doxy: support both ClassImp() and ClassDef()
[u/mrichter/AliRoot.git] / doxygen / thtml2doxy.py
CommitLineData
f329fa92 1#!/usr/bin/env python
2
06ccae0f 3## @package thtml2doxy_clang
4# Translates THtml C++ comments to Doxygen using libclang as parser.
5#
6# This code relies on Python bindings for libclang: libclang's interface is pretty unstable, and
7# its Python bindings are unstable as well.
8#
9# AST (Abstract Source Tree) traversal is performed entirely using libclang used as a C++ parser,
10# instead of attempting to write a parser ourselves.
11#
12# This code (expecially AST traversal) was inspired by:
13#
14# - [Implementing a code generator with libclang](http://szelei.me/code-generator/)
15# (this refers to API calls used here)
16# - [Parsing C++ in Python with Clang](http://eli.thegreenplace.net/2011/07/03/parsing-c-in-python-with-clang)
17# (outdated, API calls described there do not work anymore, but useful to understand some basic
18# concepts)
19#
20# Usage:
21#
22# `thtml2doxy_clang [--stdout|-o] [-d] [--debug=DEBUG_LEVEL] file1 [file2 [file3...]]`
23#
24# Parameters:
25#
26# - `--stdout|-o`: output all on standard output instead of writing files in place
27# - `-d`: enable debug mode (very verbose output)
28# - `--debug=DEBUG_LEVEL`: set debug level to one of `DEBUG`, `INFO`, `WARNING`, `ERROR`,
29# `CRITICAL`
30#
31# @author Dario Berzano, CERN
32# @date 2014-12-05
33
34
f329fa92 35import sys
36import os
37import re
06ccae0f 38import logging
39import getopt
ee1793c7 40import hashlib
06ccae0f 41import clang.cindex
42
43
44## Brain-dead color output for terminal.
45class Colt(str):
46
47 def red(self):
48 return self.color('\033[31m')
49
50 def green(self):
51 return self.color('\033[32m')
52
53 def yellow(self):
54 return self.color('\033[33m')
55
56 def blue(self):
57 return self.color('\033[34m')
58
59 def magenta(self):
60 return self.color('\033[35m')
61
62 def cyan(self):
63 return self.color('\033[36m')
64
65 def color(self, c):
66 return c + self + '\033[m'
67
f329fa92 68
06ccae0f 69## Comment.
42d5b6da 70class Comment(object):
06ccae0f 71
c03aba1e 72 def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func, \
73 append_empty=True):
74
3896b0ea 75 assert first_line > 0 and last_line >= first_line, 'Wrong line numbers'
06ccae0f 76 self.lines = lines
77 self.first_line = first_line
78 self.first_col = first_col
79 self.last_line = last_line
80 self.last_col = last_col
81 self.indent = indent
82 self.func = func
c03aba1e 83 self.append_empty = append_empty
06ccae0f 84
85 def has_comment(self, line):
86 return line >= self.first_line and line <= self.last_line
87
88 def __str__(self):
42d5b6da 89 return "<%s for %s: [%d,%d:%d,%d] %s>" % ( \
90 self.__class__.__name__, self.func,
91 self.first_line, self.first_col, self.last_line, self.last_col,
92 self.lines)
93
94
95## Prepend comment.
96class PrependComment(Comment):
97
48a1eb94 98 def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func, \
99 append_empty=False):
42d5b6da 100 super(PrependComment, self).__init__( \
48a1eb94 101 lines, first_line, first_col, last_line, last_col, indent, func, append_empty)
06ccae0f 102
103
104## A data member comment.
105class MemberComment:
106
21689d7a 107 def __init__(self, text, comment_flag, array_size, first_line, first_col, func):
3896b0ea 108 assert first_line > 0, 'Wrong line number'
21689d7a 109 assert comment_flag is None or comment_flag == '!' or comment_flag in [ '!', '||', '->' ]
06ccae0f 110 self.lines = [ text ]
21689d7a 111 self.comment_flag = comment_flag
06ccae0f 112 self.array_size = array_size
113 self.first_line = first_line
114 self.first_col = first_col
115 self.func = func
116
21689d7a 117 def is_transient(self):
118 return self.comment_flag == '!'
119
120 def is_dontsplit(self):
121 return self.comment_flag == '||'
122
123 def is_ptr(self):
124 return self.comment_flag == '->'
125
06ccae0f 126 def has_comment(self, line):
127 return line == self.first_line
128
129 def __str__(self):
130
21689d7a 131 if self.is_transient():
06ccae0f 132 tt = '!transient! '
21689d7a 133 elif self.is_dontsplit():
134 tt = '!dontsplit! '
135 elif self.is_ptr():
136 tt = '!ptr! '
06ccae0f 137 else:
138 tt = ''
139
140 if self.array_size is not None:
141 ars = '[%s] ' % self.array_size
142 else:
143 ars = ''
144
145 return "<MemberComment for %s: [%d,%d] %s%s%s>" % (self.func, self.first_line, self.first_col, tt, ars, self.lines[0])
146
147
148## A dummy comment that removes comment lines.
149class RemoveComment(Comment):
150
151 def __init__(self, first_line, last_line):
3896b0ea 152 assert first_line > 0 and last_line >= first_line, 'Wrong line numbers'
06ccae0f 153 self.first_line = first_line
154 self.last_line = last_line
155 self.func = '<remove>'
156
157 def __str__(self):
158 return "<RemoveComment: [%d,%d]>" % (self.first_line, self.last_line)
159
160
161## Parses method comments.
f329fa92 162#
06ccae0f 163# @param cursor Current libclang parser cursor
164# @param comments Array of comments: new ones will be appended there
165def comment_method(cursor, comments):
166
167 # we are looking for the following structure: method -> compound statement -> comment, i.e. we
168 # need to extract the first comment in the compound statement composing the method
169
170 in_compound_stmt = False
171 expect_comment = False
172 emit_comment = False
173
174 comment = []
175 comment_function = cursor.spelling or cursor.displayname
176 comment_line_start = -1
177 comment_line_end = -1
178 comment_col_start = -1
179 comment_col_end = -1
180 comment_indent = -1
181
182 for token in cursor.get_tokens():
183
184 if token.cursor.kind == clang.cindex.CursorKind.COMPOUND_STMT:
185 if not in_compound_stmt:
186 in_compound_stmt = True
187 expect_comment = True
188 comment_line_end = -1
189 else:
190 if in_compound_stmt:
191 in_compound_stmt = False
192 emit_comment = True
193
194 # tkind = str(token.kind)[str(token.kind).index('.')+1:]
195 # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:]
196
197 if in_compound_stmt:
198
199 if expect_comment:
200
201 extent = token.extent
202 line_start = extent.start.line
203 line_end = extent.end.line
204
205 if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{':
206 pass
207
208 elif token.kind == clang.cindex.TokenKind.COMMENT and (comment_line_end == -1 or (line_start == comment_line_end+1 and line_end-line_start == 0)):
209 comment_line_end = line_end
210 comment_col_end = extent.end.column
211
212 if comment_indent == -1 or (extent.start.column-1) < comment_indent:
213 comment_indent = extent.start.column-1
214
215 if comment_line_start == -1:
216 comment_line_start = line_start
217 comment_col_start = extent.start.column
218 comment.extend( token.spelling.split('\n') )
219
220 # multiline comments are parsed in one go, therefore don't expect subsequent comments
221 if line_end - line_start > 0:
222 emit_comment = True
223 expect_comment = False
224
225 else:
226 emit_comment = True
227 expect_comment = False
228
229 if emit_comment:
230
6f0e3bf3 231 if comment_line_start > 0:
06ccae0f 232
ee1793c7 233 comment = refactor_comment( comment, infilename=str(cursor.location.file) )
6f0e3bf3 234
235 if len(comment) > 0:
236 logging.debug("Comment found for function %s" % Colt(comment_function).magenta())
237 comments.append( Comment(comment, comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) )
238 else:
5dccb084 239 logging.debug('Empty comment found for function %s: collapsing' % Colt(comment_function).magenta())
240 comments.append( Comment([''], comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) )
241 #comments.append(RemoveComment(comment_line_start, comment_line_end))
6f0e3bf3 242
243 else:
244 logging.warning('No comment found for function %s' % Colt(comment_function).magenta())
06ccae0f 245
246 comment = []
247 comment_line_start = -1
248 comment_line_end = -1
249 comment_col_start = -1
250 comment_col_end = -1
251 comment_indent = -1
252
253 emit_comment = False
254 break
255
256
257## Parses comments to class data members.
258#
259# @param cursor Current libclang parser cursor
260# @param comments Array of comments: new ones will be appended there
261def comment_datamember(cursor, comments):
262
263 # Note: libclang 3.5 seems to have problems parsing a certain type of FIELD_DECL, so we revert
264 # to a partial manual parsing. When parsing fails, the cursor's "extent" is not set properly,
265 # returning a line range 0-0. We therefore make the not-so-absurd assumption that the datamember
266 # definition is fully on one line, and we take the line number from cursor.location.
267
268 line_num = cursor.location.line
269 raw = None
270 prev = None
271 found = False
272
21689d7a 273 # Huge overkill: current line saved in "raw", previous in "prev"
06ccae0f 274 with open(str(cursor.location.file)) as fp:
275 cur_line = 0
276 for raw in fp:
277 cur_line = cur_line + 1
278 if cur_line == line_num:
279 found = True
280 break
281 prev = raw
282
283 assert found, 'A line that should exist was not found in file' % cursor.location.file
284
c756848b 285 recomm = r'(//(!|\|\||->)|///?)(\[(.+?)\])?<?\s*(.*?)\s*$'
21689d7a 286 recomm_prevline = r'^\s*///\s*(.*?)\s*$'
06ccae0f 287
288 mcomm = re.search(recomm, raw)
289 if mcomm:
54203c62 290 # If it does not match, we do not have a comment
06ccae0f 291 member_name = cursor.spelling;
21689d7a 292 comment_flag = mcomm.group(2)
06ccae0f 293 array_size = mcomm.group(4)
294 text = mcomm.group(5)
295
296 col_num = mcomm.start()+1;
297
298 if array_size is not None and prev is not None:
299 # ROOT arrays with comments already converted to Doxygen have the member description on the
300 # previous line
21689d7a 301 mcomm_prevline = re.search(recomm_prevline, prev)
302 if mcomm_prevline:
303 text = mcomm_prevline.group(1)
06ccae0f 304 comments.append(RemoveComment(line_num-1, line_num-1))
305
306 logging.debug('Comment found for member %s' % Colt(member_name).magenta())
307
308 comments.append( MemberComment(
309 text,
21689d7a 310 comment_flag,
06ccae0f 311 array_size,
312 line_num,
313 col_num,
314 member_name ))
315
06ccae0f 316
317## Parses class description (beginning of file).
318#
319# The clang parser does not work in this case so we do it manually, but it is very simple: we keep
320# the first consecutive sequence of single-line comments (//) we find - provided that it occurs
321# before any other comment found so far in the file (the comments array is inspected to ensure
322# this).
f329fa92 323#
06ccae0f 324# Multi-line comments (/* ... */) are not considered as they are commonly used to display
325# copyright notice.
f329fa92 326#
06ccae0f 327# @param filename Name of the current file
328# @param comments Array of comments: new ones will be appended there
48a1eb94 329# @param look_no_further_than_line Stop before reaching this line when looking for class comment
330def comment_classdesc(filename, comments, look_no_further_than_line):
06ccae0f 331
332 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
333
ccf0fca4 334 reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)'
06ccae0f 335 class_name_doxy = None
336
ea2f3fcb 337 reauthor = r'(?i)^\s*\\?(authors?|origin):?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$'
06ccae0f 338 redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$'
339 author = None
340 date = None
341
342 comment_lines = []
343
344 start_line = -1
345 end_line = -1
346
347 line_num = 0
348
ccf0fca4 349 is_macro = filename.endswith('.C')
350
06ccae0f 351 with open(filename, 'r') as fp:
352
353 for raw in fp:
354
355 line_num = line_num + 1
356
48a1eb94 357 if look_no_further_than_line is not None and line_num == look_no_further_than_line:
358 logging.debug('Stopping at line %d while looking for class/file description' % \
359 look_no_further_than_line)
360 break
361
424eef90 362 if raw.strip() == '' and start_line > 0:
06ccae0f 363 # Skip empty lines
06ccae0f 364 continue
365
366 stripped = strip_html(raw)
367 mcomm = re.search(recomm, stripped)
368 if mcomm:
369
424eef90 370 if start_line == -1:
06ccae0f 371
372 # First line. Check that we do not overlap with other comments
373 comment_overlaps = False
374 for c in comments:
375 if c.has_comment(line_num):
376 comment_overlaps = True
377 break
378
379 if comment_overlaps:
380 # No need to look for other comments
381 break
382
383 start_line = line_num
384
dde83b2b 385 end_line = line_num
06ccae0f 386 append = True
387
388 mclass_doxy = re.search(reclass_doxy, mcomm.group(1))
389 if mclass_doxy:
ccf0fca4 390 class_name_doxy = mclass_doxy.group(2)
06ccae0f 391 append = False
392 else:
393 mauthor = re.search(reauthor, mcomm.group(1))
394 if mauthor:
ea2f3fcb 395 author = mauthor.group(2)
06ccae0f 396 if date is None:
397 # Date specified in the standalone \date field has priority
ea2f3fcb 398 date = mauthor.group(4)
06ccae0f 399 append = False
400 else:
401 mdate = re.search(redate, mcomm.group(1))
402 if mdate:
403 date = mdate.group(1)
404 append = False
405
406 if append:
407 comment_lines.append( mcomm.group(1) )
408
409 else:
424eef90 410 if start_line > 0:
06ccae0f 411 break
412
413 if class_name_doxy is None:
414
415 # No \class specified: guess it from file name
416 reclass = r'^(.*/)?(.*?)(\..*)?$'
417 mclass = re.search( reclass, filename )
418 if mclass:
419 class_name_doxy = mclass.group(2)
420 else:
421 assert False, 'Regexp unable to extract classname from file'
422
48a1eb94 423 # Macro or class?
424 if is_macro:
425 file_class_line = '\\file ' + class_name_doxy + '.C'
426 else:
427 file_class_line = '\\class ' + class_name_doxy
428
424eef90 429 if start_line > 0:
06ccae0f 430
ccf0fca4 431 # Prepend \class or \file specifier (and an empty line)
48a1eb94 432 comment_lines[:0] = [ file_class_line ]
424eef90 433 comment_lines.append('')
06ccae0f 434
84039997 435 # Append author and date if they exist
424eef90 436 if author is not None:
437 comment_lines.append( '\\author ' + author )
06ccae0f 438
424eef90 439 if date is not None:
440 comment_lines.append( '\\date ' + date )
441
ee1793c7 442 comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
424eef90 443 logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
444 comments.append(Comment(
445 comment_lines,
446 start_line, 1, end_line, 1,
447 0, class_name_doxy
448 ))
449
450 else:
451
48a1eb94 452 logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
453 Colt(class_name_doxy).magenta())
454
455 comments.append(PrependComment(
456 [ file_class_line ],
457 1, 1, 1, 1,
458 0, class_name_doxy, append_empty=True
459 ))
06ccae0f 460
461
b9781053 462## Looks for a special ROOT ClassImp() entry.
463#
464# Doxygen might get confused by `ClassImp()` entries as they are macros normally written without
465# the ending `;`: this function wraps the definition inside a condition in order to make Doxygen
466# ignore it.
467#
468# @param filename Name of the current file
469# @param comments Array of comments: new ones will be appended there
470def comment_classimp(filename, comments):
471
472 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
473
474 line_num = 0
f0afbd73 475 reclassimp = r'^(\s*)Class(Imp|Def)\((.*?)\).*$'
b9781053 476
c03aba1e 477 in_classimp_cond = False
478 restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
479 reendcond = r'^\s*///\s*\\endcond\s*$'
480
b9781053 481 with open(filename, 'r') as fp:
482
c03aba1e 483 line_classimp = -1
484 line_startcond = -1
485 line_endcond = -1
486 classimp_class = None
487 classimp_indent = None
488
b9781053 489 for line in fp:
490
491 line_num = line_num + 1
c03aba1e 492
b9781053 493 mclassimp = re.search(reclassimp, line)
494 if mclassimp:
495
496 # Adjust indent
c03aba1e 497 classimp_indent = len( mclassimp.group(1) )
b9781053 498
c03aba1e 499 line_classimp = line_num
f0afbd73 500 classimp_class = mclassimp.group(3)
501 imp_or_def = mclassimp.group(2)
b9781053 502 logging.debug(
503 'Comment found for ' +
f0afbd73 504 Colt( 'Class%s(' % imp_or_def ).magenta() +
c03aba1e 505 Colt( classimp_class ).cyan() +
b9781053 506 Colt( ')' ).magenta() )
507
c03aba1e 508 else:
b9781053 509
c03aba1e 510 mstartcond = re.search(restartcond, line)
511 if mstartcond:
512 logging.debug('Found Doxygen opening condition for ClassImp in {%s}' % line)
513 in_classimp_cond = True
514 line_startcond = line_num
515
516 elif in_classimp_cond:
517
518 mendcond = re.search(reendcond, line)
519 if mendcond:
520 logging.debug('Found Doxygen closing condition for ClassImp')
521 in_classimp_cond = False
522 line_endcond = line_num
523
524 # Did we find something?
525 if line_classimp != -1:
526
527 if line_startcond != -1:
528 comments.append(Comment(
529 ['\cond CLASSIMP'],
530 line_startcond, 1, line_startcond, 1,
f0afbd73 531 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
c03aba1e 532 append_empty=False
533 ))
534 else:
535 comments.append(PrependComment(
536 ['\cond CLASSIMP'],
537 line_classimp, 1, line_classimp, 1,
f0afbd73 538 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
c03aba1e 539 ))
540
541 if line_endcond != -1:
542 comments.append(Comment(
543 ['\endcond'],
544 line_endcond, 1, line_endcond, 1,
f0afbd73 545 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
c03aba1e 546 append_empty=False
547 ))
548 else:
549 comments.append(PrependComment(
550 ['\endcond'],
551 line_classimp+1, 1, line_classimp+1, 1,
f0afbd73 552 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
c03aba1e 553 ))
b9781053 554
555
06ccae0f 556## Traverse the AST recursively starting from the current cursor.
557#
558# @param cursor A Clang parser cursor
559# @param filename Name of the current file
560# @param comments Array of comments: new ones will be appended there
561# @param recursion Current recursion depth
84039997 562# @param in_func True if we are inside a function or method
48a1eb94 563# @param classdesc_line_limit Do not look for comments after this line
564#
565# @return A tuple containing the classdesc_line_limit as first item
566def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
06ccae0f 567
568 # libclang traverses included files as well: we do not want this behavior
569 if cursor.location.file is not None and str(cursor.location.file) != filename:
570 logging.debug("Skipping processing of included %s" % cursor.location.file)
571 return
572
573 text = cursor.spelling or cursor.displayname
574 kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
575
ccf0fca4 576 is_macro = filename.endswith('.C')
577
06ccae0f 578 indent = ''
579 for i in range(0, recursion):
580 indent = indent + ' '
581
2d3fd2ec 582 if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
e215c6c6 583 clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
06ccae0f 584
48a1eb94 585 if classdesc_line_limit is None:
586 classdesc_line_limit = cursor.location.line
587
06ccae0f 588 # cursor ran into a C++ method
589 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
590 comment_method(cursor, comments)
84039997 591 in_func = True
06ccae0f 592
84039997 593 elif not is_macro and not in_func and \
594 cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
06ccae0f 595
48a1eb94 596 if classdesc_line_limit is None:
597 classdesc_line_limit = cursor.location.line
598
06ccae0f 599 # cursor ran into a data member declaration
600 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
601 comment_datamember(cursor, comments)
602
603 else:
604
605 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
606
607 for child_cursor in cursor.get_children():
48a1eb94 608 classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
06ccae0f 609
610 if recursion == 0:
b9781053 611 comment_classimp(filename, comments)
48a1eb94 612 comment_classdesc(filename, comments, classdesc_line_limit)
613
614 return classdesc_line_limit
06ccae0f 615
616
617## Strip some HTML tags from the given string. Returns clean string.
618#
619# @param s Input string
620def strip_html(s):
798167cb 621 rehtml = r'(?i)</?(P|BR)/?>'
06ccae0f 622 return re.sub(rehtml, '', s)
623
624
625## Remove garbage from comments and convert special tags from THtml to Doxygen.
626#
627# @param comment An array containing the lines of the original comment
ee1793c7 628def refactor_comment(comment, do_strip_html=True, infilename=None):
06ccae0f 629
630 recomm = r'^(/{2,}|/\*)? ?(\s*.*?)\s*((/{2,})?\s*|\*/)$'
631 regarbage = r'^(?i)\s*([\s*=-_#]+|(Begin|End)_Html)\s*$'
632
03a6b7aa 633 # Support for LaTeX blocks spanning on multiple lines
634 relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
635 in_latex = False
636 latex_block = False
637
638 # Support for LaTeX blocks on a single line
9db428b7 639 reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
640
35b193b4 641 # Match <pre> (to turn it into the ~~~ Markdown syntax)
642 reblock = r'(?i)^(\s*)</?PRE>\s*$'
643
ee1793c7 644 # Macro blocks for pictures generation
645 in_macro = False
646 current_macro = []
647 remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
648
06ccae0f 649 new_comment = []
650 insert_blank = False
651 wait_first_non_blank = True
652 for line_comment in comment:
653
ee1793c7 654 # Check if we are in a macro block
655 mmacro = re.search(remacro, line_comment)
656 if mmacro:
657 if in_macro:
658 in_macro = False
659
660 # Dump macro
661 outimg = write_macro(infilename, current_macro) + '.png'
662 current_macro = []
663
664 # Insert image
665 new_comment.append( '![Picture from ROOT macro](%s)' % (outimg) )
666
667 logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
668
669 else:
670 in_macro = True
671
672 continue
673 elif in_macro:
674 current_macro.append( line_comment )
675 continue
676
06ccae0f 677 # Strip some HTML tags
678 if do_strip_html:
679 line_comment = strip_html(line_comment)
680
681 mcomm = re.search( recomm, line_comment )
682 if mcomm:
683 new_line_comment = mcomm.group(2)
684 mgarbage = re.search( regarbage, new_line_comment )
685
686 if new_line_comment == '' or mgarbage is not None:
687 insert_blank = True
688 else:
689 if insert_blank and not wait_first_non_blank:
690 new_comment.append('')
691 insert_blank = False
692 wait_first_non_blank = False
9db428b7 693
694 # Postprocessing: LaTeX formulas in ROOT format
695 # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
696 # There can be several ROOT LaTeX forumlas per line
697 while True:
698 minline_latex = re.search( reinline_latex, new_line_comment )
699 if minline_latex:
700 new_line_comment = '%s\\f$%s\\f$%s' % \
701 ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
702 minline_latex.group(3) )
703 else:
704 break
705
03a6b7aa 706 # ROOT LaTeX: do we have a Begin/End_LaTeX block?
707 # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
708 # block here left without a corresponding ending block
709 mlatex = re.search( relatex, new_line_comment )
710 if mlatex:
711
712 # before and after parts have been already stripped
713 l_before = mlatex.group(2)
714 l_after = mlatex.group(4)
715 is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
716
717 if l_before is None:
718 l_before = ''
719 if l_after is None:
720 l_after = ''
721
722 if is_begin:
723
724 # Begin of LaTeX part
725
726 in_latex = True
727 if l_before == '' and l_after == '':
728
729 # Opening tag alone: mark the beginning of a block: \f[ ... \f]
730 latex_block = True
731 new_comment.append( '\\f[' )
732
733 else:
734 # Mark the beginning of inline: \f$ ... \f$
735 latex_block = False
736 new_comment.append(
737 '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
738 )
739
740 else:
741
742 # End of LaTeX part
743 in_latex = False
744
745 if latex_block:
746
747 # Closing a LaTeX block
748 if l_before != '':
749 new_comment.append( l_before.replace('#', '\\') )
750 new_comment.append( '\\f]' )
751 if l_after != '':
752 new_comment.append( l_after )
753
754 else:
755
756 # Closing a LaTeX inline
757 new_comment.append(
758 '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
759 )
760
761 # Prevent appending lines (we have already done that)
762 new_line_comment = None
763
35b193b4 764 # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
765 # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
766 if new_line_comment is not None and not in_latex:
767
768 mblock = re.search( reblock, new_line_comment )
769 if mblock:
770 new_comment.append( mblock.group(1)+'~~~' )
771 new_line_comment = None
772
03a6b7aa 773 if new_line_comment is not None:
774 if in_latex:
775 new_line_comment = new_line_comment.replace('#', '\\')
776 new_comment.append( new_line_comment )
06ccae0f 777
778 else:
779 assert False, 'Comment regexp does not match'
780
781 return new_comment
782
783
ee1793c7 784## Dumps an image-generating macro to the correct place. Returns a string with the image path,
785# without the extension.
786#
787# @param infilename File name of the source file
788# @param macro_lines Array of macro lines
789def write_macro(infilename, macro_lines):
790
791 # Calculate hash
792 digh = hashlib.sha1()
793 for l in macro_lines:
794 digh.update(l)
795 digh.update('\n')
796 short_digest = digh.hexdigest()[0:7]
797
798 outdir = '%s/imgdoc' % os.path.dirname(infilename)
799 outprefix = '%s/%s_%s' % (
800 outdir,
801 os.path.basename(infilename).replace('.', '_'),
802 short_digest
803 )
804 outmacro = '%s.C' % outprefix
805
806 # Make directory
807 if not os.path.isdir(outdir):
808 # do not catch: let everything die on error
809 logging.debug('Creating directory %s' % Colt(outdir).magenta())
810 os.mkdir(outdir)
811
812 # Create file (do not catch errors either)
813 with open(outmacro, 'w') as omfp:
814 logging.debug('Writing macro %s' % Colt(outmacro).magenta())
815 for l in macro_lines:
816 omfp.write(l)
817 omfp.write('\n')
818
819 return outprefix
820
821
06ccae0f 822## Rewrites all comments from the given file handler.
823#
824# @param fhin The file handler to read from
825# @param fhout The file handler to write to
826# @param comments Array of comments
827def rewrite_comments(fhin, fhout, comments):
828
829 line_num = 0
830 in_comment = False
831 skip_empty = False
832 comm = None
833 prev_comm = None
b9781053 834 restore_lines = None
06ccae0f 835
836 rindent = r'^(\s*)'
837
b9781053 838 def dump_comment_block(cmt, restore=None):
fc82f9c7 839 text_indent = ''
b9781053 840 ask_skip_empty = False
841
fc82f9c7 842 for i in range(0, cmt.indent):
843 text_indent = text_indent + ' '
844
845 for lc in cmt.lines:
846 fhout.write('%s///' % text_indent )
847 lc = lc.rstrip()
848 if len(lc) != 0:
849 fhout.write(' ')
850 fhout.write(lc)
851 fhout.write('\n')
852
853 # Empty new line at the end of the comment
c03aba1e 854 if cmt.append_empty:
b9781053 855 fhout.write('\n')
856 ask_skip_empty = True
857
858 # Restore lines if possible
859 if restore:
860 for lr in restore:
861 fhout.write(lr)
862 fhout.write('\n')
863
864 # Tell the caller whether it should skip the next empty line found
865 return ask_skip_empty
c0094f3c 866
867
06ccae0f 868 for line in fhin:
869
870 line_num = line_num + 1
871
872 # Find current comment
873 prev_comm = comm
874 comm = None
875 for c in comments:
876 if c.has_comment(line_num):
877 comm = c
878
879 if comm:
880
b9781053 881 # First thing to check: are we in the same comment as before?
c03aba1e 882 if comm is not prev_comm and isinstance(comm, Comment) and isinstance(prev_comm, Comment) \
883 and not isinstance(prev_comm, RemoveComment):
b9781053 884
885 skip_empty = dump_comment_block(prev_comm, restore_lines)
886 in_comment = False
887 restore_lines = None
888 prev_comm = None # we have just dumped it: pretend it never existed in this loop
889
890 #
891 # Check type of comment and react accordingly
892 #
893
06ccae0f 894 if isinstance(comm, MemberComment):
c0094f3c 895
896 # end comment block
897 if in_comment:
b9781053 898 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 899 in_comment = False
b9781053 900 restore_lines = None
c0094f3c 901
06ccae0f 902 non_comment = line[ 0:comm.first_col-1 ]
903
21689d7a 904 if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
06ccae0f 905
21689d7a 906 # This is a special case: comment will be split in two lines: one before the comment for
907 # Doxygen as "member description", and the other right after the comment on the same line
908 # to be parsed by ROOT's C++ parser
909
910 # Keep indent on the generated line of comment before member definition
06ccae0f 911 mindent = re.search(rindent, line)
21689d7a 912
913 # Get correct comment flag, if any
914 if comm.comment_flag is not None:
915 cflag = comm.comment_flag
916 else:
917 cflag = ''
918
919 # Get correct array size, if any
920 if comm.array_size is not None:
921 asize = '[%s]' % comm.array_size
06ccae0f 922 else:
21689d7a 923 asize = ''
06ccae0f 924
21689d7a 925 # Write on two lines
926 fhout.write('%s/// %s\n%s//%s%s\n' % (
06ccae0f 927 mindent.group(1),
928 comm.lines[0],
929 non_comment,
21689d7a 930 cflag,
931 asize
06ccae0f 932 ))
933
934 else:
935
21689d7a 936 # Single-line comments with the "transient" flag can be kept on one line in a way that
937 # they are correctly interpreted by both ROOT and Doxygen
938
939 if comm.is_transient():
06ccae0f 940 tt = '!'
941 else:
942 tt = '/'
943
944 fhout.write('%s//%s< %s\n' % (
945 non_comment,
946 tt,
947 comm.lines[0]
948 ))
949
950 elif isinstance(comm, RemoveComment):
c0094f3c 951 # End comment block and skip this line
952 if in_comment:
b9781053 953 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 954 in_comment = False
b9781053 955 restore_lines = None
06ccae0f 956
957 elif prev_comm is None:
c0094f3c 958
b9781053 959 # Beginning of a new comment block of type Comment or PrependComment
06ccae0f 960 in_comment = True
961
b9781053 962 if isinstance(comm, PrependComment):
963 # Prepare array of lines to dump right after the comment
964 restore_lines = [ line.rstrip('\n') ]
965 else:
966 # Extract the non-comment part and print it if it exists
967 non_comment = line[ 0:comm.first_col-1 ].rstrip()
968 if non_comment != '':
969 fhout.write( non_comment + '\n' )
970
971 elif isinstance(comm, Comment):
972
973 if restore_lines is not None:
974 # From the 2nd line on of comment to prepend
975 restore_lines.append( line.rstrip('\n') )
976
977 else:
978 assert False, 'Unhandled parser state. line=%d comm=%s prev_comm=%s' % \
979 (line_num, comm, prev_comm)
06ccae0f 980
981 else:
982
b9781053 983 # Not a comment line
984
06ccae0f 985 if in_comment:
986
987 # We have just exited a comment block of type Comment
b9781053 988 skip_empty = dump_comment_block(prev_comm, restore_lines)
06ccae0f 989 in_comment = False
b9781053 990 restore_lines = None
06ccae0f 991
b9781053 992 # Dump the non-comment line
06ccae0f 993 line_out = line.rstrip('\n')
994 if skip_empty:
995 skip_empty = False
996 if line_out.strip() != '':
997 fhout.write( line_out + '\n' )
998 else:
999 fhout.write( line_out + '\n' )
1000
f329fa92 1001
1002## The main function.
1003#
06ccae0f 1004# Return value is the executable's return value.
f329fa92 1005def main(argv):
1006
06ccae0f 1007 # Setup logging on stderr
1008 log_level = logging.INFO
1009 logging.basicConfig(
1010 level=log_level,
1011 format='%(levelname)-8s %(funcName)-20s %(message)s',
1012 stream=sys.stderr
1013 )
f329fa92 1014
06ccae0f 1015 # Parse command-line options
1016 output_on_stdout = False
7afec95e 1017 include_flags = []
06ccae0f 1018 try:
7afec95e 1019 opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
06ccae0f 1020 for o, a in opts:
1021 if o == '--debug':
1022 log_level = getattr( logging, a.upper(), None )
1023 if not isinstance(log_level, int):
1024 raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
1025 elif o == '-d':
1026 log_level = logging.DEBUG
1027 elif o == '-o' or o == '--stdout':
06ccae0f 1028 output_on_stdout = True
7afec95e 1029 elif o == '-I':
1030 if os.path.isdir(a):
1031 include_flags.extend( [ '-I', a ] )
1032 else:
1033 logging.fatal('Include directory not found: %s' % Colt(a).magenta())
1034 return 2
06ccae0f 1035 else:
1036 assert False, 'Unhandled argument'
1037 except getopt.GetoptError as e:
1038 logging.fatal('Invalid arguments: %s' % e)
1039 return 1
f329fa92 1040
06ccae0f 1041 logging.getLogger('').setLevel(log_level)
f329fa92 1042
06ccae0f 1043 # Attempt to load libclang from a list of known locations
1044 libclang_locations = [
1045 '/usr/lib/llvm-3.5/lib/libclang.so.1',
1046 '/usr/lib/libclang.so',
1047 '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
1048 ]
1049 libclang_found = False
f329fa92 1050
06ccae0f 1051 for lib in libclang_locations:
1052 if os.path.isfile(lib):
1053 clang.cindex.Config.set_library_file(lib)
1054 libclang_found = True
1055 break
f329fa92 1056
06ccae0f 1057 if not libclang_found:
1058 logging.fatal('Cannot find libclang')
1059 return 1
1060
1061 # Loop over all files
1062 for fn in args:
1063
1064 logging.info('Input file: %s' % Colt(fn).magenta())
1065 index = clang.cindex.Index.create()
7afec95e 1066 clang_args = [ '-x', 'c++' ]
1067 clang_args.extend( include_flags )
1068 translation_unit = index.parse(fn, args=clang_args)
06ccae0f 1069
1070 comments = []
1071 traverse_ast( translation_unit.cursor, fn, comments )
1072 for c in comments:
1073
1074 logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
f329fa92 1075
06ccae0f 1076 if isinstance(c, MemberComment):
1077
21689d7a 1078 if c.is_transient():
1079 flag_text = Colt('transient ').yellow()
1080 elif c.is_dontsplit():
1081 flag_text = Colt('dontsplit ').yellow()
1082 elif c.is_ptr():
1083 flag_text = Colt('ptr ').yellow()
06ccae0f 1084 else:
21689d7a 1085 flag_text = ''
06ccae0f 1086
1087 if c.array_size is not None:
1088 array_text = Colt('arraysize=%s ' % c.array_size).yellow()
1089 else:
1090 array_text = ''
1091
1092 logging.debug(
1093 "%s %s%s{%s}" % ( \
1094 Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
21689d7a 1095 flag_text,
06ccae0f 1096 array_text,
1097 Colt(c.lines[0]).cyan()
1098 ))
1099
1100 elif isinstance(c, RemoveComment):
1101
1102 logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
1103
1104 else:
1105 for l in c.lines:
1106 logging.debug(
1107 Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
1108 "{%s}" % Colt(l).cyan()
1109 )
f329fa92 1110
1111 try:
06ccae0f 1112
1113 if output_on_stdout:
1114 with open(fn, 'r') as fhin:
1115 rewrite_comments( fhin, sys.stdout, comments )
1116 else:
1117 fn_back = fn + '.thtml2doxy_backup'
1118 os.rename( fn, fn_back )
1119
1120 with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
1121 rewrite_comments( fhin, fhout, comments )
1122
1123 os.remove( fn_back )
1124 logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
1125 except (IOError,OSError) as e:
1126 logging.error('File operation failed: %s' % e)
f329fa92 1127
1128 return 0
1129
06ccae0f 1130
f329fa92 1131if __name__ == '__main__':
06ccae0f 1132 sys.exit( main( sys.argv[1:] ) )