class"
comm_idx = 0
regac = r'\s*%s\s+class\.?\s*' % class_name_doxy
mgac = None
for comm in comment_lines:
mgac = re.search(regac, comm)
if mgac:
break
comm_idx = comm_idx + 1
if mgac:
logging.debug('Removing dumb comment line: {%s}' % Colt(comment_lines[comm_idx]).magenta())
del comment_lines[comm_idx]
comment_lines = refactor_comment(comment_lines, do_strip_html=False, infilename=filename)
# Now we look for a possible \brief
if brief is None:
comm_idx = 0
for comm in comment_lines:
if comm.startswith('\\class') or comm.startswith('\\file') or comm == '':
pass
else:
if len(comm) <= brief_len_threshold:
brief = comm
break
comm_idx = comm_idx + 1
if brief is not None:
comment_lines = refactor_comment(
[ comment_lines[0], '\\brief ' + brief ] + comment_lines[1:comm_idx] + comment_lines[comm_idx+1:],
do_strip_html=False, infilename=filename)
logging.debug('Comment found for class %s' % Colt(class_name_doxy).magenta())
comments.append(Comment(
comment_lines,
start_line, 1, end_line, 1,
0, class_name_doxy
))
else:
logging.warning('No comment found for class %s: creating a dummy entry at the beginning' % \
Colt(class_name_doxy).magenta())
comments.append(PrependComment(
[ file_class_line ],
1, 1, 1, 1,
0, class_name_doxy, append_empty=True
))
## Looks for a special ROOT ClassImp() entry.
#
# Doxygen might get confused by `ClassImp()` entries as they are macros normally written without
# the ending `;`: this function wraps the definition inside a condition in order to make Doxygen
# ignore it.
#
# @param filename Name of the current file
# @param comments Array of comments: new ones will be appended there
def comment_classimp(filename, comments):
recomm = r'^\s*///?(\s*.*?)\s*/*\s*$'
line_num = 0
reclassimp = r'^(\s*)Class(Imp|Def)\((.*?)\).*$'
in_classimp_cond = False
restartcond = r'^\s*///\s*\\cond\s+CLASSIMP\s*$'
reendcond = r'^\s*///\s*\\endcond\s*$'
with open(filename, 'r') as fp:
# Array of tuples: classimp, startcond, endcond
found_classimp = []
# Reset to nothing found
line_classimp = -1
line_startcond = -1
line_endcond = -1
classimp_class = None
classimp_indent = None
for line in fp:
line_num = line_num + 1
mclassimp = re.search(reclassimp, line)
if mclassimp:
# Dump previous one if appropriate, and reset
if line_classimp != -1:
found_classimp.append( (line_classimp, line_startcond, line_endcond) )
line_classimp = -1
line_startcond = -1
line_endcond = -1
# Adjust indent
classimp_indent = len( mclassimp.group(1) )
line_classimp = line_num
classimp_class = mclassimp.group(3)
imp_or_def = mclassimp.group(2)
logging.debug(
'Comment found for ' +
Colt( 'Class%s(' % imp_or_def ).magenta() +
Colt( classimp_class ).cyan() +
Colt( ')' ).magenta() )
else:
mstartcond = re.search(restartcond, line)
if mstartcond:
# Dump previous one if appropriate, and reset
if line_classimp != -1:
found_classimp.append( (line_classimp, line_startcond, line_endcond) )
line_classimp = -1
line_startcond = -1
line_endcond = -1
logging.debug('Found Doxygen opening condition for ClassImp')
in_classimp_cond = True
line_startcond = line_num
elif in_classimp_cond:
mendcond = re.search(reendcond, line)
if mendcond:
logging.debug('Found Doxygen closing condition for ClassImp')
in_classimp_cond = False
line_endcond = line_num
# Dump previous one if appropriate, and reset (out of the loop)
if line_classimp != -1:
found_classimp.append( (line_classimp, line_startcond, line_endcond) )
line_classimp = -1
line_startcond = -1
line_endcond = -1
for line_classimp,line_startcond,line_endcond in found_classimp:
# Loop over the ClassImp conditions we've found
if line_startcond != -1:
logging.debug('Looks like we are in a condition here %d,%d,%d' % (line_classimp, line_startcond, line_endcond))
comments.append(Comment(
['\cond CLASSIMP'],
line_startcond, 1, line_startcond, 1,
classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
append_empty=False
))
else:
logging.debug('Looks like we are NOT NOT in a condition here %d,%d,%d' % (line_classimp, line_startcond, line_endcond))
comments.append(PrependComment(
['\cond CLASSIMP'],
line_classimp, 1, line_classimp, 1,
classimp_indent, 'ClassImp/Def(%s)' % classimp_class
))
if line_endcond != -1:
comments.append(Comment(
['\endcond'],
line_endcond, 1, line_endcond, 1,
classimp_indent, 'ClassImp/Def(%s)' % classimp_class,
append_empty=False
))
else:
comments.append(PrependComment(
['\endcond'],
line_classimp+1, 1, line_classimp+1, 1,
classimp_indent, 'ClassImp/Def(%s)' % classimp_class
))
## 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
# @param in_func True if we are inside a function or method
# @param classdesc_line_limit Do not look for comments after this line
#
# @return A tuple containing the classdesc_line_limit as first item
def traverse_ast(cursor, filename, comments, recursion=0, in_func=False, classdesc_line_limit=None):
# 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:]
is_macro = filename.endswith('.C')
indent = ''
for i in range(0, recursion):
indent = indent + ' '
if cursor.kind in [ clang.cindex.CursorKind.CXX_METHOD, clang.cindex.CursorKind.CONSTRUCTOR,
clang.cindex.CursorKind.DESTRUCTOR, clang.cindex.CursorKind.FUNCTION_DECL ]:
if classdesc_line_limit is None:
classdesc_line_limit = cursor.location.line
# cursor ran into a C++ method
logging.debug( "%5d %s%s(%s)" % (cursor.location.line, indent, Colt(kind).magenta(), Colt(text).blue()) )
comment_method(cursor, comments)
in_func = True
elif not is_macro and not in_func and \
cursor.kind in [ clang.cindex.CursorKind.FIELD_DECL, clang.cindex.CursorKind.VAR_DECL ]:
if classdesc_line_limit is None:
classdesc_line_limit = cursor.location.line
# 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():
classdesc_line_limit = traverse_ast(child_cursor, filename, comments, recursion+1, in_func, classdesc_line_limit)
if recursion == 0:
comment_classimp(filename, comments)
comment_classdesc(filename, comments, classdesc_line_limit)
return classdesc_line_limit
## Strip some HTML tags from the given string. Returns clean string.
#
# @param s Input string
def strip_html(s):
rehtml = r'(?i)?(P|BR)/?>'
return re.sub(rehtml, '', s)
## 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, do_strip_html=True, infilename=None):
recomm = r'^(/{2,}|/\*)? ?(\s*)(.*?)\s*((/{2,})?\s*|\*/)$'
regarbage = r'^(?i)\s*([\s*=_#-]+|(Begin|End)_Html)\s*$'
# Support for LaTeX blocks spanning on multiple lines
relatex = r'(?i)^((.*?)\s+)?(BEGIN|END)_LATEX([.,;:\s]+.*)?$'
in_latex = False
latex_block = False
# Support for LaTeX blocks on a single line
reinline_latex = r'(?i)(.*)BEGIN_LATEX\s+(.*?)\s+END_LATEX(.*)$'
# Match (to turn it into the ~~~ Markdown syntax)
reblock = r'(?i)^(\s*)?PRE>\s*$'
# Macro blocks for pictures generation
in_macro = False
current_macro = []
remacro = r'(?i)^\s*(BEGIN|END)_MACRO(\((.*?)\))?\s*$'
# Minimum indent level: scale back everything
lowest_indent_level = None
# Indentation threshold: if too much indented, don't indent at all
indent_level_threshold = 7
new_comment = []
insert_blank = False
wait_first_non_blank = True
for line_comment in comment:
# Strip some HTML tags
if do_strip_html:
line_comment = strip_html(line_comment)
mcomm = re.search( recomm, line_comment )
if mcomm:
new_line_comment = mcomm.group(2) + mcomm.group(3) # indent + comm
# Check if we are in a macro block
mmacro = re.search(remacro, new_line_comment)
if mmacro:
if in_macro:
in_macro = False
# Dump macro
outimg = write_macro(infilename, current_macro) + '.png'
current_macro = []
# Insert image
new_comment.append( '![Picture from ROOT macro](%s)' % (os.path.basename(outimg)) )
logging.debug( 'Found macro for generating image %s' % Colt(outimg).magenta() )
else:
in_macro = True
continue
elif in_macro:
current_macro.append( new_line_comment )
continue
mgarbage = re.search( regarbage, new_line_comment )
if mgarbage is None and not mcomm.group(3).startswith('\\') and mcomm.group(3) != '':
# not a special command line: count indent
indent_level = len( mcomm.group(2) )
if lowest_indent_level is None or indent_level < lowest_indent_level:
lowest_indent_level = indent_level
# if indentation level is too much, consider it zero
if indent_level > indent_level_threshold:
new_line_comment = mcomm.group(3) # remove ALL indentation
if new_line_comment == '' or mgarbage is not None:
insert_blank = True
else:
if insert_blank and not wait_first_non_blank:
new_comment.append('')
insert_blank = False
wait_first_non_blank = False
# Postprocessing: LaTeX formulas in ROOT format
# Marked by BEGIN_LATEX ... END_LATEX and they use # in place of \
# There can be several ROOT LaTeX forumlas per line
while True:
minline_latex = re.search( reinline_latex, new_line_comment )
if minline_latex:
new_line_comment = '%s\\f$%s\\f$%s' % \
( minline_latex.group(1), minline_latex.group(2).replace('#', '\\'),
minline_latex.group(3) )
else:
break
# ROOT LaTeX: do we have a Begin/End_LaTeX block?
# Note: the presence of LaTeX "closures" does not exclude the possibility to have a begin
# block here left without a corresponding ending block
mlatex = re.search( relatex, new_line_comment )
if mlatex:
# before and after parts have been already stripped
l_before = mlatex.group(2)
l_after = mlatex.group(4)
is_begin = mlatex.group(3).upper() == 'BEGIN' # if not, END
if l_before is None:
l_before = ''
if l_after is None:
l_after = ''
if is_begin:
# Begin of LaTeX part
in_latex = True
if l_before == '' and l_after == '':
# Opening tag alone: mark the beginning of a block: \f[ ... \f]
latex_block = True
new_comment.append( '\\f[' )
else:
# Mark the beginning of inline: \f$ ... \f$
latex_block = False
new_comment.append(
'%s \\f$%s' % ( l_before, l_after.replace('#', '\\') )
)
else:
# End of LaTeX part
in_latex = False
if latex_block:
# Closing a LaTeX block
if l_before != '':
new_comment.append( l_before.replace('#', '\\') )
new_comment.append( '\\f]' )
if l_after != '':
new_comment.append( l_after )
else:
# Closing a LaTeX inline
new_comment.append(
'%s\\f$%s' % ( l_before.replace('#', '\\'), l_after )
)
# Prevent appending lines (we have already done that)
new_line_comment = None
# If we are not in a LaTeX block, look for tags and transform them into Doxygen code
# blocks (using ~~~ ... ~~~). Only tags on a single line are supported
if new_line_comment is not None and not in_latex:
mblock = re.search( reblock, new_line_comment )
if mblock:
new_comment.append( mblock.group(1)+'~~~' )
new_line_comment = None
if new_line_comment is not None:
if in_latex:
new_line_comment = new_line_comment.replace('#', '\\')
new_comment.append( new_line_comment )
else:
assert False, 'Comment regexp does not match'
# Fixing indentation level
if lowest_indent_level is not None:
logging.debug('Lowest indentation level found: %d' % lowest_indent_level)
new_comment_indent = []
reblankstart = r'^\s+'
for line in new_comment:
if re.search(reblankstart, line):
new_comment_indent.append( line[lowest_indent_level:] )
else:
new_comment_indent.append( line )
new_comment = new_comment_indent
else:
logging.debug('No indentation scaling applied')
return new_comment
## Dumps an image-generating macro to the correct place. Returns a string with the image path,
# without the extension.
#
# @param infilename File name of the source file
# @param macro_lines Array of macro lines
def write_macro(infilename, macro_lines):
# Calculate hash
digh = hashlib.sha1()
for l in macro_lines:
digh.update(l)
digh.update('\n')
short_digest = digh.hexdigest()[0:7]
infiledir = os.path.dirname(infilename)
if infiledir == '':
infiledir = '.'
outdir = '%s/imgdoc' % infiledir
outprefix = '%s/%s_%s' % (
outdir,
os.path.basename(infilename).replace('.', '_'),
short_digest
)
outmacro = '%s.C' % outprefix
# Make directory
if not os.path.isdir(outdir):
# do not catch: let everything die on error
logging.debug('Creating directory %s' % Colt(outdir).magenta())
os.mkdir(outdir)
# Create file (do not catch errors either)
with open(outmacro, 'w') as omfp:
logging.debug('Writing macro %s' % Colt(outmacro).magenta())
for l in macro_lines:
omfp.write(l)
omfp.write('\n')
return outprefix
## 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):
line_num = 0
in_comment = False
skip_empty = False
comm = None
prev_comm = None
restore_lines = None
rindent = r'^(\s*)'
def dump_comment_block(cmt, restore=None):
text_indent = ''
ask_skip_empty = False
for i in range(0, cmt.indent):
text_indent = text_indent + ' '
for lc in cmt.lines:
fhout.write('%s///' % text_indent )
lc = lc.rstrip()
if len(lc) != 0:
fhout.write(' ')
fhout.write(lc)
fhout.write('\n')
# Empty new line at the end of the comment
if cmt.append_empty:
fhout.write('\n')
ask_skip_empty = True
# Restore lines if possible
if restore:
for lr in restore:
fhout.write(lr)
fhout.write('\n')
# Tell the caller whether it should skip the next empty line found
return ask_skip_empty
for line in fhin:
line_num = line_num + 1
# Find current comment
prev_comm = comm
comm = None
comm_list = []
for c in comments:
if c.has_comment(line_num):
comm = c
comm_list.append(c)
if len(comm_list) > 1:
merged = True
if len(comm_list) == 2:
c1,c2 = comm_list
if isinstance(c1, Comment) and isinstance(c2, Comment):
c1.lines = c1.lines + c2.lines # list merge
comm = c1
logging.debug('Two adjacent comments merged. Result: {%s}' % Colt(comm).cyan())
else:
merged = False
else:
merged = False
if merged == False:
logging.warning('Too many unmergeable comments on the same line (%d), picking the last one' % len(comm_list))
for c in comm_list:
logging.warning('>> %s' % c)
comm = c # considering the last one
if comm:
# First thing to check: are we in the same comment as before?
if comm is not prev_comm and \
isinstance(comm, Comment) and \
isinstance(prev_comm, Comment) and \
not isinstance(prev_comm, RemoveComment):
# We are NOT in the same comment as before, and this comment is dumpable
skip_empty = dump_comment_block(prev_comm, restore_lines)
in_comment = False
restore_lines = None
prev_comm = None # we have just dumped it: pretend it never existed in this loop
#
# Check type of comment and react accordingly
#
if isinstance(comm, MemberComment):
# end comment block
if in_comment:
skip_empty = dump_comment_block(prev_comm, restore_lines)
in_comment = False
restore_lines = None
non_comment = line[ 0:comm.first_col-1 ]
if comm.array_size is not None or comm.is_dontsplit() or comm.is_ptr():
# This is a special case: comment will be split in two lines: one before the comment for
# Doxygen as "member description", and the other right after the comment on the same line
# to be parsed by ROOT's C++ parser
# Keep indent on the generated line of comment before member definition
mindent = re.search(rindent, line)
# Get correct comment flag, if any
if comm.comment_flag is not None:
cflag = comm.comment_flag
else:
cflag = ''
# Get correct array size, if any
if comm.array_size is not None:
asize = '[%s]' % comm.array_size
else:
asize = ''
# Write on two lines
fhout.write('%s/// %s\n%s//%s%s\n' % (
mindent.group(1),
comm.lines[0],
non_comment,
cflag,
asize
))
else:
# Single-line comments with the "transient" flag can be kept on one line in a way that
# they are correctly interpreted by both ROOT and Doxygen
if comm.is_transient():
tt = '!'
else:
tt = '/'
fhout.write('%s//%s< %s\n' % (
non_comment,
tt,
comm.lines[0]
))
elif isinstance(comm, RemoveComment):
# End comment block and skip this line
if in_comment:
skip_empty = dump_comment_block(prev_comm, restore_lines)
in_comment = False
restore_lines = None
elif restore_lines is None:
# Beginning of a new comment block of type Comment or PrependComment
in_comment = True
if isinstance(comm, PrependComment):
# Prepare array of lines to dump right after the comment
restore_lines = [ line.rstrip('\n') ]
logging.debug('Commencing lines to restore: {%s}' % Colt(restore_lines[0]).cyan())
else:
# Extract the non-comment part and print it if it exists. If this is the first line of a
# comment, it might happen something like `valid_code; // this is a comment`.
if comm.first_line == line_num:
non_comment = line[ 0:comm.first_col-1 ].rstrip()
if non_comment != '':
fhout.write( non_comment + '\n' )
elif isinstance(comm, Comment):
if restore_lines is not None:
# From the 2nd line on of comment to prepend
restore_lines.append( line.rstrip('\n') )
logging.debug('Appending lines to restore. All lines: {%s}' % Colt(restore_lines).cyan())
else:
assert False, 'Unhandled parser state: line=%d comm={%s} prev_comm={%s}' % \
(line_num, comm, prev_comm)
else:
# Not a comment line
if in_comment:
# We have just exited a comment block of type Comment
skip_empty = dump_comment_block(prev_comm, restore_lines)
in_comment = False
restore_lines = None
# Dump the non-comment line
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' )
# Is there some comment left here?
if restore_lines is not None:
dump_comment_block(comm, restore_lines)
# Is there some other comment beyond the last line?
for c in comments:
if c.has_comment(line_num+1):
dump_comment_block(c, None)
break
## The main 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
include_flags = []
try:
opts, args = getopt.getopt( argv, 'odI:', [ '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':
output_on_stdout = True
elif o == '-I':
if os.path.isdir(a):
include_flags.extend( [ '-I', a ] )
else:
logging.fatal('Include directory not found: %s' % Colt(a).magenta())
return 2
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',
'/usr/lib/libclang.so',
'/Library/Developer/CommandLineTools/usr/lib/libclang.dylib'
]
libclang_found = False
for lib in libclang_locations:
if os.path.isfile(lib):
clang.cindex.Config.set_library_file(lib)
libclang_found = True
break
if not libclang_found:
logging.fatal('Cannot find libclang')
return 1
# Loop over all files
for fn in args:
logging.info('Input file: %s' % Colt(fn).magenta())
index = clang.cindex.Index.create()
clang_args = [ '-x', 'c++' ]
clang_args.extend( include_flags )
translation_unit = index.parse(fn, args=clang_args)
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():
flag_text = Colt('transient ').yellow()
elif c.is_dontsplit():
flag_text = Colt('dontsplit ').yellow()
elif c.is_ptr():
flag_text = Colt('ptr ').yellow()
else:
flag_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(),
flag_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[1:] ) )