import sys
import os
import re
+import logging
+import getopt
import clang.cindex
return c + self + '\033[m'
+## Comment.
+class Comment:
+
+ def __init__(self, lines, first_line, first_col, last_line, last_col, indent, func):
+ self.lines = lines
+ self.first_line = first_line
+ self.first_col = first_col
+ self.last_line = last_line
+ self.last_col = last_col
+ self.indent = indent
+ self.func = func
+
+ def has_comment(self, line):
+ return line >= self.first_line and line <= self.last_line
+
+ def __str__(self):
+ return "<Comment for %s: [%d,%d:%d,%d] %s>" % (self.func, self.first_line, self.first_col, self.last_line, self.last_col, self.lines)
+
+
+## A data member comment.
+class MemberComment:
+
+ def __init__(self, text, is_transient, array_size, first_line, first_col, func):
+ self.lines = [ text ]
+ self.is_transient = is_transient
+ self.array_size = array_size
+ self.first_line = first_line
+ self.first_col = first_col
+ self.func = func
+
+ def has_comment(self, line):
+ return line == self.first_line
+
+ def __str__(self):
+
+ if self.is_transient:
+ tt = '!transient! '
+ else:
+ tt = ''
+
+ if self.array_size is not None:
+ ars = '[%s] ' % self.array_size
+ else:
+ ars = ''
+
+ return "<MemberComment for %s: [%d,%d] %s%s%s>" % (self.func, self.first_line, self.first_col, tt, ars, self.lines[0])
+
+
+## A dummy comment that removes comment lines.
+class RemoveComment(Comment):
+
+ def __init__(self, first_line, last_line):
+ self.first_line = first_line
+ self.last_line = last_line
+ self.func = '<remove>'
+
+ def __str__(self):
+ return "<RemoveComment: [%d,%d]>" % (self.first_line, self.last_line)
+
+
+## Parses method comments.
+#
+# @param cursor Current libclang parser cursor
+# @param comments Array of comments: new ones will be appended there
+def comment_method(cursor, comments):
+
+ # we are looking for the following structure: method -> compound statement -> comment, i.e. we
+ # need to extract the first comment in the compound statement composing the method
+
+ in_compound_stmt = False
+ expect_comment = False
+ emit_comment = False
+
+ comment = []
+ comment_function = cursor.spelling or cursor.displayname
+ comment_line_start = -1
+ comment_line_end = -1
+ comment_col_start = -1
+ comment_col_end = -1
+ comment_indent = -1
+
+ for token in cursor.get_tokens():
+
+ if token.cursor.kind == clang.cindex.CursorKind.COMPOUND_STMT:
+ if not in_compound_stmt:
+ in_compound_stmt = True
+ expect_comment = True
+ comment_line_end = -1
+ else:
+ if in_compound_stmt:
+ in_compound_stmt = False
+ emit_comment = True
+
+ # tkind = str(token.kind)[str(token.kind).index('.')+1:]
+ # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:]
+
+ if in_compound_stmt:
+
+ if expect_comment:
+
+ extent = token.extent
+ line_start = extent.start.line
+ line_end = extent.end.line
+
+ if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{':
+ pass
+
+ 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)):
+ comment_line_end = line_end
+ comment_col_end = extent.end.column
+
+ if comment_indent == -1 or (extent.start.column-1) < comment_indent:
+ comment_indent = extent.start.column-1
+
+ if comment_line_start == -1:
+ comment_line_start = line_start
+ comment_col_start = extent.start.column
+ comment.extend( token.spelling.split('\n') )
+
+ # multiline comments are parsed in one go, therefore don't expect subsequent comments
+ if line_end - line_start > 0:
+ emit_comment = True
+ expect_comment = False
+
+ else:
+ emit_comment = True
+ expect_comment = False
+
+ if emit_comment:
+
+ comment = refactor_comment( comment )
+
+ if len(comment) > 0:
+ logging.debug("Comment found for function %s" % Colt(comment_function).magenta())
+ comments.append( Comment(comment, comment_line_start, comment_col_start, comment_line_end, comment_col_end, comment_indent, comment_function) )
+
+ comment = []
+ comment_line_start = -1
+ comment_line_end = -1
+ comment_col_start = -1
+ comment_col_end = -1
+ comment_indent = -1
+
+ emit_comment = False
+ break
+
+
+## Parses comments to class data members.
+#
+# @param cursor Current libclang parser cursor
+# @param comments Array of comments: new ones will be appended there
+def comment_datamember(cursor, comments):
+
+ # Note: libclang 3.5 seems to have problems parsing a certain type of FIELD_DECL, so we revert
+ # to a partial manual parsing. When parsing fails, the cursor's "extent" is not set properly,
+ # returning a line range 0-0. We therefore make the not-so-absurd assumption that the datamember
+ # definition is fully on one line, and we take the line number from cursor.location.
+
+ line_num = cursor.location.line
+ raw = None
+ prev = None
+ found = False
+
+ # Huge overkill
+ with open(str(cursor.location.file)) as fp:
+ cur_line = 0
+ for raw in fp:
+ cur_line = cur_line + 1
+ if cur_line == line_num:
+ found = True
+ break
+ prev = raw
+
+ assert found, 'A line that should exist was not found in file' % cursor.location.file
+
+ recomm = r'(//(!)|///?)(\[(.*?)\])?<?\s*(.*?)\s*$'
+ recomm_doxyary = r'^\s*///\s*(.*?)\s*$'
+
+ mcomm = re.search(recomm, raw)
+ if mcomm:
+ member_name = cursor.spelling;
+ is_transient = mcomm.group(2) is not None
+ array_size = mcomm.group(4)
+ text = mcomm.group(5)
+
+ col_num = mcomm.start()+1;
+
+ if array_size is not None and prev is not None:
+ # ROOT arrays with comments already converted to Doxygen have the member description on the
+ # previous line
+ mcomm_doxyary = re.search(recomm_doxyary, prev)
+ if mcomm_doxyary:
+ text = mcomm_doxyary.group(1)
+ comments.append(RemoveComment(line_num-1, line_num-1))
+
+ logging.debug('Comment found for member %s' % Colt(member_name).magenta())
+
+ comments.append( MemberComment(
+ text,
+ is_transient,
+ array_size,
+ line_num,
+ col_num,
+ member_name ))
+
+ else:
+ assert False, 'Regular expression does not match member comment'
+
+
## Traverse the AST recursively starting from the current cursor.
#
# @param cursor A Clang parser cursor
+# @param filename Name of the current file
+# @param comments Array of comments: new ones will be appended there
# @param recursion Current recursion depth
-def traverse_ast(cursor, recursion=0):
+def traverse_ast(cursor, filename, comments, recursion=0):
+
+ # libclang traverses included files as well: we do not want this behavior
+ if cursor.location.file is not None and str(cursor.location.file) != filename:
+ logging.debug("Skipping processing of included %s" % cursor.location.file)
+ return
text = cursor.spelling or cursor.displayname
kind = str(cursor.kind)[str(cursor.kind).index('.')+1:]
for i in range(0, recursion):
indent = indent + ' '
- if cursor.kind == clang.cindex.CursorKind.CXX_METHOD:
+ if cursor.kind == clang.cindex.CursorKind.CXX_METHOD or cursor.kind == clang.cindex.CursorKind.CONSTRUCTOR or cursor.kind == clang.cindex.CursorKind.DESTRUCTOR:
# cursor ran into a C++ method
- print "%s%s(%s)" % (indent, Colt(kind).magenta(), Colt(text).blue())
+ logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
+ comment_method(cursor, comments)
+
+ elif cursor.kind == clang.cindex.CursorKind.FIELD_DECL:
- # we are looking for the following structure: method -> compound statement -> comment, i.e. we
- # need to extract the first comment in the compound statement composing the method
+ # cursor ran into a data member declaration
+ logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
+ comment_datamember(cursor, comments)
+
+ else:
+
+ logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, kind, text) )
+
+ for child_cursor in cursor.get_children():
+ traverse_ast(child_cursor, filename, comments, recursion+1)
- in_compound_stmt = False
- expect_comment = False
- last_comment_line = -1
- for token in cursor.get_tokens():
+## Remove garbage from comments and convert special tags from THtml to Doxygen.
+#
+# @param comment An array containing the lines of the original comment
+def refactor_comment(comment):
- if token.cursor.kind == clang.cindex.CursorKind.COMPOUND_STMT:
- if not in_compound_stmt:
- in_compound_stmt = True
- expect_comment = True
- last_comment_line = -1
+ recomm = r'^(/{2,}|/\*)?\s*(.*?)\s*((/{2,})?\s*|\*/)$'
+ regarbage = r'^[\s*=-_#]+$'
+
+ new_comment = []
+ insert_blank = False
+ wait_first_non_blank = True
+ for line_comment in comment:
+ mcomm = re.search( recomm, line_comment )
+ if mcomm:
+ new_line_comment = mcomm.group(2)
+ mgarbage = re.search( regarbage, new_line_comment )
+ if new_line_comment == '' or mgarbage is not None:
+ insert_blank = True
else:
- if in_compound_stmt:
- in_compound_stmt = False
- break
+ if insert_blank and not wait_first_non_blank:
+ new_comment.append('')
+ insert_blank = False
+ wait_first_non_blank = False
+ new_comment.append( new_line_comment )
+ else:
+ assert False, 'Comment regexp does not match'
+
+ return new_comment
- # tkind = str(token.kind)[str(token.kind).index('.')+1:]
- # ckind = str(token.cursor.kind)[str(token.cursor.kind).index('.')+1:]
- if in_compound_stmt:
+## Rewrites all comments from the given file handler.
+#
+# @param fhin The file handler to read from
+# @param fhout The file handler to write to
+# @param comments Array of comments
+def rewrite_comments(fhin, fhout, comments):
- if expect_comment:
+ line_num = 0
+ in_comment = False
+ skip_empty = False
+ comm = None
+ prev_comm = None
- extent = token.extent
- line_start = extent.start.line
- line_end = extent.end.line
+ rindent = r'^(\s*)'
- if token.kind == clang.cindex.TokenKind.PUNCTUATION and token.spelling == '{':
- pass
+ for line in fhin:
- elif token.kind == clang.cindex.TokenKind.COMMENT and (last_comment_line == -1 or line_start == last_comment_line+1):
- #print Colt("%s %s:%s = %s" % (indent, ckind, tkind, token.spelling)).green()
- last_comment_line = line_end
- new_comment = refactor_comment(token.spelling)
+ line_num = line_num + 1
- for comment_line in new_comment:
- print Colt("%s [%d-%d]" % (indent, line_start, line_end)).green(),
- print Colt(comment_line).cyan()
+ # Find current comment
+ prev_comm = comm
+ comm = None
+ for c in comments:
+ if c.has_comment(line_num):
+ comm = c
- # multiline comments are parsed in one go, therefore don't expect subsequent comments
- if line_end - line_start > 0:
- expect_comment = False
+ if comm:
+ if isinstance(comm, MemberComment):
+ non_comment = line[ 0:comm.first_col-1 ]
+
+ if comm.array_size is not None:
+
+ mindent = re.search(rindent, line)
+ if comm.is_transient:
+ tt = '!'
else:
- expect_comment = False
+ tt = ''
- # else:
- # print Colt("%s %s:%s = %s" % (indent, ckind, tkind, token.spelling)).yellow()
+ # Special case: we need multiple lines not to confuse ROOT's C++ parser
+ fhout.write('%s/// %s\n%s//%s[%s]\n' % (
+ mindent.group(1),
+ comm.lines[0],
+ non_comment,
+ tt,
+ comm.array_size
+ ))
+ else:
- else:
+ if comm.is_transient:
+ tt = '!'
+ else:
+ tt = '/'
- print "%s%s(%s)" % (indent, kind, text)
+ fhout.write('%s//%s< %s\n' % (
+ non_comment,
+ tt,
+ comm.lines[0]
+ ))
- for child_cursor in cursor.get_children():
- traverse_ast(child_cursor, recursion+1)
+ elif isinstance(comm, RemoveComment):
+ # Do nothing: just skip line
+ pass
-## Remove garbage from comments and convert special tags from THtml to Doxygen.
-#
-# @param comment The original comment
-def refactor_comment(comment):
+ elif prev_comm is None:
+ # Beginning of a new comment block of type Comment
+ in_comment = True
- resingle = r'^/{2,}\s*(.*?)\s*(/{2,})?\s*$'
- remulti_first = r'^/\*\s*(.*?)\s*\*?\s*$'
- remulti_last = r'^\s*(.*?)\s*\*/$'
+ # Extract the non-comment part and print it if it exists
+ non_comment = line[ 0:comm.first_col-1 ].rstrip()
+ if non_comment != '':
+ fhout.write( non_comment + '\n' )
- new_comment = comment.split('\n')
+ else:
- if len(new_comment) == 1:
- msingle = re.search(resingle, comment)
- if msingle:
- new_comment[0] = msingle.group(1)
+ if in_comment:
- else:
+ # We have just exited a comment block of type Comment
+ in_comment = False
- for i in range(0, len(new_comment)):
- if i == 0:
- mmulti = re.search(remulti_first, new_comment[i])
- if mmulti:
- new_comment[i] = mmulti.group(1)
- elif i == len(new_comment)-1:
- mmulti = re.search(remulti_last, new_comment[i])
- if mmulti:
- new_comment[i] = mmulti.group(1)
- else:
- new_comment[i] = new_comment[i].strip()
+ # Dump revamped comment, if applicable
+ text_indent = ''
+ for i in range(0,prev_comm.indent):
+ text_indent = text_indent + ' '
- return new_comment
+ for lc in prev_comm.lines:
+ fhout.write( "%s/// %s\n" % (text_indent, lc) );
+ fhout.write('\n')
+ skip_empty = True
+
+ line_out = line.rstrip('\n')
+ if skip_empty:
+ skip_empty = False
+ if line_out.strip() != '':
+ fhout.write( line_out + '\n' )
+ else:
+ fhout.write( line_out + '\n' )
## The main function.
#
-# **Note:** this program only has this function.
+# Return value is the executable's return value.
def main(argv):
+ # Setup logging on stderr
+ log_level = logging.INFO
+ logging.basicConfig(
+ level=log_level,
+ format='%(levelname)-8s %(funcName)-20s %(message)s',
+ stream=sys.stderr
+ )
+
+ # Parse command-line options
+ output_on_stdout = False
+ try:
+ opts, args = getopt.getopt( argv, 'od', [ 'debug=', 'stdout' ] )
+ for o, a in opts:
+ if o == '--debug':
+ log_level = getattr( logging, a.upper(), None )
+ if not isinstance(log_level, int):
+ raise getopt.GetoptError('log level must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL')
+ elif o == '-d':
+ log_level = logging.DEBUG
+ elif o == '-o' or o == '--stdout':
+ logging.debug('Output on stdout instead of replacing original files')
+ output_on_stdout = True
+ else:
+ assert False, 'Unhandled argument'
+ except getopt.GetoptError as e:
+ logging.fatal('Invalid arguments: %s' % e)
+ return 1
+
+ logging.getLogger('').setLevel(log_level)
+
# Attempt to load libclang from a list of known locations
libclang_locations = [
'/usr/lib/llvm-3.5/lib/libclang.so.1',
break
if not libclang_found:
- print Colt('[Error] Cannot find libclang, aborting').red()
+ logging.fatal('Cannot find libclang')
return 1
# Loop over all files
- for fn in argv[1:]:
+ for fn in args:
+ logging.info('Input file: %s' % Colt(fn).magenta())
index = clang.cindex.Index.create()
translation_unit = index.parse(fn, args=['-x', 'c++'])
- traverse_ast( translation_unit.cursor )
+
+ comments = []
+ traverse_ast( translation_unit.cursor, fn, comments )
+ for c in comments:
+
+ logging.debug("Comment found for entity %s:" % Colt(c.func).magenta())
+
+ if isinstance(c, MemberComment):
+
+ if c.is_transient:
+ transient_text = Colt('transient ').yellow()
+ else:
+ transient_text = ''
+
+ if c.array_size is not None:
+ array_text = Colt('arraysize=%s ' % c.array_size).yellow()
+ else:
+ array_text = ''
+
+ logging.debug(
+ "%s %s%s{%s}" % ( \
+ Colt("[%d,%d]" % (c.first_line, c.first_col)).green(),
+ transient_text,
+ array_text,
+ Colt(c.lines[0]).cyan()
+ ))
+
+ elif isinstance(c, RemoveComment):
+
+ logging.debug( Colt('[%d,%d]' % (c.first_line, c.last_line)).green() )
+
+ else:
+ for l in c.lines:
+ logging.debug(
+ Colt("[%d,%d:%d,%d] " % (c.first_line, c.first_col, c.last_line, c.last_col)).green() +
+ "{%s}" % Colt(l).cyan()
+ )
+
+ try:
+
+ if output_on_stdout:
+ with open(fn, 'r') as fhin:
+ rewrite_comments( fhin, sys.stdout, comments )
+ else:
+ fn_back = fn + '.thtml2doxy_backup'
+ os.rename( fn, fn_back )
+
+ with open(fn_back, 'r') as fhin, open(fn, 'w') as fhout:
+ rewrite_comments( fhin, fhout, comments )
+
+ os.remove( fn_back )
+ logging.info("File %s converted to Doxygen: check differences before committing!" % Colt(fn).magenta())
+ except (IOError,OSError) as e:
+ logging.error('File operation failed: %s' % e)
return 0
if __name__ == '__main__':
- sys.exit( main( sys.argv ) )
+ sys.exit( main( sys.argv[1:] ) )