc', 'device_only',
 184                      'supported_on', 'label', 'policies', 'items',
 185                      'example_value', 'features', 'deprecated', 'future',
 186                      'id', 'schema', 'max_size'):
 187         self.warning_count += 1
 188         print ('In policy %s: Warning: Unknown key: %s' %
 189                (policy.get('name'), key))
 190
 191     # Each policy must have a name.
 192     self._CheckContains(policy, 'name', str, regexp_check=NO_WHITESPACE)
 193
 194     # Each policy must have a type.
 195     policy_types = ('group', 'main', 'string', 'int', 'list', 'int-enum',
 196                     'string-enum', 'dict', 'external')
 197     policy_type = self._CheckContains(policy, 'type', str)
 198     if policy_type not in policy_types:
 199       self._Error('Policy type must be one of: ' + ', '.join(policy_types),
 200                   'policy', policy.get('name'), policy_type)
 201       return  # Can't continue for unsupported type.
 202
 203     # Each policy must have a caption message.
 204     self._CheckContains(policy, 'caption', str)
 205
 206     # Each policy must have a description message.
 207     self._CheckContains(policy, 'desc', str)
 208
 209     # If 'label' is present, it must be a string.
 210     self._CheckContains(policy, 'label', str, True)
 211
 212     # If 'deprecated' is present, it must be a bool.
 213     self._CheckContains(policy, 'deprecated', bool, True)
 214
 215     # If 'future' is present, it must be a bool.
 216     self._CheckContains(policy, 'future', bool, True)
 217
 218     if policy_type == 'group':
 219       # Groups must not be nested.
 220       if is_in_group:
 221         self._Error('Policy groups must not be nested.', 'policy', policy)
 222
 223       # Each policy group must have a list of policies.
 224       policies = self._CheckContains(policy, 'policies', list)
 225
 226       # Check sub-policies.
 227       if policies is not None:
 228         for nested_policy in policies:
 229           self._CheckPolicy(nested_policy, True, policy_ids)
 230
 231       # Groups must not have an |id|.
 232       if 'id' in policy:
 233         self._Error('Policies of type "group" must not have an "id" field.',
 234                     'policy', policy)
 235
 236       # Statistics.
 237       self.num_groups += 1
 238
 239     else:  # policy_type != group
 240       # Each policy must have a protobuf ID.
 241       id = self._CheckContains(policy, 'id', int)
 242       self._AddPolicyID(id, policy_ids, policy)
 243
 244       # 'schema' is the new 'type'.
 245       # TODO(joaodasilva): remove the 'type' checks once 'schema' is used
 246       # everywhere.
 247       self._CheckPolicySchema(policy, policy_type)
 248
 249       # Each policy must have a supported_on list.
 250       supported_on = self._CheckContains(policy, 'supported_on', list)
 251       if supported_on is not None:
 252         for s in supported_on:
 253           if not isinstance(s, str):
 254             self._Error('Entries in "supported_on" must be strings.', 'policy',
 255                         policy, supported_on)
 256
 257       # Each policy must have a 'features' dict.
 258       features = self._CheckContains(policy, 'features', dict)
 259
 260       # All the features must have a documenting message.
 261       if features:
 262         for feature in features:
 263           if not feature in self.features:
 264             self._Error('Unknown feature "%s". Known features must have a '
 265                         'documentation string in the messages dictionary.' %
 266                         feature, 'policy', policy.get('name', policy))
 267
 268       # All user policies must have a per_profile feature flag.
 269       if (not policy.get('device_only', False) and
 270           not policy.get('deprecated', False) and
 271           not filter(re.compile('^chrome_frame:.*').match, supported_on)):
 272         self._CheckContains(features, 'per_profile', bool,
 273                             container_name='features',
 274                             identifier=policy.get('name'))
 275
 276       # All policies must declare whether they allow changes at runtime.
 277       self._CheckContains(features, 'dynamic_refresh', bool,
 278                           container_name='features',
 279                           identifier=policy.get('name'))
 280
 281       # Each policy must have an 'example_value' of appropriate type.
 282       if policy_type == 'main':
 283         value_type = bool
 284       elif policy_type in ('string', 'string-enum'):
 285         value_type = str
 286       elif policy_type in ('int', 'int-enum'):
 287         value_type = int
 288       elif policy_type == 'list':
 289         value_type = list
 290       elif policy_type == 'external':
 291         value_type = dict
 292       elif policy_type == 'dict':
 293         value_type = [ dict, list ]
 294       else:
 295         raise NotImplementedError('Unimplemented policy type: %s' % policy_type)
 296       self._CheckContains(policy, 'example_value', value_type)
 297
 298       # Statistics.
 299       self.num_policies += 1
 300       if is_in_group:
 301         self.num_policies_in_groups += 1
 302
 303     if policy_type in ('int-enum', 'string-enum'):
 304       # Enums must contain a list of items.
 305       items = self._CheckContains(policy, 'items', list)
 306       if items is not None:
 307         if len(items) < 1:
 308           self._Error('"items" must not be empty.', 'policy', policy, items)
 309         for item in items:
 310           # Each item must have a name.
 311           # Note: |policy.get('name')| is used instead of |policy['name']|
 312           # because it returns None rather than failing when no key called
 313           # 'name' exists.
 314           self._CheckContains(item, 'name', str, container_name='item',
 315                               identifier=policy.get('name'),
 316                               regexp_check=NO_WHITESPACE)
 317
 318           # Each item must have a value of the correct type.
 319           self._CheckContains(item, 'value', value_type, container_name='item',
 320                               identifier=policy.get('name'))
 321
 322           # Each item must have a caption.
 323           self._CheckContains(item, 'caption', str, container_name='item',
 324                               identifier=policy.get('name'))
 325
 326     if policy_type == 'external':
 327       # Each policy referencing external data must specify a maximum data size.
 328       self._CheckContains(policy, 'max_size', int)
 329
 330   def _CheckMessage(self, key, value):
 331     # |key| must be a string, |value| a dict.
 332     if not isinstance(key, str):
 333       self._Error('Each message key must be a string.', 'message', key, key)
 334       return
 335
 336     if not isinstance(value, dict):
 337       self._Error('Each message must be a dictionary.', 'message', key, value)
 338       return
 339
 340     # Each message must have a desc.
 341     self._CheckContains(value, 'desc', str, parent_element='message',
 342                         identifier=key)
 343
 344     # Each message must have a text.
 345     self._CheckContains(value, 'text', str, parent_element='message',
 346                         identifier=key)
 347
 348     # There should not be any unknown keys in |value|.
 349     for vkey in value:
 350       if vkey not in ('desc', 'text'):
 351         self.warning_count += 1
 352         print 'In message %s: Warning: Unknown key: %s' % (key, vkey)
 353
 354   def _LeadingWhitespace(self, line):
 355     match = LEADING_WHITESPACE.match(line)
 356     if match:
 357       return match.group(1)
 358     return ''
 359
 360   def _TrailingWhitespace(self, line):
 361     match = TRAILING_WHITESPACE.match(line)
 362     if match:
 363       return match.group(1)
 364     return ''
 365
 366   def _LineError(self, message, line_number):
 367     self.error_count += 1
 368     print 'In line %d: Error: %s' % (line_number, message)
 369
 370   def _LineWarning(self, message, line_number):
 371     self.warning_count += 1
 372     print ('In line %d: Warning: Automatically fixing formatting: %s'
 373            % (line_number, message))
 374
 375   def _CheckFormat(self, filename):
 376     if self.options.fix:
 377       fixed_lines = []
 378     with open(filename) as f:
 379       indent = 0
 380       line_number = 0
 381       for line in f:
 382         line_number += 1
 383         line = line.rstrip('\n')
 384         # Check for trailing whitespace.
 385         trailing_whitespace = self._TrailingWhitespace(line)
 386         if len(trailing_whitespace) > 0:
 387           if self.options.fix:
 388             line = line.rstrip()
 389             self._LineWarning('Trailing whitespace.', line_number)
 390           else:
 391             self._LineError('Trailing whitespace.', line_number)
 392         if self.options.fix:
 393           if len(line) == 0:
 394             fixed_lines += ['\n']
 395             continue
 396         else:
 397           if line == trailing_whitespace:
 398             # This also catches the case of an empty line.
 399             continue
 400         # Check for correct amount of leading whitespace.
 401         leading_whitespace = self._LeadingWhitespace(line)
 402         if leading_whitespace.count('\t') > 0:
 403           if self.options.fix:
 404             leading_whitespace = leading_whitespace.replace('\t', '  ')
 405             line = leading_whitespace + line.lstrip()
 406             self._LineWarning('Tab character found.', line_number)
 407           else:
 408             self._LineError('Tab character found.', line_number)
 409         if line[len(leading_whitespace)] in (']', '}'):
 410           indent -= 2
 411         if line[0] != '#':  # Ignore 0-indented comments.
 412           if len(leading_whitespace) != indent:
 413             if self.options.fix:
 414               line = ' ' * indent + line.lstrip()
 415               self._LineWarning('Indentation should be ' + str(indent) +
 416                                 ' spaces.', line_number)
 417             else:
 418               self._LineError('Bad indentation. Should be ' + str(indent) +
 419                               ' spaces.', line_number)
 420         if line[-1] in ('[', '{'):
 421           indent += 2
 422         if self.options.fix:
 423           fixed_lines.append(line + '\n')
 424
 425     # If --fix is specified: backup the file (deleting any existing backup),
 426     # then write the fixed version with the old filename.
 427     if self.options.fix:
 428       if self.options.backup:
 429         backupfilename = filename + '.bak'
 430         if os.path.exists(backupfilename):
 431           os.remove(backupfilename)
 432         os.rename(filename, backupfilename)
 433       with open(filename, 'w') as f:
 434         f.writelines(fixed_lines)
 435
 436   def Main(self, filename, options):
 437     try:
 438       with open(filename) as f:
 439         data = eval(f.read())
 440     except:
 441       import traceback
 442       traceback.print_exc(file=sys.stdout)
 443       self._Error('Invalid Python/JSON syntax.')
 444       return 1
 445     if data == None:
 446       self._Error('Invalid Python/JSON syntax.')
 447       return 1
 448     self.options = options
 449
 450     # First part: check JSON structure.
 451
 452     # Check (non-policy-specific) message definitions.
 453     messages = self._CheckContains(data, 'messages', dict,
 454                                    parent_element=None,
 455                                    container_name='The root element',
 456                                    offending=None)
 457     if messages is not None:
 458       for message in messages:
 459         self._CheckMessage(message, messages[message])
 460         if message.startswith('doc_feature_'):
 461           self.features.append(message[12:])
 462
 463     # Check policy definitions.
 464     policy_definitions = self._CheckContains(data, 'policy_definitions', list,
 465                                              parent_element=None,
 466                                              container_name='The root element',
 467                                              offending=None)
 468     if policy_definitions is not None:
 469       policy_ids = set()
 470       for policy in policy_definitions:
 471         self._CheckPolicy(policy, False, policy_ids)
 472       self._CheckPolicyIDs(policy_ids)
 473
 474     # Second part: check formatting.
 475     self._CheckFormat(filename)
 476
 477     # Third part: summary and exit.
 478     print ('Finished checking %s. %d errors, %d warnings.' %
 479         (filename, self.error_count, self.warning_count))
 480     if self.options.stats:
 481       if self.num_groups > 0:
 482         print ('%d policies, %d of those in %d groups (containing on '
 483                'average %.1f policies).' %
 484                (self.num_policies, self.num_policies_in_groups, self.num_groups,
 485                  (1.0 * self.num_policies_in_groups / self.num_groups)))
 486       else:
 487         print self.num_policies, 'policies, 0 policy groups.'
 488     if self.error_count > 0:
 489       return 1
 490     return 0
 491
 492   def Run(self, argv, filename=None):
 493     parser = optparse.OptionParser(
 494         usage='usage: %prog [options] filename',
 495         description='Syntax check a policy_templates.json file.')
 496     parser.add_option('--fix', action='store_true',
 497                       help='Automatically fix formatting.')
 498     parser.add_option('--backup', action='store_true',
 499                       help='Create backup of original file (before fixing).')
 500     parser.add_option('--stats', action='store_true',
 501                       help='Generate statistics.')
 502     (options, args) = parser.parse_args(argv)
 503     if filename is None:
 504       if len(args) != 2:
 505         parser.print_help()
 506         sys.exit(1)
 507       filename = args[1]
 508     return self.Main(filename, options)
 509
 510
 511 if __name__ == '__main__':
 512   sys.exit(PolicyTemplateChecker().Run(sys.argv))