Update master to aliroot
[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17# (outdated, API calls described there do not work anymore, but useful to understand some basic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#
b4dd7d25 324# Multi-line comments (/* ... */) that *immediately* follow a series of single-line comments
325# (*i.e.* without empty lines in-between) are also considered. A class description can eventually
326# be a series of single-line and multi-line comments, with no blank spaces between them, and always
327# starting with a single-line sequence.
328#
329# The reason why they cannot start with a multi-line sequence is that those are commonly used to
330# display a copyright notice.
f329fa92 331#
06ccae0f 332# @param filename Name of the current file
333# @param comments Array of comments: new ones will be appended there
48a1eb94 334# @param look_no_further_than_line Stop before reaching this line when looking for class comment
335def comment_classdesc(filename, comments, look_no_further_than_line):
06ccae0f 336
b4dd7d25 337 # Single-line comment
a3e90541 338 recomm = r'^\s*///?(\s*(.*?))\s*/*\s*$'
06ccae0f 339
b4dd7d25 340 # Multi-line comment (only either /* or */ on a single line)
341 remlcomm_in = r'^\s*/\*\s*$'
342 remlcomm_out = r'^\s*\*/\s*$'
343 in_mlcomm = False
344
ccf0fca4 345 reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)'
06ccae0f 346 class_name_doxy = None
347
ea2f3fcb 348 reauthor = r'(?i)^\s*\\?(authors?|origin):?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$'
06ccae0f 349 redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$'
fc7afee9 350 rebrief = r'(?i)^\s*\\brief\s*(.*)\s*$'
06ccae0f 351 author = None
352 date = None
fc7afee9 353 brief = None
354 brief_len_threshold = 80
06ccae0f 355
356 comment_lines = []
357
358 start_line = -1
359 end_line = -1
360
361 line_num = 0
b4dd7d25 362 last_comm_line_num = 0
06ccae0f 363
ccf0fca4 364 is_macro = filename.endswith('.C')
365
06ccae0f 366 with open(filename, 'r') as fp:
367
368 for raw in fp:
369
370 line_num = line_num + 1
371
48a1eb94 372 if look_no_further_than_line is not None and line_num == look_no_further_than_line:
373 logging.debug('Stopping at line %d while looking for class/file description' % \
374 look_no_further_than_line)
375 break
376
b4dd7d25 377 if in_mlcomm == False and raw.strip() == '' and start_line > 0:
06ccae0f 378 # Skip empty lines
06ccae0f 379 continue
380
381 stripped = strip_html(raw)
a3e90541 382 mcomm = None
383 this_comment = None
384
385 if not in_mlcomm:
386 mcomm = re.search(recomm, stripped)
387
388 if last_comm_line_num == 0 or last_comm_line_num == line_num-1:
389
390 if mcomm and not mcomm.group(2).startswith('#'):
391 # Single-line comment
392 this_comment = mcomm.group(1)
393 elif start_line > -1:
394 # Not a single-line comment. But it cannot be the first.
395 if in_mlcomm == False:
396 mmlcomm = re.search(remlcomm_in, stripped)
397 if mmlcomm:
398 in_mlcomm = True
399 this_comment = ''
400 else:
401 mmlcomm = re.search(remlcomm_out, stripped)
402 if mmlcomm:
403 in_mlcomm = False
404 this_comment = ''
405 else:
406 this_comment = stripped
407
408 if this_comment is not None:
06ccae0f 409
424eef90 410 if start_line == -1:
06ccae0f 411
412 # First line. Check that we do not overlap with other comments
413 comment_overlaps = False
414 for c in comments:
415 if c.has_comment(line_num):
416 comment_overlaps = True
417 break
418
419 if comment_overlaps:
420 # No need to look for other comments
421 break
422
423 start_line = line_num
424
dde83b2b 425 end_line = line_num
06ccae0f 426 append = True
427
b4dd7d25 428 mclass_doxy = re.search(reclass_doxy, this_comment)
06ccae0f 429 if mclass_doxy:
ccf0fca4 430 class_name_doxy = mclass_doxy.group(2)
06ccae0f 431 append = False
432 else:
b4dd7d25 433 mauthor = re.search(reauthor, this_comment)
06ccae0f 434 if mauthor:
ea2f3fcb 435 author = mauthor.group(2)
06ccae0f 436 if date is None:
437 # Date specified in the standalone \date field has priority
ea2f3fcb 438 date = mauthor.group(4)
06ccae0f 439 append = False
440 else:
b4dd7d25 441 mdate = re.search(redate, this_comment)
06ccae0f 442 if mdate:
443 date = mdate.group(1)
444 append = False
fc7afee9 445 else:
b4dd7d25 446 mbrief = re.search(rebrief, this_comment)
fc7afee9 447 if mbrief:
448 brief = mbrief.group(1)
449 append = False
06ccae0f 450
451 if append:
b4dd7d25 452 comment_lines.append( this_comment )
06ccae0f 453
454 else:
424eef90 455 if start_line > 0:
06ccae0f 456 break
457
b4dd7d25 458 # This line had a valid comment
459 last_comm_line_num = line_num
460
06ccae0f 461 if class_name_doxy is None:
462
463 # No \class specified: guess it from file name
464 reclass = r'^(.*/)?(.*?)(\..*)?$'
465 mclass = re.search( reclass, filename )
466 if mclass:
467 class_name_doxy = mclass.group(2)
468 else:
469 assert False, 'Regexp unable to extract classname from file'
470
48a1eb94 471 # Macro or class?
472 if is_macro:
473 file_class_line = '\\file ' + class_name_doxy + '.C'
474 else:
475 file_class_line = '\\class ' + class_name_doxy
476
424eef90 477 if start_line > 0:
06ccae0f 478
fc7afee9 479 prepend_to_comment = []
480
481 # Prepend \class or \file specifier, then the \brief, then an empty line
482 prepend_to_comment.append( file_class_line )
483
484 if brief is not None:
485 prepend_to_comment.append( '\\brief ' + brief )
486 prepend_to_comment.append( '' )
487
488 comment_lines = prepend_to_comment + comment_lines # join lists
06ccae0f 489
84039997 490 # Append author and date if they exist
424eef90 491 if author is not None:
492 comment_lines.append( '\\author ' + author )
06ccae0f 493
424eef90 494 if date is not None:
495 comment_lines.append( '\\date ' + date )
496
b4dd7d25 497 # We should erase the "dumb" comments, such as "<class_name> class"
498 comm_idx = 0
499 regac = r'\s*%s\s+class\.?\s*' % class_name_doxy
500 mgac = None
501 for comm in comment_lines:
502 mgac = re.search(regac, comm)
503 if mgac:
504 break
505 comm_idx = comm_idx + 1
506 if mgac:
507 logging.debug('Removing dumb comment line: {%s}' % Colt(comment_lines[comm_idx]).magenta())
508 del comment_lines[comm_idx]
509
ee1793c7 510 comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
fc7afee9 511
512 # Now we look for a possible \brief
513 if brief is None:
514 comm_idx = 0
515 for comm in comment_lines:
516 if comm.startswith('\\class') or comm.startswith('\\file') or comm == '':
517 pass
518 else:
519 if len(comm) <= brief_len_threshold:
520 brief = comm
521 break
522 comm_idx = comm_idx + 1
523 if brief is not None:
f57ac84c 524 comment_lines = refactor_comment(
525 [ comment_lines[0], '\\brief ' + brief ] + comment_lines[1:comm_idx] + comment_lines[comm_idx+1:],
526 do_strip_html=False, infilename=filename)
fc7afee9 527
424eef90 528 logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
529 comments.append(Comment(
530 comment_lines,
531 start_line, 1, end_line, 1,
532 0, class_name_doxy
533 ))
534
535 else:
536
48a1eb94 537 logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
538 Colt(class_name_doxy).magenta())
539
540 comments.append(PrependComment(
541 [ file_class_line ],
542 1, 1, 1, 1,
543 0, class_name_doxy, append_empty=True
544 ))
06ccae0f 545
546
b9781053 547## Looks for a special ROOT ClassImp() entry.
548#
549# Doxygen might get confused by `ClassImp()` entries as they are macros normally written without
550# the ending `;`: this function wraps the definition inside a condition in order to make Doxygen
551# ignore it.
552#
553# @param filename Name of the current file
554# @param comments Array of comments: new ones will be appended there
555def comment_classimp(filename, comments):
556
557 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
558
559 line_num = 0
f0afbd73 560 reclassimp = r'^(\s*)Class(Imp|Def)\((.*?)\).*$'
b9781053 561
c03aba1e 562 in_classimp_cond = False
563 restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
564 reendcond = r'^\s*///\s*\\endcond\s*$'
565
b9781053 566 with open(filename, 'r') as fp:
567
7ab203b5 568 # Array of tuples: classimp, startcond, endcond
569 found_classimp = []
b9781053 570
7ab203b5 571 # Reset to nothing found
572 line_classimp = -1
573 line_startcond = -1
574 line_endcond = -1
575 classimp_class = None
576 classimp_indent = None
577
578 for line in fp:
f70a9ca0 579
b9781053 580 line_num = line_num + 1
c03aba1e 581
b9781053 582 mclassimp = re.search(reclassimp, line)
583 if mclassimp:
584
7ab203b5 585 # Dump previous one if appropriate, and reset
586 if line_classimp != -1:
587 found_classimp.append( (line_classimp, line_startcond, line_endcond) )
588 line_classimp = -1
589 line_startcond = -1
590 line_endcond = -1
591
b9781053 592 # Adjust indent
c03aba1e 593 classimp_indent = len( mclassimp.group(1) )
b9781053 594
c03aba1e 595 line_classimp = line_num
f0afbd73 596 classimp_class = mclassimp.group(3)
597 imp_or_def = mclassimp.group(2)
b9781053 598 logging.debug(
599 'Comment found for ' +
f0afbd73 600 Colt( 'Class%s(' % imp_or_def ).magenta() +
c03aba1e 601 Colt( classimp_class ).cyan() +
b9781053 602 Colt( ')' ).magenta() )
603
c03aba1e 604 else:
b9781053 605
c03aba1e 606 mstartcond = re.search(restartcond, line)
607 if mstartcond:
7ab203b5 608
609 # Dump previous one if appropriate, and reset
610 if line_classimp != -1:
611 found_classimp.append( (line_classimp, line_startcond, line_endcond) )
612 line_classimp = -1
613 line_startcond = -1
614 line_endcond = -1
615
616 logging.debug('Found Doxygen opening condition for ClassImp')
c03aba1e 617 in_classimp_cond = True
618 line_startcond = line_num
619
620 elif in_classimp_cond:
621
622 mendcond = re.search(reendcond, line)
623 if mendcond:
624 logging.debug('Found Doxygen closing condition for ClassImp')
625 in_classimp_cond = False
626 line_endcond = line_num
627
7ab203b5 628 # Dump previous one if appropriate, and reset (out of the loop)
629 if line_classimp != -1:
630 found_classimp.append( (line_classimp, line_startcond, line_endcond) )
631 line_classimp = -1
632 line_startcond = -1
633 line_endcond = -1
c03aba1e 634
7ab203b5 635 for line_classimp,line_startcond,line_endcond in found_classimp:
636
637 # Loop over the ClassImp conditions we've found
638
639 if line_startcond != -1:
640 logging.debug('Looks like we are in a condition here %d,%d,%d' % (line_classimp, line_startcond, line_endcond))
641 comments.append(Comment(
642 ['\cond CLASSIMP'],
643 line_startcond, 1, line_startcond, 1,
644 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
645 append_empty=False
646 ))
647 else:
648 logging.debug('Looks like we are NOT NOT in a condition here %d,%d,%d' % (line_classimp, line_startcond, line_endcond))
649 comments.append(PrependComment(
650 ['\cond CLASSIMP'],
651 line_classimp, 1, line_classimp, 1,
652 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
653 ))
654
655 if line_endcond != -1:
656 comments.append(Comment(
657 ['\endcond'],
658 line_endcond, 1, line_endcond, 1,
659 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
660 append_empty=False
661 ))
662 else:
663 comments.append(PrependComment(
664 ['\endcond'],
665 line_classimp+1, 1, line_classimp+1, 1,
666 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
667 ))
b9781053 668
669
06ccae0f 670## Traverse the AST recursively starting from the current cursor.
671#
672# @param cursor A Clang parser cursor
673# @param filename Name of the current file
674# @param comments Array of comments: new ones will be appended there
675# @param recursion Current recursion depth
84039997 676# @param in_func True if we are inside a function or method
48a1eb94 677# @param classdesc_line_limit Do not look for comments after this line
678#
679# @return A tuple containing the classdesc_line_limit as first item
680def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
06ccae0f 681
682 # libclang traverses included files as well: we do not want this behavior
683 if cursor.location.file is not None and str(cursor.location.file) != filename:
684 logging.debug("Skipping processing of included %s" % cursor.location.file)
685 return
686
687 text = cursor.spelling or cursor.displayname
688 kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
689
ccf0fca4 690 is_macro = filename.endswith('.C')
691
06ccae0f 692 indent = ''
693 for i in range(0, recursion):
694 indent = indent + ' '
695
2d3fd2ec 696 if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
e215c6c6 697 clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
06ccae0f 698
48a1eb94 699 if classdesc_line_limit is None:
700 classdesc_line_limit = cursor.location.line
701
06ccae0f 702 # cursor ran into a C++ method
703 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
704 comment_method(cursor, comments)
84039997 705 in_func = True
06ccae0f 706
84039997 707 elif not is_macro and not in_func and \
708 cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
06ccae0f 709
48a1eb94 710 if classdesc_line_limit is None:
711 classdesc_line_limit = cursor.location.line
712
06ccae0f 713 # cursor ran into a data member declaration
714 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
715 comment_datamember(cursor, comments)
716
717 else:
718
719 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
720
721 for child_cursor in cursor.get_children():
48a1eb94 722 classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
06ccae0f 723
724 if recursion == 0:
b9781053 725 comment_classimp(filename, comments)
48a1eb94 726 comment_classdesc(filename, comments, classdesc_line_limit)
727
728 return classdesc_line_limit
06ccae0f 729
730
731## Strip some HTML tags from the given string. Returns clean string.
732#
733# @param s Input string
734def strip_html(s):
798167cb 735 rehtml = r'(?i)</?(P|BR)/?>'
06ccae0f 736 return re.sub(rehtml, '', s)
737
738
739## Remove garbage from comments and convert special tags from THtml to Doxygen.
740#
741# @param comment An array containing the lines of the original comment
ee1793c7 742def refactor_comment(comment, do_strip_html=True, infilename=None):
06ccae0f 743
5a3cf816 744 recomm = r'^(/{2,}|/\*)? ?(\s*)(.*?)\s*((/{2,})?\s*|\*/)$'
5eaf1dc7 745 regarbage = r'^(?i)\s*([\s*=_#-]+|(Begin|End)_Html)\s*$'
06ccae0f 746
03a6b7aa 747 # Support for LaTeX blocks spanning on multiple lines
748 relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
749 in_latex = False
750 latex_block = False
751
752 # Support for LaTeX blocks on a single line
9db428b7 753 reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
754
35b193b4 755 # Match <pre> (to turn it into the ~~~ Markdown syntax)
756 reblock = r'(?i)^(\s*)</?PRE>\s*$'
757
ee1793c7 758 # Macro blocks for pictures generation
759 in_macro = False
760 current_macro = []
761 remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
762
5a3cf816 763 # Minimum indent level: scale back everything
764 lowest_indent_level = None
765
766 # Indentation threshold: if too much indented, don't indent at all
1d38347b 767 indent_level_threshold = 7
5a3cf816 768
06ccae0f 769 new_comment = []
770 insert_blank = False
771 wait_first_non_blank = True
772 for line_comment in comment:
773
774 # Strip some HTML tags
775 if do_strip_html:
776 line_comment = strip_html(line_comment)
777
778 mcomm = re.search( recomm, line_comment )
779 if mcomm:
5a3cf816 780 new_line_comment = mcomm.group(2) + mcomm.group(3) # indent + comm
781
6bce9020 782 # Check if we are in a macro block
783 mmacro = re.search(remacro, new_line_comment)
784 if mmacro:
785 if in_macro:
786 in_macro = False
787
788 # Dump macro
789 outimg = write_macro(infilename, current_macro) + '.png'
790 current_macro = []
791
792 # Insert image
793 new_comment.append( '![Picture from ROOT macro](%s)' % (os.path.basename(outimg)) )
794
795 logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
796
797 else:
798 in_macro = True
799
800 continue
801 elif in_macro:
802 current_macro.append( new_line_comment )
803 continue
804
06ccae0f 805 mgarbage = re.search( regarbage, new_line_comment )
806
5a3cf816 807 if mgarbage is None and not mcomm.group(3).startswith('\\') and mcomm.group(3) != '':
808 # not a special command line: count indent
809 indent_level = len( mcomm.group(2) )
810 if lowest_indent_level is None or indent_level < lowest_indent_level:
811 lowest_indent_level = indent_level
812
813 # if indentation level is too much, consider it zero
814 if indent_level > indent_level_threshold:
815 new_line_comment = mcomm.group(3) # remove ALL indentation
816
06ccae0f 817 if new_line_comment == '' or mgarbage is not None:
818 insert_blank = True
819 else:
820 if insert_blank and not wait_first_non_blank:
821 new_comment.append('')
822 insert_blank = False
823 wait_first_non_blank = False
9db428b7 824
825 # Postprocessing: LaTeX formulas in ROOT format
826 # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
827 # There can be several ROOT LaTeX forumlas per line
828 while True:
829 minline_latex = re.search( reinline_latex, new_line_comment )
830 if minline_latex:
831 new_line_comment = '%s\\f$%s\\f$%s' % \
832 ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
833 minline_latex.group(3) )
834 else:
835 break
836
03a6b7aa 837 # ROOT LaTeX: do we have a Begin/End_LaTeX block?
838 # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
839 # block here left without a corresponding ending block
840 mlatex = re.search( relatex, new_line_comment )
841 if mlatex:
842
843 # before and after parts have been already stripped
844 l_before = mlatex.group(2)
845 l_after = mlatex.group(4)
846 is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
847
848 if l_before is None:
849 l_before = ''
850 if l_after is None:
851 l_after = ''
852
853 if is_begin:
854
855 # Begin of LaTeX part
856
857 in_latex = True
858 if l_before == '' and l_after == '':
859
860 # Opening tag alone: mark the beginning of a block: \f[ ... \f]
861 latex_block = True
862 new_comment.append( '\\f[' )
863
864 else:
865 # Mark the beginning of inline: \f$ ... \f$
866 latex_block = False
867 new_comment.append(
868 '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
869 )
870
871 else:
872
873 # End of LaTeX part
874 in_latex = False
875
876 if latex_block:
877
878 # Closing a LaTeX block
879 if l_before != '':
880 new_comment.append( l_before.replace('#', '\\') )
881 new_comment.append( '\\f]' )
882 if l_after != '':
883 new_comment.append( l_after )
884
885 else:
886
887 # Closing a LaTeX inline
888 new_comment.append(
889 '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
890 )
891
892 # Prevent appending lines (we have already done that)
893 new_line_comment = None
894
35b193b4 895 # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
896 # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
897 if new_line_comment is not None and not in_latex:
898
899 mblock = re.search( reblock, new_line_comment )
900 if mblock:
901 new_comment.append( mblock.group(1)+'~~~' )
902 new_line_comment = None
903
03a6b7aa 904 if new_line_comment is not None:
905 if in_latex:
906 new_line_comment = new_line_comment.replace('#', '\\')
907 new_comment.append( new_line_comment )
06ccae0f 908
909 else:
910 assert False, 'Comment regexp does not match'
911
5a3cf816 912 # Fixing indentation level
913 if lowest_indent_level is not None:
914 logging.debug('Lowest indentation level found: %d' % lowest_indent_level)
915
916 new_comment_indent = []
917 reblankstart = r'^\s+'
918 for line in new_comment:
919 if re.search(reblankstart, line):
920 new_comment_indent.append( line[lowest_indent_level:] )
921 else:
922 new_comment_indent.append( line )
923
924 new_comment = new_comment_indent
925
926 else:
927 logging.debug('No indentation scaling applied')
928
06ccae0f 929 return new_comment
930
931
ee1793c7 932## Dumps an image-generating macro to the correct place. Returns a string with the image path,
933# without the extension.
934#
935# @param infilename File name of the source file
936# @param macro_lines Array of macro lines
937def write_macro(infilename, macro_lines):
938
939 # Calculate hash
940 digh = hashlib.sha1()
941 for l in macro_lines:
942 digh.update(l)
943 digh.update('\n')
944 short_digest = digh.hexdigest()[0:7]
945
3c338ad2 946 infiledir = os.path.dirname(infilename)
947 if infiledir == '':
948 infiledir = '.'
949 outdir = '%s/imgdoc' % infiledir
ee1793c7 950 outprefix = '%s/%s_%s' % (
951 outdir,
952 os.path.basename(infilename).replace('.', '_'),
953 short_digest
954 )
955 outmacro = '%s.C' % outprefix
956
957 # Make directory
958 if not os.path.isdir(outdir):
959 # do not catch: let everything die on error
960 logging.debug('Creating directory %s' % Colt(outdir).magenta())
961 os.mkdir(outdir)
962
963 # Create file (do not catch errors either)
964 with open(outmacro, 'w') as omfp:
965 logging.debug('Writing macro %s' % Colt(outmacro).magenta())
966 for l in macro_lines:
967 omfp.write(l)
968 omfp.write('\n')
969
970 return outprefix
971
972
06ccae0f 973## Rewrites all comments from the given file handler.
974#
975# @param fhin The file handler to read from
976# @param fhout The file handler to write to
977# @param comments Array of comments
978def rewrite_comments(fhin, fhout, comments):
979
980 line_num = 0
981 in_comment = False
982 skip_empty = False
983 comm = None
984 prev_comm = None
b9781053 985 restore_lines = None
06ccae0f 986
987 rindent = r'^(\s*)'
988
b9781053 989 def dump_comment_block(cmt, restore=None):
fc82f9c7 990 text_indent = ''
b9781053 991 ask_skip_empty = False
992
fc82f9c7 993 for i in range(0, cmt.indent):
994 text_indent = text_indent + ' '
995
996 for lc in cmt.lines:
997 fhout.write('%s///' % text_indent )
998 lc = lc.rstrip()
999 if len(lc) != 0:
1000 fhout.write(' ')
1001 fhout.write(lc)
1002 fhout.write('\n')
1003
1004 # Empty new line at the end of the comment
c03aba1e 1005 if cmt.append_empty:
b9781053 1006 fhout.write('\n')
1007 ask_skip_empty = True
1008
1009 # Restore lines if possible
1010 if restore:
1011 for lr in restore:
1012 fhout.write(lr)
1013 fhout.write('\n')
1014
1015 # Tell the caller whether it should skip the next empty line found
1016 return ask_skip_empty
c0094f3c 1017
1018
06ccae0f 1019 for line in fhin:
1020
1021 line_num = line_num + 1
1022
1023 # Find current comment
1024 prev_comm = comm
1025 comm = None
1fd0e6c3 1026 comm_list = []
06ccae0f 1027 for c in comments:
1028 if c.has_comment(line_num):
1029 comm = c
1fd0e6c3 1030 comm_list.append(c)
1031
1032 if len(comm_list) > 1:
1033
1034 merged = True
1035
1036 if len(comm_list) == 2:
1037 c1,c2 = comm_list
1038 if isinstance(c1, Comment) and isinstance(c2, Comment):
1039 c1.lines = c1.lines + c2.lines # list merge
1040 comm = c1
1041 logging.debug('Two adjacent comments merged. Result: {%s}' % Colt(comm).cyan())
1042 else:
1043 merged = False
1044 else:
1045 merged = False
1046
1047 if merged == False:
1048 logging.warning('Too many unmergeable comments on the same line (%d), picking the last one' % len(comm_list))
1049 for c in comm_list:
1050 logging.warning('>> %s' % c)
1051 comm = c # considering the last one
06ccae0f 1052
1053 if comm:
1054
b9781053 1055 # First thing to check: are we in the same comment as before?
a2245741 1056 if comm is not prev_comm and \
1057 isinstance(comm, Comment) and \
1058 isinstance(prev_comm, Comment) and \
1059 not isinstance(prev_comm, RemoveComment):
1060
1061 # We are NOT in the same comment as before, and this comment is dumpable
b9781053 1062
1063 skip_empty = dump_comment_block(prev_comm, restore_lines)
1064 in_comment = False
1065 restore_lines = None
1066 prev_comm = None # we have just dumped it: pretend it never existed in this loop
1067
1068 #
1069 # Check type of comment and react accordingly
1070 #
1071
06ccae0f 1072 if isinstance(comm, MemberComment):
c0094f3c 1073
1074 # end comment block
1075 if in_comment:
b9781053 1076 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 1077 in_comment = False
b9781053 1078 restore_lines = None
c0094f3c 1079
06ccae0f 1080 non_comment = line[ 0:comm.first_col-1 ]
1081
21689d7a 1082 if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
06ccae0f 1083
21689d7a 1084 # This is a special case: comment will be split in two lines: one before the comment for
1085 # Doxygen as "member description", and the other right after the comment on the same line
1086 # to be parsed by ROOT's C++ parser
1087
1088 # Keep indent on the generated line of comment before member definition
06ccae0f 1089 mindent = re.search(rindent, line)
21689d7a 1090
1091 # Get correct comment flag, if any
1092 if comm.comment_flag is not None:
1093 cflag = comm.comment_flag
1094 else:
1095 cflag = ''
1096
1097 # Get correct array size, if any
1098 if comm.array_size is not None:
1099 asize = '[%s]' % comm.array_size
06ccae0f 1100 else:
21689d7a 1101 asize = ''
06ccae0f 1102
21689d7a 1103 # Write on two lines
1104 fhout.write('%s/// %s\n%s//%s%s\n' % (
06ccae0f 1105 mindent.group(1),
1106 comm.lines[0],
1107 non_comment,
21689d7a 1108 cflag,
1109 asize
06ccae0f 1110 ))
1111
1112 else:
1113
21689d7a 1114 # Single-line comments with the "transient" flag can be kept on one line in a way that
1115 # they are correctly interpreted by both ROOT and Doxygen
1116
1117 if comm.is_transient():
06ccae0f 1118 tt = '!'
1119 else:
1120 tt = '/'
1121
1122 fhout.write('%s//%s< %s\n' % (
1123 non_comment,
1124 tt,
1125 comm.lines[0]
1126 ))
1127
1128 elif isinstance(comm, RemoveComment):
c0094f3c 1129 # End comment block and skip this line
1130 if in_comment:
b9781053 1131 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 1132 in_comment = False
b9781053 1133 restore_lines = None
06ccae0f 1134
22679479 1135 elif restore_lines is None:
c0094f3c 1136
b9781053 1137 # Beginning of a new comment block of type Comment or PrependComment
06ccae0f 1138 in_comment = True
1139
b9781053 1140 if isinstance(comm, PrependComment):
1141 # Prepare array of lines to dump right after the comment
1142 restore_lines = [ line.rstrip('\n') ]
b53c40e9 1143 logging.debug('Commencing lines to restore: {%s}' % Colt(restore_lines[0]).cyan())
b9781053 1144 else:
5ebea9f0 1145 # Extract the non-comment part and print it if it exists. If this is the first line of a
1146 # comment, it might happen something like `valid_code; // this is a comment`.
1147 if comm.first_line == line_num:
1148 non_comment = line[ 0:comm.first_col-1 ].rstrip()
1149 if non_comment != '':
1150 fhout.write( non_comment + '\n' )
b9781053 1151
1152 elif isinstance(comm, Comment):
1153
1154 if restore_lines is not None:
1155 # From the 2nd line on of comment to prepend
1156 restore_lines.append( line.rstrip('\n') )
b53c40e9 1157 logging.debug('Appending lines to restore. All lines: {%s}' % Colt(restore_lines).cyan())
b9781053 1158
1159 else:
a2245741 1160 assert False, 'Unhandled parser state: line=%d comm={%s} prev_comm={%s}' % \
b9781053 1161 (line_num, comm, prev_comm)
06ccae0f 1162
1163 else:
1164
b9781053 1165 # Not a comment line
1166
06ccae0f 1167 if in_comment:
1168
1169 # We have just exited a comment block of type Comment
b9781053 1170 skip_empty = dump_comment_block(prev_comm, restore_lines)
06ccae0f 1171 in_comment = False
b9781053 1172 restore_lines = None
06ccae0f 1173
b9781053 1174 # Dump the non-comment line
06ccae0f 1175 line_out = line.rstrip('\n')
1176 if skip_empty:
1177 skip_empty = False
1178 if line_out.strip() != '':
1179 fhout.write( line_out + '\n' )
1180 else:
1181 fhout.write( line_out + '\n' )
1182
7ff022be 1183 # Is there some comment left here?
1184 if restore_lines is not None:
1185 dump_comment_block(comm, restore_lines)
1186
1187 # Is there some other comment beyond the last line?
1188 for c in comments:
1189 if c.has_comment(line_num+1):
1190 dump_comment_block(c, None)
1191 break
1192
f329fa92 1193
1194## The main function.
1195#
06ccae0f 1196# Return value is the executable's return value.
f329fa92 1197def main(argv):
1198
06ccae0f 1199 # Setup logging on stderr
1200 log_level = logging.INFO
1201 logging.basicConfig(
1202 level=log_level,
1203 format='%(levelname)-8s %(funcName)-20s %(message)s',
1204 stream=sys.stderr
1205 )
f329fa92 1206
06ccae0f 1207 # Parse command-line options
1208 output_on_stdout = False
7afec95e 1209 include_flags = []
06ccae0f 1210 try:
7afec95e 1211 opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
06ccae0f 1212 for o, a in opts:
1213 if o == '--debug':
1214 log_level = getattr( logging, a.upper(), None )
1215 if not isinstance(log_level, int):
1216 raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
1217 elif o == '-d':
1218 log_level = logging.DEBUG
1219 elif o == '-o' or o == '--stdout':
06ccae0f 1220 output_on_stdout = True
7afec95e 1221 elif o == '-I':
1222 if os.path.isdir(a):
1223 include_flags.extend( [ '-I', a ] )
1224 else:
1225 logging.fatal('Include directory not found: %s' % Colt(a).magenta())
1226 return 2
06ccae0f 1227 else:
1228 assert False, 'Unhandled argument'
1229 except getopt.GetoptError as e:
1230 logging.fatal('Invalid arguments: %s' % e)
1231 return 1
f329fa92 1232
06ccae0f 1233 logging.getLogger('').setLevel(log_level)
f329fa92 1234
06ccae0f 1235 # Attempt to load libclang from a list of known locations
1236 libclang_locations = [
1237 '/usr/lib/llvm-3.5/lib/libclang.so.1',
1238 '/usr/lib/libclang.so',
1239 '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
1240 ]
1241 libclang_found = False
f329fa92 1242
06ccae0f 1243 for lib in libclang_locations:
1244 if os.path.isfile(lib):
1245 clang.cindex.Config.set_library_file(lib)
1246 libclang_found = True
1247 break
f329fa92 1248
06ccae0f 1249 if not libclang_found:
1250 logging.fatal('Cannot find libclang')
1251 return 1
1252
1253 # Loop over all files
1254 for fn in args:
1255
1256 logging.info('Input file: %s' % Colt(fn).magenta())
1257 index = clang.cindex.Index.create()
7afec95e 1258 clang_args = [ '-x', 'c++' ]
1259 clang_args.extend( include_flags )
1260 translation_unit = index.parse(fn, args=clang_args)
06ccae0f 1261
1262 comments = []
1263 traverse_ast( translation_unit.cursor, fn, comments )
1264 for c in comments:
1265
1266 logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
f329fa92 1267
06ccae0f 1268 if isinstance(c, MemberComment):
1269
21689d7a 1270 if c.is_transient():
1271 flag_text = Colt('transient ').yellow()
1272 elif c.is_dontsplit():
1273 flag_text = Colt('dontsplit ').yellow()
1274 elif c.is_ptr():
1275 flag_text = Colt('ptr ').yellow()
06ccae0f 1276 else:
21689d7a 1277 flag_text = ''
06ccae0f 1278
1279 if c.array_size is not None:
1280 array_text = Colt('arraysize=%s ' % c.array_size).yellow()
1281 else:
1282 array_text = ''
1283
1284 logging.debug(
1285 "%s %s%s{%s}" % ( \
1286 Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
21689d7a 1287 flag_text,
06ccae0f 1288 array_text,
1289 Colt(c.lines[0]).cyan()
1290 ))
1291
1292 elif isinstance(c, RemoveComment):
1293
1294 logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
1295
1296 else:
1297 for l in c.lines:
1298 logging.debug(
1299 Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
1300 "{%s}" % Colt(l).cyan()
1301 )
f329fa92 1302
1303 try:
06ccae0f 1304
1305 if output_on_stdout:
1306 with open(fn, 'r') as fhin:
1307 rewrite_comments( fhin, sys.stdout, comments )
1308 else:
1309 fn_back = fn + '.thtml2doxy_backup'
1310 os.rename( fn, fn_back )
1311
1312 with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
1313 rewrite_comments( fhin, fhout, comments )
1314
1315 os.remove( fn_back )
1316 logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
1317 except (IOError,OSError) as e:
1318 logging.error('File operation failed: %s' % e)
f329fa92 1319
1320 return 0
1321
06ccae0f 1322
f329fa92 1323if __name__ == '__main__':
06ccae0f 1324 sys.exit( main( sys.argv[1:] ) )