3 ## @package thtml2doxy_clang
4 # Translates THtml C++ comments to Doxygen using libclang as parser.
6 # This code relies on Python bindings for libclang: libclang's interface is pretty unstable, and
7 # its Python bindings are unstable as well.
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.
12 # This code (expecially AST traversal) was inspired by:
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17 # (outdated, API calls described there do not work anymore, but useful to understand some basic
22 # `thtml2doxy_clang [--stdout|-o] [-d] [--debug=DEBUG_LEVEL] file1 [file2 [file3...]]`
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`,
31 # @author Dario Berzano, CERN
44 ## Brain-dead color output for terminal.
48 return self.color('\033[31m')
51 return self.color('\033[32m')
54 return self.color('\033[33m')
57 return self.color('\033[34m')
60 return self.color('\033[35m')
63 return self.color('\033[36m')
66 return c + self + '\033[m'
70 class Comment(object):
72 def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func, \
75 assert first_line > 0 and last_line >= first_line, 'Wrong line numbers'
77 self.first_line = first_line
78 self.first_col = first_col
79 self.last_line = last_line
80 self.last_col = last_col
83 self.append_empty = append_empty
85 def has_comment(self, line):
86 return line >= self.first_line and line <= self.last_line
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,
96 class PrependComment(Comment):
98 def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func, \
100 super(PrependComment, self).__init__( \
101 lines, first_line, first_col, last_line, last_col, indent, func, append_empty)
104 ## A data member comment.
107 def __init__(self, text, comment_flag, array_size, first_line, first_col, func):
108 assert first_line > 0, 'Wrong line number'
109 assert comment_flag is None or comment_flag == '!' or comment_flag in [ '!', '||', '->' ]
110 self.lines = [ text ]
111 self.comment_flag = comment_flag
112 self.array_size = array_size
113 self.first_line = first_line
114 self.first_col = first_col
117 def is_transient(self):
118 return self.comment_flag == '!'
120 def is_dontsplit(self):
121 return self.comment_flag == '||'
124 return self.comment_flag == '->'
126 def has_comment(self, line):
127 return line == self.first_line
131 if self.is_transient():
133 elif self.is_dontsplit():
140 if self.array_size is not None:
141 ars = '[%s] ' % self.array_size
145 return "<MemberComment for %s: [%d,%d] %s%s%s>" % (self.func, self.first_line, self.first_col, tt, ars, self.lines[0])
148 ## A dummy comment that removes comment lines.
149 class RemoveComment(Comment):
151 def __init__(self, first_line, last_line):
152 assert first_line > 0 and last_line >= first_line, 'Wrong line numbers'
153 self.first_line = first_line
154 self.last_line = last_line
155 self.func = '<remove>'
158 return "<RemoveComment: [%d,%d]>" % (self.first_line, self.last_line)
161 ## Parses method comments.
163 # @param cursor Current libclang parser cursor
164 # @param comments Array of comments: new ones will be appended there
165 def comment_method(cursor, comments):
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
170 in_compound_stmt = False
171 expect_comment = False
175 comment_function = cursor.spelling or cursor.displayname
176 comment_line_start = -1
177 comment_line_end = -1
178 comment_col_start = -1
182 for token in cursor.get_tokens():
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
191 in_compound_stmt = False
194 # tkind = str(token.kind)[str(token.kind).index('.')+1:]
195 # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:]
201 extent = token.extent
202 line_start = extent.start.line
203 line_end = extent.end.line
205 if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{':
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
212 if comment_indent == -1 or (extent.start.column-1) < comment_indent:
213 comment_indent = extent.start.column-1
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') )
220 # multiline comments are parsed in one go, therefore don't expect subsequent comments
221 if line_end - line_start > 0:
223 expect_comment = False
227 expect_comment = False
231 if comment_line_start > 0:
233 comment = refactor_comment( comment, infilename=str(cursor.location.file) )
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) )
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))
244 logging.warning('No comment found for function %s' % Colt(comment_function).magenta())
247 comment_line_start = -1
248 comment_line_end = -1
249 comment_col_start = -1
257 ## Parses comments to class data members.
259 # @param cursor Current libclang parser cursor
260 # @param comments Array of comments: new ones will be appended there
261 def comment_datamember(cursor, comments):
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.
268 line_num = cursor.location.line
273 # Huge overkill: current line saved in "raw", previous in "prev"
274 with open(str(cursor.location.file)) as fp:
277 cur_line = cur_line + 1
278 if cur_line == line_num:
283 assert found, 'A line that should exist was not found in file' % cursor.location.file
285 recomm = r'(//(!|\|\||->)|///?)(\[(.+?)\])?<?\s*(.*?)\s*$'
286 recomm_prevline = r'^\s*///\s*(.*?)\s*$'
288 mcomm = re.search(recomm, raw)
290 # If it does not match, we do not have a comment
291 member_name = cursor.spelling;
292 comment_flag = mcomm.group(2)
293 array_size = mcomm.group(4)
294 text = mcomm.group(5)
296 col_num = mcomm.start()+1;
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
301 mcomm_prevline = re.search(recomm_prevline, prev)
303 text = mcomm_prevline.group(1)
304 comments.append(RemoveComment(line_num-1, line_num-1))
306 logging.debug('Comment found for member %s' % Colt(member_name).magenta())
308 comments.append( MemberComment(
317 ## Parses class description (beginning of file).
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
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.
329 # The reason why they cannot start with a multi-line sequence is that those are commonly used to
330 # display a copyright notice.
332 # @param filename Name of the current file
333 # @param comments Array of comments: new ones will be appended there
334 # @param look_no_further_than_line Stop before reaching this line when looking for class comment
335 def comment_classdesc(filename, comments, look_no_further_than_line):
337 # Single-line comment
338 recomm = r'^\s*///?(\s*(.*?))\s*/*\s*$'
340 # Multi-line comment (only either /* or */ on a single line)
341 remlcomm_in = r'^\s*/\*\s*$'
342 remlcomm_out = r'^\s*\*/\s*$'
345 reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)'
346 class_name_doxy = None
348 reauthor = r'(?i)^\s*\\?(authors?|origin):?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$'
349 redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$'
350 rebrief = r'(?i)^\s*\\brief\s*(.*)\s*$'
354 brief_len_threshold = 80
362 last_comm_line_num = 0
364 is_macro = filename.endswith('.C')
366 with open(filename, 'r') as fp:
370 line_num = line_num + 1
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)
377 if in_mlcomm == False and raw.strip() == '' and start_line > 0:
381 stripped = strip_html(raw)
386 mcomm = re.search(recomm, stripped)
388 if last_comm_line_num == 0 or last_comm_line_num == line_num-1:
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)
401 mmlcomm = re.search(remlcomm_out, stripped)
406 this_comment = stripped
408 if this_comment is not None:
412 # First line. Check that we do not overlap with other comments
413 comment_overlaps = False
415 if c.has_comment(line_num):
416 comment_overlaps = True
420 # No need to look for other comments
423 start_line = line_num
428 mclass_doxy = re.search(reclass_doxy, this_comment)
430 class_name_doxy = mclass_doxy.group(2)
433 mauthor = re.search(reauthor, this_comment)
435 author = mauthor.group(2)
437 # Date specified in the standalone \date field has priority
438 date = mauthor.group(4)
441 mdate = re.search(redate, this_comment)
443 date = mdate.group(1)
446 mbrief = re.search(rebrief, this_comment)
448 brief = mbrief.group(1)
452 comment_lines.append( this_comment )
458 # This line had a valid comment
459 last_comm_line_num = line_num
461 if class_name_doxy is None:
463 # No \class specified: guess it from file name
464 reclass = r'^(.*/)?(.*?)(\..*)?$'
465 mclass = re.search( reclass, filename )
467 class_name_doxy = mclass.group(2)
469 assert False, 'Regexp unable to extract classname from file'
473 file_class_line = '\\file ' + class_name_doxy + '.C'
475 file_class_line = '\\class ' + class_name_doxy
479 prepend_to_comment = []
481 # Prepend \class or \file specifier, then the \brief, then an empty line
482 prepend_to_comment.append( file_class_line )
484 if brief is not None:
485 prepend_to_comment.append( '\\brief ' + brief )
486 prepend_to_comment.append( '' )
488 comment_lines = prepend_to_comment + comment_lines # join lists
490 # Append author and date if they exist
491 if author is not None:
492 comment_lines.append( '\\author ' + author )
495 comment_lines.append( '\\date ' + date )
497 # We should erase the "dumb" comments, such as "<class_name> class"
499 regac = r'\s*%s\s+class\.?\s*' % class_name_doxy
501 for comm in comment_lines:
502 mgac = re.search(regac, comm)
505 comm_idx = comm_idx + 1
507 logging.debug('Removing dumb comment line: {%s}' % Colt(comment_lines[comm_idx]).magenta())
508 del comment_lines[comm_idx]
510 comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
512 # Now we look for a possible \brief
515 for comm in comment_lines:
516 if comm.startswith('\\class') or comm.startswith('\\file') or comm == '':
519 if len(comm) <= brief_len_threshold:
522 comm_idx = comm_idx + 1
523 if brief is not None:
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)
528 logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
529 comments.append(Comment(
531 start_line, 1, end_line, 1,
537 logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
538 Colt(class_name_doxy).magenta())
540 comments.append(PrependComment(
543 0, class_name_doxy, append_empty=True
547 ## Looks for a special ROOT ClassImp() entry.
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
553 # @param filename Name of the current file
554 # @param comments Array of comments: new ones will be appended there
555 def comment_classimp(filename, comments):
557 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
560 reclassimp = r'^(\s*)Class(Imp|Def)\((.*?)\).*$'
562 in_classimp_cond = False
563 restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
564 reendcond = r'^\s*///\s*\\endcond\s*$'
566 with open(filename, 'r') as fp:
570 # Reset to nothing found
574 classimp_class = None
575 classimp_indent = None
577 line_num = line_num + 1
579 mclassimp = re.search(reclassimp, line)
583 classimp_indent = len( mclassimp.group(1) )
585 line_classimp = line_num
586 classimp_class = mclassimp.group(3)
587 imp_or_def = mclassimp.group(2)
589 'Comment found for ' +
590 Colt( 'Class%s(' % imp_or_def ).magenta() +
591 Colt( classimp_class ).cyan() +
592 Colt( ')' ).magenta() )
596 mstartcond = re.search(restartcond, line)
598 logging.debug('Found Doxygen opening condition for ClassImp in {%s}' % line)
599 in_classimp_cond = True
600 line_startcond = line_num
602 elif in_classimp_cond:
604 mendcond = re.search(reendcond, line)
606 logging.debug('Found Doxygen closing condition for ClassImp')
607 in_classimp_cond = False
608 line_endcond = line_num
610 # Did we find something?
611 if line_classimp != -1:
613 if line_startcond != -1:
614 comments.append(Comment(
616 line_startcond, 1, line_startcond, 1,
617 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
621 comments.append(PrependComment(
623 line_classimp, 1, line_classimp, 1,
624 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
627 if line_endcond != -1:
628 comments.append(Comment(
630 line_endcond, 1, line_endcond, 1,
631 classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
635 comments.append(PrependComment(
637 line_classimp+1, 1, line_classimp+1, 1,
638 classimp_indent, 'ClassImp/Def(%s)' % classimp_class
642 ## Traverse the AST recursively starting from the current cursor.
644 # @param cursor A Clang parser cursor
645 # @param filename Name of the current file
646 # @param comments Array of comments: new ones will be appended there
647 # @param recursion Current recursion depth
648 # @param in_func True if we are inside a function or method
649 # @param classdesc_line_limit Do not look for comments after this line
651 # @return A tuple containing the classdesc_line_limit as first item
652 def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
654 # libclang traverses included files as well: we do not want this behavior
655 if cursor.location.file is not None and str(cursor.location.file) != filename:
656 logging.debug("Skipping processing of included %s" % cursor.location.file)
659 text = cursor.spelling or cursor.displayname
660 kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
662 is_macro = filename.endswith('.C')
665 for i in range(0, recursion):
666 indent = indent + ' '
668 if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
669 clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
671 if classdesc_line_limit is None:
672 classdesc_line_limit = cursor.location.line
674 # cursor ran into a C++ method
675 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
676 comment_method(cursor, comments)
679 elif not is_macro and not in_func and \
680 cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
682 if classdesc_line_limit is None:
683 classdesc_line_limit = cursor.location.line
685 # cursor ran into a data member declaration
686 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
687 comment_datamember(cursor, comments)
691 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
693 for child_cursor in cursor.get_children():
694 classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
697 comment_classimp(filename, comments)
698 comment_classdesc(filename, comments, classdesc_line_limit)
700 return classdesc_line_limit
703 ## Strip some HTML tags from the given string. Returns clean string.
705 # @param s Input string
707 rehtml = r'(?i)</?(P|BR)/?>'
708 return re.sub(rehtml, '', s)
711 ## Remove garbage from comments and convert special tags from THtml to Doxygen.
713 # @param comment An array containing the lines of the original comment
714 def refactor_comment(comment, do_strip_html=True, infilename=None):
716 recomm = r'^(/{2,}|/\*)? ?(\s*)(.*?)\s*((/{2,})?\s*|\*/)$'
717 regarbage = r'^(?i)\s*([\s*=_#-]+|(Begin|End)_Html)\s*$'
719 # Support for LaTeX blocks spanning on multiple lines
720 relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
724 # Support for LaTeX blocks on a single line
725 reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
727 # Match <pre> (to turn it into the ~~~ Markdown syntax)
728 reblock = r'(?i)^(\s*)</?PRE>\s*$'
730 # Macro blocks for pictures generation
733 remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
735 # Minimum indent level: scale back everything
736 lowest_indent_level = None
738 # Indentation threshold: if too much indented, don't indent at all
739 indent_level_threshold = 7
743 wait_first_non_blank = True
744 for line_comment in comment:
746 # Strip some HTML tags
748 line_comment = strip_html(line_comment)
750 mcomm = re.search( recomm, line_comment )
752 new_line_comment = mcomm.group(2) + mcomm.group(3) # indent + comm
754 # Check if we are in a macro block
755 mmacro = re.search(remacro, new_line_comment)
761 outimg = write_macro(infilename, current_macro) + '.png'
765 new_comment.append( '![Picture from ROOT macro](%s)' % (os.path.basename(outimg)) )
767 logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
774 current_macro.append( new_line_comment )
777 mgarbage = re.search( regarbage, new_line_comment )
779 if mgarbage is None and not mcomm.group(3).startswith('\\') and mcomm.group(3) != '':
780 # not a special command line: count indent
781 indent_level = len( mcomm.group(2) )
782 if lowest_indent_level is None or indent_level < lowest_indent_level:
783 lowest_indent_level = indent_level
785 # if indentation level is too much, consider it zero
786 if indent_level > indent_level_threshold:
787 new_line_comment = mcomm.group(3) # remove ALL indentation
789 if new_line_comment == '' or mgarbage is not None:
792 if insert_blank and not wait_first_non_blank:
793 new_comment.append('')
795 wait_first_non_blank = False
797 # Postprocessing: LaTeX formulas in ROOT format
798 # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
799 # There can be several ROOT LaTeX forumlas per line
801 minline_latex = re.search( reinline_latex, new_line_comment )
803 new_line_comment = '%s\\f$%s\\f$%s' % \
804 ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
805 minline_latex.group(3) )
809 # ROOT LaTeX: do we have a Begin/End_LaTeX block?
810 # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
811 # block here left without a corresponding ending block
812 mlatex = re.search( relatex, new_line_comment )
815 # before and after parts have been already stripped
816 l_before = mlatex.group(2)
817 l_after = mlatex.group(4)
818 is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
827 # Begin of LaTeX part
830 if l_before == '' and l_after == '':
832 # Opening tag alone: mark the beginning of a block: \f[ ... \f]
834 new_comment.append( '\\f[' )
837 # Mark the beginning of inline: \f$ ... \f$
840 '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
850 # Closing a LaTeX block
852 new_comment.append( l_before.replace('#', '\\') )
853 new_comment.append( '\\f]' )
855 new_comment.append( l_after )
859 # Closing a LaTeX inline
861 '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
864 # Prevent appending lines (we have already done that)
865 new_line_comment = None
867 # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
868 # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
869 if new_line_comment is not None and not in_latex:
871 mblock = re.search( reblock, new_line_comment )
873 new_comment.append( mblock.group(1)+'~~~' )
874 new_line_comment = None
876 if new_line_comment is not None:
878 new_line_comment = new_line_comment.replace('#', '\\')
879 new_comment.append( new_line_comment )
882 assert False, 'Comment regexp does not match'
884 # Fixing indentation level
885 if lowest_indent_level is not None:
886 logging.debug('Lowest indentation level found: %d' % lowest_indent_level)
888 new_comment_indent = []
889 reblankstart = r'^\s+'
890 for line in new_comment:
891 if re.search(reblankstart, line):
892 new_comment_indent.append( line[lowest_indent_level:] )
894 new_comment_indent.append( line )
896 new_comment = new_comment_indent
899 logging.debug('No indentation scaling applied')
904 ## Dumps an image-generating macro to the correct place. Returns a string with the image path,
905 # without the extension.
907 # @param infilename File name of the source file
908 # @param macro_lines Array of macro lines
909 def write_macro(infilename, macro_lines):
912 digh = hashlib.sha1()
913 for l in macro_lines:
916 short_digest = digh.hexdigest()[0:7]
918 infiledir = os.path.dirname(infilename)
921 outdir = '%s/imgdoc' % infiledir
922 outprefix = '%s/%s_%s' % (
924 os.path.basename(infilename).replace('.', '_'),
927 outmacro = '%s.C' % outprefix
930 if not os.path.isdir(outdir):
931 # do not catch: let everything die on error
932 logging.debug('Creating directory %s' % Colt(outdir).magenta())
935 # Create file (do not catch errors either)
936 with open(outmacro, 'w') as omfp:
937 logging.debug('Writing macro %s' % Colt(outmacro).magenta())
938 for l in macro_lines:
945 ## Rewrites all comments from the given file handler.
947 # @param fhin The file handler to read from
948 # @param fhout The file handler to write to
949 # @param comments Array of comments
950 def rewrite_comments(fhin, fhout, comments):
961 def dump_comment_block(cmt, restore=None):
963 ask_skip_empty = False
965 for i in range(0, cmt.indent):
966 text_indent = text_indent + ' '
969 fhout.write('%s///' % text_indent )
976 # Empty new line at the end of the comment
979 ask_skip_empty = True
981 # Restore lines if possible
987 # Tell the caller whether it should skip the next empty line found
988 return ask_skip_empty
993 line_num = line_num + 1
995 # Find current comment
1000 if c.has_comment(line_num):
1004 if len(comm_list) > 1:
1008 if len(comm_list) == 2:
1010 if isinstance(c1, Comment) and isinstance(c2, Comment):
1011 c1.lines = c1.lines + c2.lines # list merge
1013 logging.debug('Two adjacent comments merged. Result: {%s}' % Colt(comm).cyan())
1020 logging.warning('Too many unmergeable comments on the same line (%d), picking the last one' % len(comm_list))
1022 logging.warning('>> %s' % c)
1023 comm = c # considering the last one
1027 # First thing to check: are we in the same comment as before?
1028 if comm is not prev_comm and \
1029 isinstance(comm, Comment) and \
1030 isinstance(prev_comm, Comment) and \
1031 not isinstance(prev_comm, RemoveComment):
1033 # We are NOT in the same comment as before, and this comment is dumpable
1035 skip_empty = dump_comment_block(prev_comm, restore_lines)
1037 restore_lines = None
1038 prev_comm = None # we have just dumped it: pretend it never existed in this loop
1041 # Check type of comment and react accordingly
1044 if isinstance(comm, MemberComment):
1048 skip_empty = dump_comment_block(prev_comm, restore_lines)
1050 restore_lines = None
1052 non_comment = line[ 0:comm.first_col-1 ]
1054 if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
1056 # This is a special case: comment will be split in two lines: one before the comment for
1057 # Doxygen as "member description", and the other right after the comment on the same line
1058 # to be parsed by ROOT's C++ parser
1060 # Keep indent on the generated line of comment before member definition
1061 mindent = re.search(rindent, line)
1063 # Get correct comment flag, if any
1064 if comm.comment_flag is not None:
1065 cflag = comm.comment_flag
1069 # Get correct array size, if any
1070 if comm.array_size is not None:
1071 asize = '[%s]' % comm.array_size
1075 # Write on two lines
1076 fhout.write('%s/// %s\n%s//%s%s\n' % (
1086 # Single-line comments with the "transient" flag can be kept on one line in a way that
1087 # they are correctly interpreted by both ROOT and Doxygen
1089 if comm.is_transient():
1094 fhout.write('%s//%s< %s\n' % (
1100 elif isinstance(comm, RemoveComment):
1101 # End comment block and skip this line
1103 skip_empty = dump_comment_block(prev_comm, restore_lines)
1105 restore_lines = None
1107 elif restore_lines is None:
1109 # Beginning of a new comment block of type Comment or PrependComment
1112 if isinstance(comm, PrependComment):
1113 # Prepare array of lines to dump right after the comment
1114 restore_lines = [ line.rstrip('\n') ]
1115 logging.debug('Commencing lines to restore: {%s}' % Colt(restore_lines[0]).cyan())
1117 # Extract the non-comment part and print it if it exists. If this is the first line of a
1118 # comment, it might happen something like `valid_code; // this is a comment`.
1119 if comm.first_line == line_num:
1120 non_comment = line[ 0:comm.first_col-1 ].rstrip()
1121 if non_comment != '':
1122 fhout.write( non_comment + '\n' )
1124 elif isinstance(comm, Comment):
1126 if restore_lines is not None:
1127 # From the 2nd line on of comment to prepend
1128 restore_lines.append( line.rstrip('\n') )
1129 logging.debug('Appending lines to restore. All lines: {%s}' % Colt(restore_lines).cyan())
1132 assert False, 'Unhandled parser state: line=%d comm={%s} prev_comm={%s}' % \
1133 (line_num, comm, prev_comm)
1137 # Not a comment line
1141 # We have just exited a comment block of type Comment
1142 skip_empty = dump_comment_block(prev_comm, restore_lines)
1144 restore_lines = None
1146 # Dump the non-comment line
1147 line_out = line.rstrip('\n')
1150 if line_out.strip() != '':
1151 fhout.write( line_out + '\n' )
1153 fhout.write( line_out + '\n' )
1155 # Is there some comment left here?
1156 if restore_lines is not None:
1157 dump_comment_block(comm, restore_lines)
1159 # Is there some other comment beyond the last line?
1161 if c.has_comment(line_num+1):
1162 dump_comment_block(c, None)
1166 ## The main function.
1168 # Return value is the executable's return value.
1171 # Setup logging on stderr
1172 log_level = logging.INFO
1173 logging.basicConfig(
1175 format='%(levelname)-8s %(funcName)-20s %(message)s',
1179 # Parse command-line options
1180 output_on_stdout = False
1183 opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
1186 log_level = getattr( logging, a.upper(), None )
1187 if not isinstance(log_level, int):
1188 raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
1190 log_level = logging.DEBUG
1191 elif o == '-o' or o == '--stdout':
1192 output_on_stdout = True
1194 if os.path.isdir(a):
1195 include_flags.extend( [ '-I', a ] )
1197 logging.fatal('Include directory not found: %s' % Colt(a).magenta())
1200 assert False, 'Unhandled argument'
1201 except getopt.GetoptError as e:
1202 logging.fatal('Invalid arguments: %s' % e)
1205 logging.getLogger('').setLevel(log_level)
1207 # Attempt to load libclang from a list of known locations
1208 libclang_locations = [
1209 '/usr/lib/llvm-3.5/lib/libclang.so.1',
1210 '/usr/lib/libclang.so',
1211 '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
1213 libclang_found = False
1215 for lib in libclang_locations:
1216 if os.path.isfile(lib):
1217 clang.cindex.Config.set_library_file(lib)
1218 libclang_found = True
1221 if not libclang_found:
1222 logging.fatal('Cannot find libclang')
1225 # Loop over all files
1228 logging.info('Input file: %s' % Colt(fn).magenta())
1229 index = clang.cindex.Index.create()
1230 clang_args = [ '-x', 'c++' ]
1231 clang_args.extend( include_flags )
1232 translation_unit = index.parse(fn, args=clang_args)
1235 traverse_ast( translation_unit.cursor, fn, comments )
1238 logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
1240 if isinstance(c, MemberComment):
1242 if c.is_transient():
1243 flag_text = Colt('transient ').yellow()
1244 elif c.is_dontsplit():
1245 flag_text = Colt('dontsplit ').yellow()
1247 flag_text = Colt('ptr ').yellow()
1251 if c.array_size is not None:
1252 array_text = Colt('arraysize=%s ' % c.array_size).yellow()
1258 Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
1261 Colt(c.lines[0]).cyan()
1264 elif isinstance(c, RemoveComment):
1266 logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
1271 Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
1272 "{%s}" % Colt(l).cyan()
1277 if output_on_stdout:
1278 with open(fn, 'r') as fhin:
1279 rewrite_comments( fhin, sys.stdout, comments )
1281 fn_back = fn + '.thtml2doxy_backup'
1282 os.rename( fn, fn_back )
1284 with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
1285 rewrite_comments( fhin, fhout, comments )
1287 os.remove( fn_back )
1288 logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
1289 except (IOError,OSError) as e:
1290 logging.error('File operation failed: %s' % e)
1295 if __name__ == '__main__':
1296 sys.exit( main( sys.argv[1:] ) )