1#!/usr/bin/env python3
2# A tool to parse the FormatStyle struct from Format.h and update the
3# documentation in ../ClangFormatStyleOptions.rst automatically.
4# Run from the directory in which this file is located to update the docs.
5
6import inspect
7import os
8import re
9import sys
10from io import TextIOWrapper
11from typing import Set
12
13CLANG_DIR = os.path.join(os.path.dirname(__file__), '../..')
14FORMAT_STYLE_FILE = os.path.join(CLANG_DIR, 'include/clang/Format/Format.h')
15INCLUDE_STYLE_FILE = os.path.join(CLANG_DIR, 'include/clang/Tooling/Inclusions/IncludeStyle.h')
16DOC_FILE = os.path.join(CLANG_DIR, 'docs/ClangFormatStyleOptions.rst')
17
18PLURALS_FILE = os.path.join(os.path.dirname(__file__), 'plurals.txt')
19
20plurals: Set[str] = set()
21with open(PLURALS_FILE, 'a+') as f:
22  f.seek(0)
23  plurals = set(f.read().splitlines())
24
25def substitute(text, tag, contents):
26  replacement = '\n.. START_%s\n\n%s\n\n.. END_%s\n' % (tag, contents, tag)
27  pattern = r'\n\.\. START_%s\n.*\n\.\. END_%s\n' % (tag, tag)
28  return re.sub(pattern, '%s', text, flags=re.S) % replacement
29
30def register_plural(singular: str, plural: str):
31  if plural not in plurals:
32    if not hasattr(register_plural, "generated_new_plural"):
33      print('Plural generation: you can use '
34      f'`git checkout -- {os.path.relpath(PLURALS_FILE)}` '
35      'to reemit warnings or `git add` to include new plurals\n')
36    register_plural.generated_new_plural = True
37
38    plurals.add(plural)
39    with open(PLURALS_FILE, 'a') as f:
40      f.write(plural + '\n')
41    cf = inspect.currentframe()
42    lineno = ''
43    if cf and cf.f_back:
44      lineno = ':' + str(cf.f_back.f_lineno)
45    print(f'{__file__}{lineno} check if plural of {singular} is {plural}', file=sys.stderr)
46  return plural
47
48def pluralize(word: str):
49  lword = word.lower()
50  if len(lword) >= 2 and lword[-1] == 'y' and lword[-2] not in 'aeiou':
51    return register_plural(word, word[:-1] + 'ies')
52  elif lword.endswith(('s', 'sh', 'ch', 'x', 'z')):
53    return register_plural(word, word[:-1] + 'es')
54  elif lword.endswith('fe'):
55    return register_plural(word, word[:-2] + 'ves')
56  elif lword.endswith('f') and not lword.endswith('ff'):
57    return register_plural(word, word[:-1] + 'ves')
58  else:
59    return register_plural(word, word + 's')
60
61
62def to_yaml_type(typestr: str):
63  if typestr == 'bool':
64    return 'Boolean'
65  elif typestr == 'int':
66    return 'Integer'
67  elif typestr == 'unsigned':
68    return 'Unsigned'
69  elif typestr == 'std::string':
70    return 'String'
71
72  subtype, napplied = re.subn(r'^std::vector<(.*)>$', r'\1', typestr)
73  if napplied == 1:
74    return 'List of ' + pluralize(to_yaml_type(subtype))
75
76  return typestr
77
78def doxygen2rst(text):
79  text = re.sub(r'<tt>\s*(.*?)\s*<\/tt>', r'``\1``', text)
80  text = re.sub(r'\\c ([^ ,;\.]+)', r'``\1``', text)
81  text = re.sub(r'\\\w+ ', '', text)
82  return text
83
84def indent(text, columns, indent_first_line=True):
85  indent_str = ' ' * columns
86  s = re.sub(r'\n([^\n])', '\n' + indent_str + '\\1', text, flags=re.S)
87  if not indent_first_line or s.startswith('\n'):
88    return s
89  return indent_str + s
90
91class Option(object):
92  def __init__(self, name, opt_type, comment, version):
93    self.name = name
94    self.type = opt_type
95    self.comment = comment.strip()
96    self.enum = None
97    self.nested_struct = None
98    self.version = version
99
100  def __str__(self):
101    s = ".. _%s:\n\n**%s** (``%s``) " % (self.name, self.name, to_yaml_type(self.type))
102    if self.version:
103      s += ':versionbadge:`clang-format %s` ' % self.version
104    s += ':ref:`¶ <%s>`\n%s' % (self.name, doxygen2rst(indent(self.comment, 2)))
105    if self.enum and self.enum.values:
106      s += indent('\n\nPossible values:\n\n%s\n' % self.enum, 2)
107    if self.nested_struct:
108      s += indent('\n\nNested configuration flags:\n\n%s\n' %self.nested_struct,
109                  2)
110    return s
111
112class NestedStruct(object):
113  def __init__(self, name, comment):
114    self.name = name
115    self.comment = comment.strip()
116    self.values = []
117
118  def __str__(self):
119    return self.comment + '\n' + '\n'.join(map(str, self.values))
120
121class NestedField(object):
122  def __init__(self, name, comment):
123    self.name = name
124    self.comment = comment.strip()
125
126  def __str__(self):
127    return '\n* ``%s`` %s' % (
128        self.name,
129        doxygen2rst(indent(self.comment, 2, indent_first_line=False)))
130
131class Enum(object):
132  def __init__(self, name, comment):
133    self.name = name
134    self.comment = comment.strip()
135    self.values = []
136
137  def __str__(self):
138    return '\n'.join(map(str, self.values))
139
140class NestedEnum(object):
141  def __init__(self, name, enumtype, comment, values):
142    self.name = name
143    self.comment = comment
144    self.values = values
145    self.type = enumtype
146
147  def __str__(self):
148    s = '\n* ``%s %s``\n%s' % (to_yaml_type(self.type), self.name,
149                                 doxygen2rst(indent(self.comment, 2)))
150    s += indent('\nPossible values:\n\n', 2)
151    s += indent('\n'.join(map(str, self.values)), 2)
152    return s
153
154class EnumValue(object):
155  def __init__(self, name, comment, config):
156    self.name = name
157    self.comment = comment
158    self.config = config
159
160  def __str__(self):
161    return '* ``%s`` (in configuration: ``%s``)\n%s' % (
162        self.name,
163        re.sub('.*_', '', self.config),
164        doxygen2rst(indent(self.comment, 2)))
165
166
167class OptionsReader:
168  def __init__(self, header: TextIOWrapper):
169    self.header = header
170    self.in_code_block = False
171    self.code_indent = 0
172    self.lineno = 0
173    self.last_err_lineno = -1
174
175  def __file_path(self):
176    return os.path.relpath(self.header.name)
177
178  def __print_line(self, line: str):
179    print(f'{self.lineno:>6} | {line}', file=sys.stderr)
180
181  def __warning(self, msg: str, line: str):
182    print(f'{self.__file_path()}:{self.lineno}: warning: {msg}:', file=sys.stderr)
183    self.__print_line(line)
184
185  def __clean_comment_line(self, line: str):
186    match = re.match(r'^/// (?P<indent> +)?\\code(\{.(?P<lang>\w+)\})?$', line)
187    if match:
188      if self.in_code_block:
189        self.__warning('`\\code` in another `\\code`', line)
190      self.in_code_block = True
191      indent_str = match.group('indent')
192      if not indent_str:
193        indent_str = ''
194      self.code_indent = len(indent_str)
195      lang = match.group('lang')
196      if not lang:
197        lang = 'c++'
198      return f'\n{indent_str}.. code-block:: {lang}\n\n'
199
200    endcode_match = re.match(r'^/// +\\endcode$', line)
201    if endcode_match:
202      if not self.in_code_block:
203        self.__warning('no correct `\\code` found before this `\\endcode`', line)
204      self.in_code_block = False
205      return ''
206
207    # check code block indentation
208    if (self.in_code_block and not line == '///' and not
209        line.startswith('///  ' + ' ' * self.code_indent)):
210      if self.last_err_lineno == self.lineno - 1:
211        self.__print_line(line)
212      else:
213        self.__warning('code block should be indented', line)
214      self.last_err_lineno = self.lineno
215
216    match = re.match(r'^/// \\warning$', line)
217    if match:
218      return '\n.. warning:: \n\n'
219
220    endwarning_match = re.match(r'^/// +\\endwarning$', line)
221    if endwarning_match:
222      return ''
223    return line[4:] + '\n'
224
225  def read_options(self):
226    class State:
227      BeforeStruct, Finished, InStruct, InNestedStruct, InNestedFieldComment, \
228        InFieldComment, InEnum, InEnumMemberComment = range(8)
229    state = State.BeforeStruct
230
231    options = []
232    enums = {}
233    nested_structs = {}
234    comment = ''
235    enum = None
236    nested_struct = None
237    version = None
238
239    for line in self.header:
240      self.lineno += 1
241      line = line.strip()
242      if state == State.BeforeStruct:
243        if line in ('struct FormatStyle {', 'struct IncludeStyle {'):
244          state = State.InStruct
245      elif state == State.InStruct:
246        if line.startswith('///'):
247          state = State.InFieldComment
248          comment = self.__clean_comment_line(line)
249        elif line == '};':
250          state = State.Finished
251          break
252      elif state == State.InFieldComment:
253        if line.startswith(r'/// \version'):
254          match = re.match(r'/// \\version\s*(?P<version>[0-9.]+)*', line)
255          if match:
256            version = match.group('version')
257        elif line.startswith('///'):
258          comment += self.__clean_comment_line(line)
259        elif line.startswith('enum'):
260          state = State.InEnum
261          name = re.sub(r'enum\s+(\w+)\s*(:((\s*\w+)+)\s*)?\{', '\\1', line)
262          enum = Enum(name, comment)
263        elif line.startswith('struct'):
264          state = State.InNestedStruct
265          name = re.sub(r'struct\s+(\w+)\s*\{', '\\1', line)
266          nested_struct = NestedStruct(name, comment)
267        elif line.endswith(';'):
268          prefix = '// '
269          if line.startswith(prefix):
270            line = line[len(prefix):]
271          state = State.InStruct
272          field_type, field_name = re.match(r'([<>:\w(,\s)]+)\s+(\w+);',
273                                            line).groups()
274
275          if not version:
276            self.__warning(f'missing version for {field_name}', line)
277          option = Option(str(field_name), str(field_type), comment, version)
278          options.append(option)
279          version = None
280        else:
281          raise Exception('Invalid format, expected comment, field or enum\n' + line)
282      elif state == State.InNestedStruct:
283        if line.startswith('///'):
284          state = State.InNestedFieldComment
285          comment = self.__clean_comment_line(line)
286        elif line == '};':
287          state = State.InStruct
288          nested_structs[nested_struct.name] = nested_struct
289      elif state == State.InNestedFieldComment:
290        if line.startswith('///'):
291          comment += self.__clean_comment_line(line)
292        else:
293          state = State.InNestedStruct
294          field_type, field_name = re.match(r'([<>:\w(,\s)]+)\s+(\w+);', line).groups()
295          if field_type in enums:
296            nested_struct.values.append(NestedEnum(field_name,
297                                                   field_type,
298                                                   comment,
299                                                   enums[field_type].values))
300          else:
301            nested_struct.values.append(NestedField(field_type + " " + field_name, comment))
302
303      elif state == State.InEnum:
304        if line.startswith('///'):
305          state = State.InEnumMemberComment
306          comment = self.__clean_comment_line(line)
307        elif line == '};':
308          state = State.InStruct
309          enums[enum.name] = enum
310        else:
311          # Enum member without documentation. Must be documented where the enum
312          # is used.
313          pass
314      elif state == State.InEnumMemberComment:
315        if line.startswith('///'):
316          comment += self.__clean_comment_line(line)
317        else:
318          state = State.InEnum
319          val = line.replace(',', '')
320          pos = val.find(" // ")
321          if pos != -1:
322            config = val[pos + 4:]
323            val = val[:pos]
324          else:
325            config = val
326          enum.values.append(EnumValue(val, comment, config))
327    if state != State.Finished:
328      raise Exception('Not finished by the end of file')
329
330    for option in options:
331      if option.type not in ['bool', 'unsigned', 'int', 'std::string',
332                             'std::vector<std::string>',
333                             'std::vector<IncludeCategory>',
334                             'std::vector<RawStringFormat>']:
335        if option.type in enums:
336          option.enum = enums[option.type]
337        elif option.type in nested_structs:
338          option.nested_struct = nested_structs[option.type]
339        else:
340          raise Exception('Unknown type: %s' % option.type)
341    return options
342
343
344with open(FORMAT_STYLE_FILE) as f:
345  opts = OptionsReader(f).read_options()
346with open(INCLUDE_STYLE_FILE) as f:
347  opts += OptionsReader(f).read_options()
348
349opts = sorted(opts, key=lambda x: x.name)
350options_text = '\n\n'.join(map(str, opts))
351
352with open(DOC_FILE) as f:
353  contents = f.read()
354
355contents = substitute(contents, 'FORMAT_STYLE_OPTIONS', options_text)
356
357with open(DOC_FILE, 'wb') as output:
358  output.write(contents.encode())
359