doxy: consider comments after last line
[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17# (outdated, API calls described there do not work anymore, but useful to understand some basic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*$'
fc7afee9 339 rebrief = r'(?i)^\s*\\brief\s*(.*)\s*$'
06ccae0f 340 author = None
341 date = None
fc7afee9 342 brief = None
343 brief_len_threshold = 80
06ccae0f 344
345 comment_lines = []
346
347 start_line = -1
348 end_line = -1
349
350 line_num = 0
351
ccf0fca4 352 is_macro = filename.endswith('.C')
353
06ccae0f 354 with open(filename, 'r') as fp:
355
356 for raw in fp:
357
358 line_num = line_num + 1
359
48a1eb94 360 if look_no_further_than_line is not None and line_num == look_no_further_than_line:
361 logging.debug('Stopping at line %d while looking for class/file description' % \
362 look_no_further_than_line)
363 break
364
424eef90 365 if raw.strip() == '' and start_line > 0:
06ccae0f 366 # Skip empty lines
06ccae0f 367 continue
368
369 stripped = strip_html(raw)
370 mcomm = re.search(recomm, stripped)
371 if mcomm:
372
424eef90 373 if start_line == -1:
06ccae0f 374
375 # First line. Check that we do not overlap with other comments
376 comment_overlaps = False
377 for c in comments:
378 if c.has_comment(line_num):
379 comment_overlaps = True
380 break
381
382 if comment_overlaps:
383 # No need to look for other comments
384 break
385
386 start_line = line_num
387
dde83b2b 388 end_line = line_num
06ccae0f 389 append = True
390
391 mclass_doxy = re.search(reclass_doxy, mcomm.group(1))
392 if mclass_doxy:
ccf0fca4 393 class_name_doxy = mclass_doxy.group(2)
06ccae0f 394 append = False
395 else:
396 mauthor = re.search(reauthor, mcomm.group(1))
397 if mauthor:
ea2f3fcb 398 author = mauthor.group(2)
06ccae0f 399 if date is None:
400 # Date specified in the standalone \date field has priority
ea2f3fcb 401 date = mauthor.group(4)
06ccae0f 402 append = False
403 else:
404 mdate = re.search(redate, mcomm.group(1))
405 if mdate:
406 date = mdate.group(1)
407 append = False
fc7afee9 408 else:
409 mbrief = re.search(rebrief, mcomm.group(1))
410 if mbrief:
411 brief = mbrief.group(1)
412 append = False
06ccae0f 413
414 if append:
415 comment_lines.append( mcomm.group(1) )
416
417 else:
424eef90 418 if start_line > 0:
06ccae0f 419 break
420
421 if class_name_doxy is None:
422
423 # No \class specified: guess it from file name
424 reclass = r'^(.*/)?(.*?)(\..*)?$'
425 mclass = re.search( reclass, filename )
426 if mclass:
427 class_name_doxy = mclass.group(2)
428 else:
429 assert False, 'Regexp unable to extract classname from file'
430
48a1eb94 431 # Macro or class?
432 if is_macro:
433 file_class_line = '\\file ' + class_name_doxy + '.C'
434 else:
435 file_class_line = '\\class ' + class_name_doxy
436
424eef90 437 if start_line > 0:
06ccae0f 438
fc7afee9 439 prepend_to_comment = []
440
441 # Prepend \class or \file specifier, then the \brief, then an empty line
442 prepend_to_comment.append( file_class_line )
443
444 if brief is not None:
445 prepend_to_comment.append( '\\brief ' + brief )
446 prepend_to_comment.append( '' )
447
448 comment_lines = prepend_to_comment + comment_lines # join lists
06ccae0f 449
84039997 450 # Append author and date if they exist
424eef90 451 if author is not None:
452 comment_lines.append( '\\author ' + author )
06ccae0f 453
424eef90 454 if date is not None:
455 comment_lines.append( '\\date ' + date )
456
ee1793c7 457 comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
fc7afee9 458
459 # Now we look for a possible \brief
460 if brief is None:
461 comm_idx = 0
462 for comm in comment_lines:
463 if comm.startswith('\\class') or comm.startswith('\\file') or comm == '':
464 pass
465 else:
466 if len(comm) <= brief_len_threshold:
467 brief = comm
468 break
469 comm_idx = comm_idx + 1
470 if brief is not None:
f57ac84c 471 comment_lines = refactor_comment(
472 [ comment_lines[0], '\\brief ' + brief ] + comment_lines[1:comm_idx] + comment_lines[comm_idx+1:],
473 do_strip_html=False, infilename=filename)
fc7afee9 474
424eef90 475 logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
476 comments.append(Comment(
477 comment_lines,
478 start_line, 1, end_line, 1,
479 0, class_name_doxy
480 ))
481
482 else:
483
48a1eb94 484 logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
485 Colt(class_name_doxy).magenta())
486
487 comments.append(PrependComment(
488 [ file_class_line ],
489 1, 1, 1, 1,
490 0, class_name_doxy, append_empty=True
491 ))
06ccae0f 492
493
b9781053 494## Looks for a special ROOT ClassImp() entry.
495#
496# Doxygen might get confused by `ClassImp()` entries as they are macros normally written without
497# the ending `;`: this function wraps the definition inside a condition in order to make Doxygen
498# ignore it.
499#
500# @param filename Name of the current file
501# @param comments Array of comments: new ones will be appended there
502def comment_classimp(filename, comments):
503
504 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
505
506 line_num = 0
f0afbd73 507 reclassimp = r'^(\s*)Class(Imp|Def)\((.*?)\).*$'
b9781053 508
c03aba1e 509 in_classimp_cond = False
510 restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
511 reendcond = r'^\s*///\s*\\endcond\s*$'
512
b9781053 513 with open(filename, 'r') as fp:
514
515 for line in fp:
516
f70a9ca0 517 # Reset to nothing found
518 line_classimp = -1
519 line_startcond = -1
520 line_endcond = -1
521 classimp_class = None
522 classimp_indent = None
523
b9781053 524 line_num = line_num + 1
c03aba1e 525
b9781053 526 mclassimp = re.search(reclassimp, line)
527 if mclassimp:
528
529 # Adjust indent
c03aba1e 530 classimp_indent = len( mclassimp.group(1) )
b9781053 531
c03aba1e 532 line_classimp = line_num
f0afbd73 533 classimp_class = mclassimp.group(3)
534 imp_or_def = mclassimp.group(2)
b9781053 535 logging.debug(
536 'Comment found for ' +
f0afbd73 537 Colt( 'Class%s(' % imp_or_def ).magenta() +
c03aba1e 538 Colt( classimp_class ).cyan() +
b9781053 539 Colt( ')' ).magenta() )
540
c03aba1e 541 else:
b9781053 542
c03aba1e 543 mstartcond = re.search(restartcond, line)
544 if mstartcond:
545 logging.debug('Found Doxygen opening condition for ClassImp in {%s}' % line)
546 in_classimp_cond = True
547 line_startcond = line_num
548
549 elif in_classimp_cond:
550
551 mendcond = re.search(reendcond, line)
552 if mendcond:
553 logging.debug('Found Doxygen closing condition for ClassImp')
554 in_classimp_cond = False
555 line_endcond = line_num
556
f70a9ca0 557 # Did we find something?
558 if line_classimp != -1:
c03aba1e 559
f70a9ca0 560 if line_startcond != -1:
561 comments.append(Comment(
562 ['\cond CLASSIMP'],
563 line_startcond, 1, line_startcond, 1,
564 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
565 append_empty=False
566 ))
567 else:
568 comments.append(PrependComment(
569 ['\cond CLASSIMP'],
570 line_classimp, 1, line_classimp, 1,
571 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
572 ))
573
574 if line_endcond != -1:
575 comments.append(Comment(
576 ['\endcond'],
577 line_endcond, 1, line_endcond, 1,
578 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
579 append_empty=False
580 ))
581 else:
582 comments.append(PrependComment(
583 ['\endcond'],
584 line_classimp+1, 1, line_classimp+1, 1,
585 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
586 ))
b9781053 587
588
06ccae0f 589## Traverse the AST recursively starting from the current cursor.
590#
591# @param cursor A Clang parser cursor
592# @param filename Name of the current file
593# @param comments Array of comments: new ones will be appended there
594# @param recursion Current recursion depth
84039997 595# @param in_func True if we are inside a function or method
48a1eb94 596# @param classdesc_line_limit Do not look for comments after this line
597#
598# @return A tuple containing the classdesc_line_limit as first item
599def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
06ccae0f 600
601 # libclang traverses included files as well: we do not want this behavior
602 if cursor.location.file is not None and str(cursor.location.file) != filename:
603 logging.debug("Skipping processing of included %s" % cursor.location.file)
604 return
605
606 text = cursor.spelling or cursor.displayname
607 kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
608
ccf0fca4 609 is_macro = filename.endswith('.C')
610
06ccae0f 611 indent = ''
612 for i in range(0, recursion):
613 indent = indent + ' '
614
2d3fd2ec 615 if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
e215c6c6 616 clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
06ccae0f 617
48a1eb94 618 if classdesc_line_limit is None:
619 classdesc_line_limit = cursor.location.line
620
06ccae0f 621 # cursor ran into a C++ method
622 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
623 comment_method(cursor, comments)
84039997 624 in_func = True
06ccae0f 625
84039997 626 elif not is_macro and not in_func and \
627 cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
06ccae0f 628
48a1eb94 629 if classdesc_line_limit is None:
630 classdesc_line_limit = cursor.location.line
631
06ccae0f 632 # cursor ran into a data member declaration
633 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
634 comment_datamember(cursor, comments)
635
636 else:
637
638 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
639
640 for child_cursor in cursor.get_children():
48a1eb94 641 classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
06ccae0f 642
643 if recursion == 0:
b9781053 644 comment_classimp(filename, comments)
48a1eb94 645 comment_classdesc(filename, comments, classdesc_line_limit)
646
647 return classdesc_line_limit
06ccae0f 648
649
650## Strip some HTML tags from the given string. Returns clean string.
651#
652# @param s Input string
653def strip_html(s):
798167cb 654 rehtml = r'(?i)</?(P|BR)/?>'
06ccae0f 655 return re.sub(rehtml, '', s)
656
657
658## Remove garbage from comments and convert special tags from THtml to Doxygen.
659#
660# @param comment An array containing the lines of the original comment
ee1793c7 661def refactor_comment(comment, do_strip_html=True, infilename=None):
06ccae0f 662
5a3cf816 663 recomm = r'^(/{2,}|/\*)? ?(\s*)(.*?)\s*((/{2,})?\s*|\*/)$'
5eaf1dc7 664 regarbage = r'^(?i)\s*([\s*=_#-]+|(Begin|End)_Html)\s*$'
06ccae0f 665
03a6b7aa 666 # Support for LaTeX blocks spanning on multiple lines
667 relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
668 in_latex = False
669 latex_block = False
670
671 # Support for LaTeX blocks on a single line
9db428b7 672 reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
673
35b193b4 674 # Match <pre> (to turn it into the ~~~ Markdown syntax)
675 reblock = r'(?i)^(\s*)</?PRE>\s*$'
676
ee1793c7 677 # Macro blocks for pictures generation
678 in_macro = False
679 current_macro = []
680 remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
681
5a3cf816 682 # Minimum indent level: scale back everything
683 lowest_indent_level = None
684
685 # Indentation threshold: if too much indented, don't indent at all
1d38347b 686 indent_level_threshold = 7
5a3cf816 687
06ccae0f 688 new_comment = []
689 insert_blank = False
690 wait_first_non_blank = True
691 for line_comment in comment:
692
ee1793c7 693 # Check if we are in a macro block
694 mmacro = re.search(remacro, line_comment)
695 if mmacro:
696 if in_macro:
697 in_macro = False
698
699 # Dump macro
700 outimg = write_macro(infilename, current_macro) + '.png'
701 current_macro = []
702
703 # Insert image
430c67d2 704 new_comment.append( '![Picture from ROOT macro](%s)' % (os.path.basename(outimg)) )
ee1793c7 705
706 logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
707
708 else:
709 in_macro = True
710
711 continue
712 elif in_macro:
713 current_macro.append( line_comment )
714 continue
715
06ccae0f 716 # Strip some HTML tags
717 if do_strip_html:
718 line_comment = strip_html(line_comment)
719
720 mcomm = re.search( recomm, line_comment )
721 if mcomm:
5a3cf816 722 new_line_comment = mcomm.group(2) + mcomm.group(3) # indent + comm
723
06ccae0f 724 mgarbage = re.search( regarbage, new_line_comment )
725
5a3cf816 726 if mgarbage is None and not mcomm.group(3).startswith('\\') and mcomm.group(3) != '':
727 # not a special command line: count indent
728 indent_level = len( mcomm.group(2) )
729 if lowest_indent_level is None or indent_level < lowest_indent_level:
730 lowest_indent_level = indent_level
731
732 # if indentation level is too much, consider it zero
733 if indent_level > indent_level_threshold:
734 new_line_comment = mcomm.group(3) # remove ALL indentation
735
06ccae0f 736 if new_line_comment == '' or mgarbage is not None:
737 insert_blank = True
738 else:
739 if insert_blank and not wait_first_non_blank:
740 new_comment.append('')
741 insert_blank = False
742 wait_first_non_blank = False
9db428b7 743
744 # Postprocessing: LaTeX formulas in ROOT format
745 # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
746 # There can be several ROOT LaTeX forumlas per line
747 while True:
748 minline_latex = re.search( reinline_latex, new_line_comment )
749 if minline_latex:
750 new_line_comment = '%s\\f$%s\\f$%s' % \
751 ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
752 minline_latex.group(3) )
753 else:
754 break
755
03a6b7aa 756 # ROOT LaTeX: do we have a Begin/End_LaTeX block?
757 # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
758 # block here left without a corresponding ending block
759 mlatex = re.search( relatex, new_line_comment )
760 if mlatex:
761
762 # before and after parts have been already stripped
763 l_before = mlatex.group(2)
764 l_after = mlatex.group(4)
765 is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
766
767 if l_before is None:
768 l_before = ''
769 if l_after is None:
770 l_after = ''
771
772 if is_begin:
773
774 # Begin of LaTeX part
775
776 in_latex = True
777 if l_before == '' and l_after == '':
778
779 # Opening tag alone: mark the beginning of a block: \f[ ... \f]
780 latex_block = True
781 new_comment.append( '\\f[' )
782
783 else:
784 # Mark the beginning of inline: \f$ ... \f$
785 latex_block = False
786 new_comment.append(
787 '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
788 )
789
790 else:
791
792 # End of LaTeX part
793 in_latex = False
794
795 if latex_block:
796
797 # Closing a LaTeX block
798 if l_before != '':
799 new_comment.append( l_before.replace('#', '\\') )
800 new_comment.append( '\\f]' )
801 if l_after != '':
802 new_comment.append( l_after )
803
804 else:
805
806 # Closing a LaTeX inline
807 new_comment.append(
808 '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
809 )
810
811 # Prevent appending lines (we have already done that)
812 new_line_comment = None
813
35b193b4 814 # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
815 # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
816 if new_line_comment is not None and not in_latex:
817
818 mblock = re.search( reblock, new_line_comment )
819 if mblock:
820 new_comment.append( mblock.group(1)+'~~~' )
821 new_line_comment = None
822
03a6b7aa 823 if new_line_comment is not None:
824 if in_latex:
825 new_line_comment = new_line_comment.replace('#', '\\')
826 new_comment.append( new_line_comment )
06ccae0f 827
828 else:
829 assert False, 'Comment regexp does not match'
830
5a3cf816 831 # Fixing indentation level
832 if lowest_indent_level is not None:
833 logging.debug('Lowest indentation level found: %d' % lowest_indent_level)
834
835 new_comment_indent = []
836 reblankstart = r'^\s+'
837 for line in new_comment:
838 if re.search(reblankstart, line):
839 new_comment_indent.append( line[lowest_indent_level:] )
840 else:
841 new_comment_indent.append( line )
842
843 new_comment = new_comment_indent
844
845 else:
846 logging.debug('No indentation scaling applied')
847
06ccae0f 848 return new_comment
849
850
ee1793c7 851## Dumps an image-generating macro to the correct place. Returns a string with the image path,
852# without the extension.
853#
854# @param infilename File name of the source file
855# @param macro_lines Array of macro lines
856def write_macro(infilename, macro_lines):
857
858 # Calculate hash
859 digh = hashlib.sha1()
860 for l in macro_lines:
861 digh.update(l)
862 digh.update('\n')
863 short_digest = digh.hexdigest()[0:7]
864
865 outdir = '%s/imgdoc' % os.path.dirname(infilename)
866 outprefix = '%s/%s_%s' % (
867 outdir,
868 os.path.basename(infilename).replace('.', '_'),
869 short_digest
870 )
871 outmacro = '%s.C' % outprefix
872
873 # Make directory
874 if not os.path.isdir(outdir):
875 # do not catch: let everything die on error
876 logging.debug('Creating directory %s' % Colt(outdir).magenta())
877 os.mkdir(outdir)
878
879 # Create file (do not catch errors either)
880 with open(outmacro, 'w') as omfp:
881 logging.debug('Writing macro %s' % Colt(outmacro).magenta())
882 for l in macro_lines:
883 omfp.write(l)
884 omfp.write('\n')
885
886 return outprefix
887
888
06ccae0f 889## Rewrites all comments from the given file handler.
890#
891# @param fhin The file handler to read from
892# @param fhout The file handler to write to
893# @param comments Array of comments
894def rewrite_comments(fhin, fhout, comments):
895
896 line_num = 0
897 in_comment = False
898 skip_empty = False
899 comm = None
900 prev_comm = None
b9781053 901 restore_lines = None
06ccae0f 902
903 rindent = r'^(\s*)'
904
b9781053 905 def dump_comment_block(cmt, restore=None):
fc82f9c7 906 text_indent = ''
b9781053 907 ask_skip_empty = False
908
fc82f9c7 909 for i in range(0, cmt.indent):
910 text_indent = text_indent + ' '
911
912 for lc in cmt.lines:
913 fhout.write('%s///' % text_indent )
914 lc = lc.rstrip()
915 if len(lc) != 0:
916 fhout.write(' ')
917 fhout.write(lc)
918 fhout.write('\n')
919
920 # Empty new line at the end of the comment
c03aba1e 921 if cmt.append_empty:
b9781053 922 fhout.write('\n')
923 ask_skip_empty = True
924
925 # Restore lines if possible
926 if restore:
927 for lr in restore:
928 fhout.write(lr)
929 fhout.write('\n')
930
931 # Tell the caller whether it should skip the next empty line found
932 return ask_skip_empty
c0094f3c 933
934
06ccae0f 935 for line in fhin:
936
937 line_num = line_num + 1
938
939 # Find current comment
940 prev_comm = comm
941 comm = None
1fd0e6c3 942 comm_list = []
06ccae0f 943 for c in comments:
944 if c.has_comment(line_num):
945 comm = c
1fd0e6c3 946 comm_list.append(c)
947
948 if len(comm_list) > 1:
949
950 merged = True
951
952 if len(comm_list) == 2:
953 c1,c2 = comm_list
954 if isinstance(c1, Comment) and isinstance(c2, Comment):
955 c1.lines = c1.lines + c2.lines # list merge
956 comm = c1
957 logging.debug('Two adjacent comments merged. Result: {%s}' % Colt(comm).cyan())
958 else:
959 merged = False
960 else:
961 merged = False
962
963 if merged == False:
964 logging.warning('Too many unmergeable comments on the same line (%d), picking the last one' % len(comm_list))
965 for c in comm_list:
966 logging.warning('>> %s' % c)
967 comm = c # considering the last one
06ccae0f 968
969 if comm:
970
b9781053 971 # First thing to check: are we in the same comment as before?
a2245741 972 if comm is not prev_comm and \
973 isinstance(comm, Comment) and \
974 isinstance(prev_comm, Comment) and \
975 not isinstance(prev_comm, RemoveComment):
976
977 # We are NOT in the same comment as before, and this comment is dumpable
b9781053 978
979 skip_empty = dump_comment_block(prev_comm, restore_lines)
980 in_comment = False
981 restore_lines = None
982 prev_comm = None # we have just dumped it: pretend it never existed in this loop
983
984 #
985 # Check type of comment and react accordingly
986 #
987
06ccae0f 988 if isinstance(comm, MemberComment):
c0094f3c 989
990 # end comment block
991 if in_comment:
b9781053 992 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 993 in_comment = False
b9781053 994 restore_lines = None
c0094f3c 995
06ccae0f 996 non_comment = line[ 0:comm.first_col-1 ]
997
21689d7a 998 if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
06ccae0f 999
21689d7a 1000 # This is a special case: comment will be split in two lines: one before the comment for
1001 # Doxygen as "member description", and the other right after the comment on the same line
1002 # to be parsed by ROOT's C++ parser
1003
1004 # Keep indent on the generated line of comment before member definition
06ccae0f 1005 mindent = re.search(rindent, line)
21689d7a 1006
1007 # Get correct comment flag, if any
1008 if comm.comment_flag is not None:
1009 cflag = comm.comment_flag
1010 else:
1011 cflag = ''
1012
1013 # Get correct array size, if any
1014 if comm.array_size is not None:
1015 asize = '[%s]' % comm.array_size
06ccae0f 1016 else:
21689d7a 1017 asize = ''
06ccae0f 1018
21689d7a 1019 # Write on two lines
1020 fhout.write('%s/// %s\n%s//%s%s\n' % (
06ccae0f 1021 mindent.group(1),
1022 comm.lines[0],
1023 non_comment,
21689d7a 1024 cflag,
1025 asize
06ccae0f 1026 ))
1027
1028 else:
1029
21689d7a 1030 # Single-line comments with the "transient" flag can be kept on one line in a way that
1031 # they are correctly interpreted by both ROOT and Doxygen
1032
1033 if comm.is_transient():
06ccae0f 1034 tt = '!'
1035 else:
1036 tt = '/'
1037
1038 fhout.write('%s//%s< %s\n' % (
1039 non_comment,
1040 tt,
1041 comm.lines[0]
1042 ))
1043
1044 elif isinstance(comm, RemoveComment):
c0094f3c 1045 # End comment block and skip this line
1046 if in_comment:
b9781053 1047 skip_empty = dump_comment_block(prev_comm, restore_lines)
c0094f3c 1048 in_comment = False
b9781053 1049 restore_lines = None
06ccae0f 1050
22679479 1051 elif restore_lines is None:
c0094f3c 1052
b9781053 1053 # Beginning of a new comment block of type Comment or PrependComment
06ccae0f 1054 in_comment = True
1055
b9781053 1056 if isinstance(comm, PrependComment):
1057 # Prepare array of lines to dump right after the comment
1058 restore_lines = [ line.rstrip('\n') ]
b53c40e9 1059 logging.debug('Commencing lines to restore: {%s}' % Colt(restore_lines[0]).cyan())
b9781053 1060 else:
1061 # Extract the non-comment part and print it if it exists
1062 non_comment = line[ 0:comm.first_col-1 ].rstrip()
1063 if non_comment != '':
1064 fhout.write( non_comment + '\n' )
1065
1066 elif isinstance(comm, Comment):
1067
1068 if restore_lines is not None:
1069 # From the 2nd line on of comment to prepend
1070 restore_lines.append( line.rstrip('\n') )
b53c40e9 1071 logging.debug('Appending lines to restore. All lines: {%s}' % Colt(restore_lines).cyan())
b9781053 1072
1073 else:
a2245741 1074 assert False, 'Unhandled parser state: line=%d comm={%s} prev_comm={%s}' % \
b9781053 1075 (line_num, comm, prev_comm)
06ccae0f 1076
1077 else:
1078
b9781053 1079 # Not a comment line
1080
06ccae0f 1081 if in_comment:
1082
1083 # We have just exited a comment block of type Comment
b9781053 1084 skip_empty = dump_comment_block(prev_comm, restore_lines)
06ccae0f 1085 in_comment = False
b9781053 1086 restore_lines = None
06ccae0f 1087
b9781053 1088 # Dump the non-comment line
06ccae0f 1089 line_out = line.rstrip('\n')
1090 if skip_empty:
1091 skip_empty = False
1092 if line_out.strip() != '':
1093 fhout.write( line_out + '\n' )
1094 else:
1095 fhout.write( line_out + '\n' )
1096
7ff022be 1097 # Is there some comment left here?
1098 if restore_lines is not None:
1099 dump_comment_block(comm, restore_lines)
1100
1101 # Is there some other comment beyond the last line?
1102 for c in comments:
1103 if c.has_comment(line_num+1):
1104 dump_comment_block(c, None)
1105 break
1106
f329fa92 1107
1108## The main function.
1109#
06ccae0f 1110# Return value is the executable's return value.
f329fa92 1111def main(argv):
1112
06ccae0f 1113 # Setup logging on stderr
1114 log_level = logging.INFO
1115 logging.basicConfig(
1116 level=log_level,
1117 format='%(levelname)-8s %(funcName)-20s %(message)s',
1118 stream=sys.stderr
1119 )
f329fa92 1120
06ccae0f 1121 # Parse command-line options
1122 output_on_stdout = False
7afec95e 1123 include_flags = []
06ccae0f 1124 try:
7afec95e 1125 opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
06ccae0f 1126 for o, a in opts:
1127 if o == '--debug':
1128 log_level = getattr( logging, a.upper(), None )
1129 if not isinstance(log_level, int):
1130 raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
1131 elif o == '-d':
1132 log_level = logging.DEBUG
1133 elif o == '-o' or o == '--stdout':
06ccae0f 1134 output_on_stdout = True
7afec95e 1135 elif o == '-I':
1136 if os.path.isdir(a):
1137 include_flags.extend( [ '-I', a ] )
1138 else:
1139 logging.fatal('Include directory not found: %s' % Colt(a).magenta())
1140 return 2
06ccae0f 1141 else:
1142 assert False, 'Unhandled argument'
1143 except getopt.GetoptError as e:
1144 logging.fatal('Invalid arguments: %s' % e)
1145 return 1
f329fa92 1146
06ccae0f 1147 logging.getLogger('').setLevel(log_level)
f329fa92 1148
06ccae0f 1149 # Attempt to load libclang from a list of known locations
1150 libclang_locations = [
1151 '/usr/lib/llvm-3.5/lib/libclang.so.1',
1152 '/usr/lib/libclang.so',
1153 '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
1154 ]
1155 libclang_found = False
f329fa92 1156
06ccae0f 1157 for lib in libclang_locations:
1158 if os.path.isfile(lib):
1159 clang.cindex.Config.set_library_file(lib)
1160 libclang_found = True
1161 break
f329fa92 1162
06ccae0f 1163 if not libclang_found:
1164 logging.fatal('Cannot find libclang')
1165 return 1
1166
1167 # Loop over all files
1168 for fn in args:
1169
1170 logging.info('Input file: %s' % Colt(fn).magenta())
1171 index = clang.cindex.Index.create()
7afec95e 1172 clang_args = [ '-x', 'c++' ]
1173 clang_args.extend( include_flags )
1174 translation_unit = index.parse(fn, args=clang_args)
06ccae0f 1175
1176 comments = []
1177 traverse_ast( translation_unit.cursor, fn, comments )
1178 for c in comments:
1179
1180 logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
f329fa92 1181
06ccae0f 1182 if isinstance(c, MemberComment):
1183
21689d7a 1184 if c.is_transient():
1185 flag_text = Colt('transient ').yellow()
1186 elif c.is_dontsplit():
1187 flag_text = Colt('dontsplit ').yellow()
1188 elif c.is_ptr():
1189 flag_text = Colt('ptr ').yellow()
06ccae0f 1190 else:
21689d7a 1191 flag_text = ''
06ccae0f 1192
1193 if c.array_size is not None:
1194 array_text = Colt('arraysize=%s ' % c.array_size).yellow()
1195 else:
1196 array_text = ''
1197
1198 logging.debug(
1199 "%s %s%s{%s}" % ( \
1200 Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
21689d7a 1201 flag_text,
06ccae0f 1202 array_text,
1203 Colt(c.lines[0]).cyan()
1204 ))
1205
1206 elif isinstance(c, RemoveComment):
1207
1208 logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
1209
1210 else:
1211 for l in c.lines:
1212 logging.debug(
1213 Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
1214 "{%s}" % Colt(l).cyan()
1215 )
f329fa92 1216
1217 try:
06ccae0f 1218
1219 if output_on_stdout:
1220 with open(fn, 'r') as fhin:
1221 rewrite_comments( fhin, sys.stdout, comments )
1222 else:
1223 fn_back = fn + '.thtml2doxy_backup'
1224 os.rename( fn, fn_back )
1225
1226 with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
1227 rewrite_comments( fhin, fhout, comments )
1228
1229 os.remove( fn_back )
1230 logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
1231 except (IOError,OSError) as e:
1232 logging.error('File operation failed: %s' % e)
f329fa92 1233
1234 return 0
1235
06ccae0f 1236
f329fa92 1237if __name__ == '__main__':
06ccae0f 1238 sys.exit( main( sys.argv[1:] ) )