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17 # (outdated, API calls described there do not work anymore, but useful to understand some basic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'(//(!|\|\||->)|///?)(\[([0-9,]+)\])?<?\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 (/* ... */) are not considered as they are commonly used to display
327 # @param filename Name of the current file
328 # @param comments Array of comments: new ones will be appended there
329 # @param look_no_further_than_line Stop before reaching this line when looking for class comment
330 def comment_classdesc(filename, comments, look_no_further_than_line):
332 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
334 reclass_doxy = r'(?i)^\s*\\(class|file):?\s*([^.]*)'
335 class_name_doxy = None
337 reauthor = r'(?i)^\s*\\?authors?:?\s*(.*?)\s*(,?\s*([0-9./-]+))?\s*$'
338 redate = r'(?i)^\s*\\?date:?\s*([0-9./-]+)\s*$'
349 is_macro = filename.endswith('.C')
351 with open(filename, 'r') as fp:
355 line_num = line_num + 1
357 if look_no_further_than_line is not None and line_num == look_no_further_than_line:
358 logging.debug('Stopping at line %d while looking for class/file description' % \
359 look_no_further_than_line)
362 if raw.strip() == '' and start_line > 0:
366 stripped = strip_html(raw)
367 mcomm = re.search(recomm, stripped)
372 # First line. Check that we do not overlap with other comments
373 comment_overlaps = False
375 if c.has_comment(line_num):
376 comment_overlaps = True
380 # No need to look for other comments
383 start_line = line_num
388 mclass_doxy = re.search(reclass_doxy, mcomm.group(1))
390 class_name_doxy = mclass_doxy.group(2)
393 mauthor = re.search(reauthor, mcomm.group(1))
395 author = mauthor.group(1)
397 # Date specified in the standalone \date field has priority
398 date = mauthor.group(3)
401 mdate = re.search(redate, mcomm.group(1))
403 date = mdate.group(1)
407 comment_lines.append( mcomm.group(1) )
413 if class_name_doxy is None:
415 # No \class specified: guess it from file name
416 reclass = r'^(.*/)?(.*?)(\..*)?$'
417 mclass = re.search( reclass, filename )
419 class_name_doxy = mclass.group(2)
421 assert False, 'Regexp unable to extract classname from file'
425 file_class_line = '\\file ' + class_name_doxy + '.C'
427 file_class_line = '\\class ' + class_name_doxy
431 # Prepend \class or \file specifier (and an empty line)
432 comment_lines[:0] = [ file_class_line ]
433 comment_lines.append('')
435 # Append author and date if they exist
436 if author is not None:
437 comment_lines.append( '\\author ' + author )
440 comment_lines.append( '\\date ' + date )
442 comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
443 logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
444 comments.append(Comment(
446 start_line, 1, end_line, 1,
452 logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
453 Colt(class_name_doxy).magenta())
455 comments.append(PrependComment(
458 0, class_name_doxy, append_empty=True
462 ## Looks for a special ROOT ClassImp() entry.
464 # Doxygen might get confused by `ClassImp()` entries as they are macros normally written without
465 # the ending `;`: this function wraps the definition inside a condition in order to make Doxygen
468 # @param filename Name of the current file
469 # @param comments Array of comments: new ones will be appended there
470 def comment_classimp(filename, comments):
472 recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
475 reclassimp = r'^(\s*)ClassImp\((.*?)\)\s*;?\s*$'
477 in_classimp_cond = False
478 restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
479 reendcond = r'^\s*///\s*\\endcond\s*$'
481 with open(filename, 'r') as fp:
486 classimp_class = None
487 classimp_indent = None
491 line_num = line_num + 1
493 mclassimp = re.search(reclassimp, line)
497 classimp_indent = len( mclassimp.group(1) )
499 line_classimp = line_num
500 classimp_class = mclassimp.group(2)
502 'Comment found for ' +
503 Colt( 'ClassImp(' ).magenta() +
504 Colt( classimp_class ).cyan() +
505 Colt( ')' ).magenta() )
509 mstartcond = re.search(restartcond, line)
511 logging.debug('Found Doxygen opening condition for ClassImp in {%s}' % line)
512 in_classimp_cond = True
513 line_startcond = line_num
515 elif in_classimp_cond:
517 mendcond = re.search(reendcond, line)
519 logging.debug('Found Doxygen closing condition for ClassImp')
520 in_classimp_cond = False
521 line_endcond = line_num
523 # Did we find something?
524 if line_classimp != -1:
526 if line_startcond != -1:
527 comments.append(Comment(
529 line_startcond, 1, line_startcond, 1,
530 classimp_indent, 'ClassImp(%s)' % classimp_class,
534 comments.append(PrependComment(
536 line_classimp, 1, line_classimp, 1,
537 classimp_indent, 'ClassImp(%s)' % classimp_class
540 if line_endcond != -1:
541 comments.append(Comment(
543 line_endcond, 1, line_endcond, 1,
544 classimp_indent, 'ClassImp(%s)' % classimp_class,
548 comments.append(PrependComment(
550 line_classimp+1, 1, line_classimp+1, 1,
551 classimp_indent, 'ClassImp(%s)' % classimp_class
555 ## Traverse the AST recursively starting from the current cursor.
557 # @param cursor A Clang parser cursor
558 # @param filename Name of the current file
559 # @param comments Array of comments: new ones will be appended there
560 # @param recursion Current recursion depth
561 # @param in_func True if we are inside a function or method
562 # @param classdesc_line_limit Do not look for comments after this line
564 # @return A tuple containing the classdesc_line_limit as first item
565 def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
567 # libclang traverses included files as well: we do not want this behavior
568 if cursor.location.file is not None and str(cursor.location.file) != filename:
569 logging.debug("Skipping processing of included %s" % cursor.location.file)
572 text = cursor.spelling or cursor.displayname
573 kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
575 is_macro = filename.endswith('.C')
578 for i in range(0, recursion):
579 indent = indent + ' '
581 if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
582 clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
584 if classdesc_line_limit is None:
585 classdesc_line_limit = cursor.location.line
587 # cursor ran into a C++ method
588 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
589 comment_method(cursor, comments)
592 elif not is_macro and not in_func and \
593 cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
595 if classdesc_line_limit is None:
596 classdesc_line_limit = cursor.location.line
598 # cursor ran into a data member declaration
599 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
600 comment_datamember(cursor, comments)
604 logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
606 for child_cursor in cursor.get_children():
607 classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
610 comment_classimp(filename, comments)
611 comment_classdesc(filename, comments, classdesc_line_limit)
613 return classdesc_line_limit
616 ## Strip some HTML tags from the given string. Returns clean string.
618 # @param s Input string
620 rehtml = r'(?i)</?(P|BR)/?>'
621 return re.sub(rehtml, '', s)
624 ## Remove garbage from comments and convert special tags from THtml to Doxygen.
626 # @param comment An array containing the lines of the original comment
627 def refactor_comment(comment, do_strip_html=True, infilename=None):
629 recomm = r'^(/{2,}|/\*)? ?(\s*.*?)\s*((/{2,})?\s*|\*/)$'
630 regarbage = r'^(?i)\s*([\s*=-_#]+|(Begin|End)_Html)\s*$'
632 # Support for LaTeX blocks spanning on multiple lines
633 relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
637 # Support for LaTeX blocks on a single line
638 reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
640 # Match <pre> (to turn it into the ~~~ Markdown syntax)
641 reblock = r'(?i)^(\s*)</?PRE>\s*$'
643 # Macro blocks for pictures generation
646 remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
650 wait_first_non_blank = True
651 for line_comment in comment:
653 # Check if we are in a macro block
654 mmacro = re.search(remacro, line_comment)
660 outimg = write_macro(infilename, current_macro) + '.png'
664 new_comment.append( '![Picture from ROOT macro](%s)' % (outimg) )
666 logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
673 current_macro.append( line_comment )
676 # Strip some HTML tags
678 line_comment = strip_html(line_comment)
680 mcomm = re.search( recomm, line_comment )
682 new_line_comment = mcomm.group(2)
683 mgarbage = re.search( regarbage, new_line_comment )
685 if new_line_comment == '' or mgarbage is not None:
688 if insert_blank and not wait_first_non_blank:
689 new_comment.append('')
691 wait_first_non_blank = False
693 # Postprocessing: LaTeX formulas in ROOT format
694 # Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
695 # There can be several ROOT LaTeX forumlas per line
697 minline_latex = re.search( reinline_latex, new_line_comment )
699 new_line_comment = '%s\\f$%s\\f$%s' % \
700 ( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
701 minline_latex.group(3) )
705 # ROOT LaTeX: do we have a Begin/End_LaTeX block?
706 # Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
707 # block here left without a corresponding ending block
708 mlatex = re.search( relatex, new_line_comment )
711 # before and after parts have been already stripped
712 l_before = mlatex.group(2)
713 l_after = mlatex.group(4)
714 is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
723 # Begin of LaTeX part
726 if l_before == '' and l_after == '':
728 # Opening tag alone: mark the beginning of a block: \f[ ... \f]
730 new_comment.append( '\\f[' )
733 # Mark the beginning of inline: \f$ ... \f$
736 '%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
746 # Closing a LaTeX block
748 new_comment.append( l_before.replace('#', '\\') )
749 new_comment.append( '\\f]' )
751 new_comment.append( l_after )
755 # Closing a LaTeX inline
757 '%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
760 # Prevent appending lines (we have already done that)
761 new_line_comment = None
763 # If we are not in a LaTeX block, look for <pre> tags and transform them into Doxygen code
764 # blocks (using ~~~ ... ~~~). Only <pre> tags on a single line are supported
765 if new_line_comment is not None and not in_latex:
767 mblock = re.search( reblock, new_line_comment )
769 new_comment.append( mblock.group(1)+'~~~' )
770 new_line_comment = None
772 if new_line_comment is not None:
774 new_line_comment = new_line_comment.replace('#', '\\')
775 new_comment.append( new_line_comment )
778 assert False, 'Comment regexp does not match'
783 ## Dumps an image-generating macro to the correct place. Returns a string with the image path,
784 # without the extension.
786 # @param infilename File name of the source file
787 # @param macro_lines Array of macro lines
788 def write_macro(infilename, macro_lines):
791 digh = hashlib.sha1()
792 for l in macro_lines:
795 short_digest = digh.hexdigest()[0:7]
797 outdir = '%s/imgdoc' % os.path.dirname(infilename)
798 outprefix = '%s/%s_%s' % (
800 os.path.basename(infilename).replace('.', '_'),
803 outmacro = '%s.C' % outprefix
806 if not os.path.isdir(outdir):
807 # do not catch: let everything die on error
808 logging.debug('Creating directory %s' % Colt(outdir).magenta())
811 # Create file (do not catch errors either)
812 with open(outmacro, 'w') as omfp:
813 logging.debug('Writing macro %s' % Colt(outmacro).magenta())
814 for l in macro_lines:
821 ## Rewrites all comments from the given file handler.
823 # @param fhin The file handler to read from
824 # @param fhout The file handler to write to
825 # @param comments Array of comments
826 def rewrite_comments(fhin, fhout, comments):
837 def dump_comment_block(cmt, restore=None):
839 ask_skip_empty = False
841 for i in range(0, cmt.indent):
842 text_indent = text_indent + ' '
845 fhout.write('%s///' % text_indent )
852 # Empty new line at the end of the comment
855 ask_skip_empty = True
857 # Restore lines if possible
863 # Tell the caller whether it should skip the next empty line found
864 return ask_skip_empty
869 line_num = line_num + 1
871 # Find current comment
875 if c.has_comment(line_num):
880 # First thing to check: are we in the same comment as before?
881 if comm is not prev_comm and isinstance(comm, Comment) and isinstance(prev_comm, Comment) \
882 and not isinstance(prev_comm, RemoveComment):
884 skip_empty = dump_comment_block(prev_comm, restore_lines)
887 prev_comm = None # we have just dumped it: pretend it never existed in this loop
890 # Check type of comment and react accordingly
893 if isinstance(comm, MemberComment):
897 skip_empty = dump_comment_block(prev_comm, restore_lines)
901 non_comment = line[ 0:comm.first_col-1 ]
903 if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
905 # This is a special case: comment will be split in two lines: one before the comment for
906 # Doxygen as "member description", and the other right after the comment on the same line
907 # to be parsed by ROOT's C++ parser
909 # Keep indent on the generated line of comment before member definition
910 mindent = re.search(rindent, line)
912 # Get correct comment flag, if any
913 if comm.comment_flag is not None:
914 cflag = comm.comment_flag
918 # Get correct array size, if any
919 if comm.array_size is not None:
920 asize = '[%s]' % comm.array_size
925 fhout.write('%s/// %s\n%s//%s%s\n' % (
935 # Single-line comments with the "transient" flag can be kept on one line in a way that
936 # they are correctly interpreted by both ROOT and Doxygen
938 if comm.is_transient():
943 fhout.write('%s//%s< %s\n' % (
949 elif isinstance(comm, RemoveComment):
950 # End comment block and skip this line
952 skip_empty = dump_comment_block(prev_comm, restore_lines)
956 elif prev_comm is None:
958 # Beginning of a new comment block of type Comment or PrependComment
961 if isinstance(comm, PrependComment):
962 # Prepare array of lines to dump right after the comment
963 restore_lines = [ line.rstrip('\n') ]
965 # Extract the non-comment part and print it if it exists
966 non_comment = line[ 0:comm.first_col-1 ].rstrip()
967 if non_comment != '':
968 fhout.write( non_comment + '\n' )
970 elif isinstance(comm, Comment):
972 if restore_lines is not None:
973 # From the 2nd line on of comment to prepend
974 restore_lines.append( line.rstrip('\n') )
977 assert False, 'Unhandled parser state. line=%d comm=%s prev_comm=%s' % \
978 (line_num, comm, prev_comm)
986 # We have just exited a comment block of type Comment
987 skip_empty = dump_comment_block(prev_comm, restore_lines)
991 # Dump the non-comment line
992 line_out = line.rstrip('\n')
995 if line_out.strip() != '':
996 fhout.write( line_out + '\n' )
998 fhout.write( line_out + '\n' )
1001 ## The main function.
1003 # Return value is the executable's return value.
1006 # Setup logging on stderr
1007 log_level = logging.INFO
1008 logging.basicConfig(
1010 format='%(levelname)-8s %(funcName)-20s %(message)s',
1014 # Parse command-line options
1015 output_on_stdout = False
1018 opts, args = getopt.getopt( argv, 'odI:', [ 'debug=', 'stdout' ] )
1021 log_level = getattr( logging, a.upper(), None )
1022 if not isinstance(log_level, int):
1023 raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
1025 log_level = logging.DEBUG
1026 elif o == '-o' or o == '--stdout':
1027 output_on_stdout = True
1029 if os.path.isdir(a):
1030 include_flags.extend( [ '-I', a ] )
1032 logging.fatal('Include directory not found: %s' % Colt(a).magenta())
1035 assert False, 'Unhandled argument'
1036 except getopt.GetoptError as e:
1037 logging.fatal('Invalid arguments: %s' % e)
1040 logging.getLogger('').setLevel(log_level)
1042 # Attempt to load libclang from a list of known locations
1043 libclang_locations = [
1044 '/usr/lib/llvm-3.5/lib/libclang.so.1',
1045 '/usr/lib/libclang.so',
1046 '/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
1048 libclang_found = False
1050 for lib in libclang_locations:
1051 if os.path.isfile(lib):
1052 clang.cindex.Config.set_library_file(lib)
1053 libclang_found = True
1056 if not libclang_found:
1057 logging.fatal('Cannot find libclang')
1060 # Loop over all files
1063 logging.info('Input file: %s' % Colt(fn).magenta())
1064 index = clang.cindex.Index.create()
1065 clang_args = [ '-x', 'c++' ]
1066 clang_args.extend( include_flags )
1067 translation_unit = index.parse(fn, args=clang_args)
1070 traverse_ast( translation_unit.cursor, fn, comments )
1073 logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
1075 if isinstance(c, MemberComment):
1077 if c.is_transient():
1078 flag_text = Colt('transient ').yellow()
1079 elif c.is_dontsplit():
1080 flag_text = Colt('dontsplit ').yellow()
1082 flag_text = Colt('ptr ').yellow()
1086 if c.array_size is not None:
1087 array_text = Colt('arraysize=%s ' % c.array_size).yellow()
1093 Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
1096 Colt(c.lines[0]).cyan()
1099 elif isinstance(c, RemoveComment):
1101 logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
1106 Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
1107 "{%s}" % Colt(l).cyan()
1112 if output_on_stdout:
1113 with open(fn, 'r') as fhin:
1114 rewrite_comments( fhin, sys.stdout, comments )
1116 fn_back = fn + '.thtml2doxy_backup'
1117 os.rename( fn, fn_back )
1119 with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
1120 rewrite_comments( fhin, fhout, comments )
1122 os.remove( fn_back )
1123 logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
1124 except (IOError,OSError) as e:
1125 logging.error('File operation failed: %s' % e)
1130 if __name__ == '__main__':
1131 sys.exit( main( sys.argv[1:] ) )